Good quality LM723 LAB Power Supply

[blackdog]

Hi, 

I found in my bin with transformers a type that is 20V at 3.5-Ampere.
And I thought this transformer could become the basis of some explanation of what is all important for a good LAB Power Supply.
A lot of information that I find important in Power Supply's is often not found in designs on the Internet or only partially.
And I would like to share this information with you.
There are many similarities with other power supply's, even from a totaly different setup, such as the Harrison and HP schematics that I prefer.
So this schematic / project is mainly meant to explain what is involved in developing a good LAB Power Supply.

If i'am wrong about my explanation or i have to do it better, please tel me!

First I want to say this, this is a project in development, and meanwhile I have done already a lot of measurements on the schematic below.
This Power Supply will not be a "One Size Fits All" Power Supply.
So don't ask me if is is posible to change it to 30V 5-Ampere, or 60V at 2-Ampere... you are on your own then. 
Processor controlled, please don't ask, just two 10 turn Model 534 potmeters for setting U and I. and a switch for enable.

The basis of this Power Supply is a standard LM723 circuit with parts around it, that make's it a good Power Supply. (my opinion)

Preliminary Specifications
The Output current wil be 2-Ampere Max.
Max output voltage 20V.
Current control from say 50mA to peak 2.2-Ampere, no sharp current limiting knee! this is not posible with one base emitor voltage.
DC Ri < 0.002 Ohm, at the moment around 0.001 Ohm.
Small output capacitor 100uF preferred or 150uF, will depend on the final measurement and the capacitor quality.
Dynamic properties, at the moment between 55 and 65 phase margin, have tot do more testing on that later.
Picture of the dyamic behaviour will come later.

KIND POWER SUPPLY!
What this means? there are so many professional Power Supply's that have Power On and Power Off abbarations and i downt like that!

I have done my best so that the output of this Power Supply never gives more than 0.2V abberation at power on/off or the enable switch.
This only happens just before the power supply is coming on and when the Power Supply is already off for about five seconds.
I will do my best to make these abberations even smaller, that's where most of the work is done.
It is a combination of the lowest possible noise behaviour on the output and the timing to keep the abberations as clean as possible.
At the moment it is almost perfect, and i mean enable/off is perfect for small and large output currents.
Switching to power supply off with al kind of loads, is perfect.
There is a very small abberation when powering on, about 0.25V, that is no problem, but i will do work some more on that part.

The explanation of the schematic below
Keep in mind that the schematic is not yet finished and will certainly change.



Let's start with the transformer, so this is sucandair 20V at 3.5-Ampere.
This gives at double-phase rectification a little more than 2-Ampere to output current without overloading the transformer.
Usually you take about 60 to 65% of the current that the transformer can supply to a resistor or lamp.
By rectifying, together with the buffer capacitor, the maximum current without overloading the transformer in continuous operation is about 60 to 65%.
This is because the transformer has to supply large peak currents to charge the capacitor as well as the current that draws the load.
These peak currents make the i2r losses in the transformer higher than with a load such as a lamp or an ordinary resistor without bridge rectifier and capacitor.

R1 and C2 is a Snubber circuit, these components remove the ringing of the transformer when the diodes come out of conductance below say 0.5V.
The transformer sees suddenly no load anymore and will ring, (The induction and the winding capacity makes a small resonator circuit)
When you wil pull the power plug or switch the power supply off, the transformer will also "ring" R1 and C2 wik kill this.
I used a standard component of the brand WIMA.

The transformer also is used to make aditional voltages.
D4 and C7 makes a "clean" Power supply for the LM723 and the current source around Q4 that is used for the current limiting.
The ripple on C7 is far smaller than the ripple on C4 and also the DC level is higer as say 2-Ampere load,
so if de mains voltage is not to low, i can still have a clean 20V output at 2-Ampere current.
R5 help's with keeping fast pulses from the mains out of this power rail, Ri of the tranformer + R5 and C7 makes a 30Hz Low Pass filter. 

The bridgerectifier is a 15Amp Low VF Glass Passivated type.
At 7,5-Ampere peak current the Vf is 0.9V, and this is low, it kan even be better with Schottky types, but for me this is good eneugh.

C1, C3, D1 and D2 are making a negative supply rail, also R2 is here used for some filtering, the real value of it comes later, for now it is 1 Ohm.

-------
End 6-1-2019
-------

The time is up for today, tomorrow I will make some mistakes in the text, English is not my mother tongue.
I will keep updating this first post as I continue to explain this schematic.

I would like to hear your comments.

Kind regards,
Bram

[BravoV]

... English is not my mother tongue.

Bram, as non native English speaker, don't worry, your's is just fine and fully understandable  , I've seen much-much worst from others.

Subscribed, as this thread reminds me of a similar sized step-down transformer that is gathering dust, really tempting, I might join this project.

Thanks for sharing.


Edit : What does LED1 do ? As constant current mode indicator ?

[iMo]

I like 723 PSUs What needs some elaboration is the more precise current protection setting, afaik.

[blackdog]

Hi,

BravoV
Later i will explane the LED in the current source, but yes you can use it as a Power.

imo
I know the "problem" with current limiting, but most of the time you don't need precision curent limiting. (sharp knee)
What I mentioned before, I find an impeccable behavior when switching on and off, fast current limitation and relatively small output capacitor more important.

It is the same with Power Supply's as with DMM's, often you need three to four with different specifications.
This Power Supply has one more good feature and that is that the hum and noise is less than 50uV at full load.
This of course depends on good wiring technique and the placement of the transformer.
So this is not the best Power Supply I can make, but made with readily available parts and kind for its load.
And of course not to forget, a single secundary transformer winding, everybody is loving this! 

And I like to hear from readers, why in the setup of this Power Supply that I have chosen, it is not possible to upgrade it to say 30V output voltage.

If my headache has dropped a bit today, I will post a modified schematic with some measurement data.

Kind regards,
Bram

[glarsson]

The comment on R19 "Direct on Base Q4" should probably refer to Q5.

[blackdog]

Hi glarsson,

Good find!
Thanks.

Bram

[blackdog]

Hi,

It's time for eye candy!

These are the voltage levels on the important power supply lines.
Yellow is the output of the power supply.
Pink is the voltage on the buffer capacitor C4.
Blue is the voltage on the buffer capacitor C3 and it has a negative value.
This value is measured at a value of C4 of 5200uF and a low mains voltage of 225V, which should normally be around 230V.

At the very left side you can see where the "0" level of the trace is located.
In the blue trace this is 1 division from above.
On the yellow and pink trace this is 1 division from the bottom.

These are the levels at 3.3V output voltage and a light load of around 15mA.



These are the levels at 3.3V output voltage and a load of 1.9-Ampere.



And for this post the last picture.
These are the levels at 20V output voltage and a load of 1.9-Ampere.


Kind regards,
Bram

[Kleinstein]

There are few point's I don't like very much about the circuit:

The obvious point is the constant current sink with an LM317 at the output - this is somewhat mixed up, but easy to fix. It may also reach the voltage limit of the 317.

The 723 gets quite a high supply. The good point at this is the extra filtering to avoid the ripple drop. 
So the circuit is limited to a relatively low maximum output voltage. I don't see a reason to give it so much (-5 V) negative voltage. 2.5 V should be plenty, as the inputs on the 723 can work down to about 1-1.5 V. I would guess one could reduce C1 quite a bit in size and this way reduce the power loss at the negative voltage regulation.  It's better to drop the voltage over the capacitor than resistors.

The reference use is a combination of the 723 internal and the TL431.

The power off protection looks a little odd, especially if one tries to charge a battery.

I know it is relatively common, especially in older power supply circuits, but I don't like the voltage adjustment with the feedback divider: this changes the loop gain with the set voltage and thus the compensation has to be a compromise between too much ringing at low voltage and sluggish regulation at high voltages.

The current source around Q4 is relatively temperature stable, but the voltage to compare too depends on the chip temperature. This makes the current setting at low currents rather temperature dependent. It would be slightly better with a PAT current source. Anyway the current limit is not really stable, as it's relative to the VBE voltage.

The 723 is a good voltage regulator, but like most other voltage regulators a poor choice for building a good lab supply. Voltage regulators are different from in lab supply in that they only need crude current limit and the compensation is made for a reasonably well defined load, while a lab supply should be stable with any reasonable passive (could be realized with an RLC combination) load.

[blackdog]

Hi,

Some more pictures.

These pictures are about the abberations of the output voltage of this power supply.
The Enable switch is at the moment fine.
But the power switch or pulling the plug, need some extra work, I already have some ideas to change this.

The schematic is changed, it has now version number 0.2

OK lets look at the abberations or the absence thereof.
I start wih the Enable switch, I put the scope on single shot and move the tumbler switch quickly back and forth to be able to display a few periods.

3.3V 15ma Enable switch used.



3.3V 15ma Enable switch used.



20V-15mA-Enable-On-Off.



20V-15mA-Enable-On-Off



20V-1.9A-Power-On.



3.3V-1.9A-Power On-a.



3.3V-1.9A-Power On-b.



3.3V-15mA-Power On-a.



3.3V-15mA-Power On-b.


So i have stil to do some work on these functions, but first build a new circuitboard, because the one i have now is not so nice anymore 

Kleinstein
Thank for your remarks!

I wil do some testing with the TL431 at 2.495V setting to see whether this voltage is enough for the LM723 to function properly.
This save's a capacitor and two resistors and less noise that I have to filter out...

With my transformer I don't get above 36V as total power supply for the LM723 at about 235V mains voltage.

I already indicated in the beginning that this is a power supply project for me and I hope others can learn something from it if they want to.
I show here how I approach the parts of the power supply and what all influence each other.
This is a project to use the parts I had for a "nice" LAB power supply.
And the second reason is to use this project to explain what's involved in designing a power supply.
For the beginners: A Lab power supply is not simple to design!

Battery charging
The battery charging must be done via a diode, just like testing an LED ALWAYS via a series resistor!

I think everyone should know how their measuring instruments work, so what is and isn't possible.
But I know that a lot of people who work with electronics have no idea what the specifications of their measuring equipment is.
I have more than 50 years of experience in electronics, and have seen several colleagues do stupid things, and occasionally I also contribute to that.
It becomes quite complex to design a power supply(and other instruments) that can deal with all the laziness and abberations of humanity. 

Setting the output voltage.
You are right, the stability/phase margin is easier to set if the gain of the power supply is at a fixed value.
I will  test a TL431 and an LM723 on a breadboard to see if I can adjust the schematic so that I can vary the reference voltage on the +input of the LM723.
And a fixed divider for the -input.

Current limiting
I am myself aware of the temperature drift of the current limitation.
This too I have indicated several times that this is not optimal.
I do not see it as a big problem, this is not a "precision" current controlled power supply.
If I want something like that, I build one according to the Harrison concept.

Let's look at it from the other side, e.g. the Rigol DP832 Power Supply.
This power supply has 470uF over its outputs.
You can set the current to 5mA, but if something goes wrong, the energy from the 470uF capacitor will be dumped into the D.U.T..
It is not much use then to set the current limit to 5mA.
The DP823 set to 12V and say 5mA will always break or damage your LED if no series resistor is applied.
It can help by using the enable button on this instrument, but most people are just lazy. 

Kind regard,
Bram

[schmitt trigger]

Bram;
I have followed many of your posts in this forum. You take the time to properly document and explain your imaginative designs.
And you don't become upset when people asks you questions.

So...I am going to ask mine: I see what you are attempting to do with C1, a capacitive divider.

But would C1 actually see the proper polarity voltage? I know there ill be significant AC ripple. But my question is whether the voltage across C1 *could* become negative

[Kleinstein]

The voltage at C1 is Ok. Under normal operation there is always a significant DC bias. Only during turn on there might be a slight chance to get a small (e.g. 0.5 V) negative voltage. C1 is more like part of a capacitor supply or charge pump circuit.


For the tests it might be interesting to note that the critical case for voltage regulation is with a relatively large low ESR capacitor at the output.
As this makes fast regulation difficult, one does not need to get fast response and low overshoots in this case - but it should be stable.

For the regulator circuit there are mainly two types of circuit, with different properties:
1) The regulators similar to the one shown here, with a low output impedance output stage (e.g. darlington as emitter-follower). These usually give easy voltage regulation and can get away with a small capacitor at the output. However it gets tricky with a voltage higher than some 20-30 V and the current regulation is more tricky and sometimes rather slow. If one does not need very fast regulation the voltage loop can be rather simple and slow, as the output stage already gives a kind of baseline performance also for the fast part.

2) A regulator circuit with an output stage that primarily sets the current (e.g. a transistor in Emitter circuit) and thus high output impedance. A common implementation is the classical circuit with a floating supply for the regulator. These regulators need an capacitor at the output and in more slow implementations it may need to be quite large. The advantage, especially of the floating regulator circuit is that it's very flexible to work to higher voltage if needed and the current regulation it easier and thus often better (except for the output capacitor). However the voltage regulation is more tricky and needs a more involved compensation to be fast and thus alow a relatively small output capacitor. This type of circuit is usually found in lab supplies.

[IconicPCB]

As drawn, LM317 and associated 47ohm resistor will NOT act as a constant current sin/source. The resistor must be inline with the current flow.

[iMo]

LM317 and 337 always keep 1.25V between Out and ADJ. So for const current source you feed current into INP and the output is the ADJ, the const current is 1.25/R + few uAmps of ADJ.
Mind the 317 and 337 need a MIN input voltage to work properly.
PS: people usually filter the 723's Vref with 10-100uF (from mid of the divider to gnd).

[IconicPCB]

Please re asses Your understanding of LM317 operation.

The way the schematic is drawn the voltage difference of 1.25 volts does not come into the reckoning.

Edit:  the resistor should be connected between the OUT and ADJUST lines and the current should be taken through the resistor at ADJUST pin anyways LM317 is arse about with respect to circuit polarities.

[blackdog]

Hi,

Schematic version
The schematic has been given a new version number and I'm going to try today if controlling at the positive input of the LM723 goes as I want.

Current Source.
Concerning the LM317 current source, don't bite my head off, it's just a schematic error, it is working fine in my test setup. 

Reference voltage.
I tested yesterday with an LM336-2.5V reference, but when testing the DC Ri it turned out to be worse.
Maybe the open loop gain will be lower as the inputs of the LM723 go to ground level.
Maybe because the current source in the emitters of the difference amplifier in the LM723 gets more difficult, because of the low voltage it has to work on.

To find it out exactly that, I should do more measurements on this.
But for now I will leave the reference voltage at 4.95V with the TL431b.
There is a good chance that R3, R4 and D3 will be replaced by a LM317 current source, and I will try to do it right the first time. *grin*

Later today some more information.

Kind regards,
Bram

[not1xor1]

Bram;
I have followed many of your posts in this forum. You take the time to properly document and explain your imaginative designs.
And you don't become upset when people asks you questions.

So...I am going to ask mine: I see what you are attempting to do with C1, a capacitive divider.

But would C1 actually see the proper polarity voltage? I know there ill be significant AC ripple. But my question is whether the voltage across C1 *could* become negative

I doubt it would be correct to call it a capacitive divider. It is just a classical charge pump.
When the transformer output connected to C1 is positive, C1 is charged via D1. When the transformer output gets negative C1 charge is shared with C3 via D2.
After a few cycles both capacitors share approximately the same voltage, i.e. about the same of the rectified positive voltage.
Of course in this case the frequency is just 50Hz not 100.

[blackdog]

Hi,

Here is an extra picture related to the uncertainty regarding the negative power line and the ripple frequencies.
The frequency counter in the lower left corner of the image is linked to C4 (channel-3) and indicates 100Hz as I am in Amsterdam, the Netherlands.
A large part of Europe has 230V and 50Hz as mains frequency properties.
The C3 ripple is built on from half periods, and that is also clearly visible here.




There's a chance I'll make this rectifier as simple as the extra rectifier for the + power supply of the LM723.
I do this together with a power source for the TL431b, so I can be sure that the 4.95V is clean and does not contain any power frequency components.
The high impedance of the LM317 current source together with the low Ri of the TL431b ensure a very high suppression of these signals.

We will see what my tests show, there are so many options to choose from...   

Kind regards,
Bram

[IconicPCB]

Hey Blackdog,

I qualified my comment with an "as drawn " statement.

My second response was in response to the counter claim,

There was no intent of malice in either of my posts

[blackdog]

Hi IconicPCB,

No problemo, i ask for remarks about the schematic or design. 

Sometimes an error can be in a schematic for days and the test circuit works just fine, this because I only drew it wrong.
I regularly ask on forums to look at my schematic for errors, I am not mister knows it all.

I also like humour, it is only quite difficult to show this on forums.
Furthermore, there are also quite large cultural differences between the different countries, so it can be tricky.  :

At the moment, my business administration takes up a lot of time, I don't know if it's possible to do another post here today.

But I am open for suggestions.

Kind regards
Bram

[IconicPCB]

No worries, i think all is good.

[blackdog]

Hi,

I am still very busy with work, so not much time to post here.

Negative power rail
I have already done several measurements on other parts of the design and as a result of this I changed the reference section.
This one now consists of a LM337 that is noise reduced and the accompanying diodes for protection are also added.

One of the reasons the TL431b has been replaced is that it was difficult to find a good setpoint that works sufficiently well with varying mains voltage.
After some measurements, an LM337 (preferably one of the ST brand) appeared to have a good noise behaviour which in the setup as shown here in the schematic has a lower noise than the TL431b.
The DC stability of the ST LM337 that I am testing now, can be called good after more than 12 hours.

Power section
I have selected two transistors from my stock, and I reached an Hfe of around 32000.
Because of this very high Hfe I'm going to try if the DC stability remains sufficient if I don't use the normal output(10) but the compensation output(13).
The main reason is to see if the phase margin gets better and the compensation capacitor can get smaller.

The second reason is that I have a BE level less and the dropout voltage will be lower.
I will have to test the above assumptions to see if they work out positively in practice.

Schematic version
Keep in mind that also this schematic is not yet finished and will certainly change at a later moment.
This topic is meant to describe the design process of a LM723 power supply and not a complete power supply kit.


The time is now up...

Please shoot at it 

Kind regards,
Bram

[Cliff Matthews]

Subscribed    Black Dog clean specs.. 20v @ 2 ish amps.. I'm in!

I have 6 candidates showing gains from 78 to 97 to sacrifice.. can these 2SA1695's join the fun?

[Kleinstein]

For the loop stability the Hfe of the output stage should not be that important. This would be more about voltage gain, which is close to 1 and the stability of the inner loop formed by Q4 and Q6. The inner loop stability could be adjusted with R25.  I don't thing it is a good idea to drive the output from pin 13 of the 723: usually a Sizlaki output stage wants to be drive from a low impedance source. So already R22 might not really help. A base to emitter resistor at Q4 could help to give some miniumum load to the 723.

It looks like there is something wrong with the negative supply generation: somehow C1  and D1 from the old circuit got lost, but they are needed, though C1 could likely be much smaller.  With the new circuit It looks even worse to have a high negative reference, as the effective reference now is the difference from the 7 V inside the 723 and the external 4.9 V reference. External 1.5 V should be better.

My preferred way would be not not use a negative auxiliary supply for the voltage regulation and instead add a fraction of the 7 V reference to the -input of the 723. So an extra resistor between pins 6 and 4 of the 723. The negative supply would still be needed for the constant current minimum load. This does not need to be that stable, so a simple 2 transistor or similar circuit should be sufficient.

For the current setting, I think the simple 2 transistor current source with it's TC could be better than the more stable LED + transistor version.

[Cliff Matthews]

25ma current source: Any idea why the 317HV was specified?
Could a a jelly-bean LM317 in series with a 6.2v zener do the same job?

[SK_Caterpilar_SK]

I have different schematic. Its a old lab powersupply for school purposes so basically idiot proof. No voltage spikes on startups no anything unwanted. Tested because I have the original unit which is performing great. The whole thing is nearly flawless. In its original config it uses the metal can variant of the 723 but I rewrote the shematic so that the pinout is compatible with the DIP version. Also uses some obsoltete transistors and zenners and diodes. I made a PCB layout for it so I can have another one to my desk. The original goes from true 0 to 20V. The only issue with it so far is that it requires a dual winding tranformer. I got the voltages and components so I can give you a list if youre interested to test it at all. It does 0 to 1A but I modified mine so it can do 30V max and 3 amps..

Soo if  youre just interested at all in testing it. I will get you the infos so you can do science on it if you want. I will send you the full image if you want because the max filezise is tiny for this to work properly.

[2N3055]

Tesla, right? That brings back memories...

[blackdog]

Hi,

Every time I'm surprised at what I write and how it is interpreted by readers. 
That makes me think even more about how I put it here why I make certain choices.
And of course what I finally type into the translation machine "Deepl" and what it then makes of it.

Kleinstein
I use the Sziklai pair output stage with selected transistors to get the highest possible current gain.
This is to avoid using the emitter follower in the IC, it will give some delay according to my assumption.
Because of the very high Hfe ot the Sziklai pair, the differential amplifier in the LM723 will not be loaded too much and the loopgain will not drop too much.
And yes, I still have to test this to see if it really works well enough, also to see if the Sziklai pair is happy with the relativ high impedance on pin-13
That's why two connections are drawn in the schematic for controlling the Sziklai pair output stage, pin 10 (normal output) or pin-13 the compensation connection.


Sziklai pair
On the Internet it is often found that the Sziklai pair is unstable, I always think, designer then you have not set them up properly...
The 1.5 to 5.6 Ohm resistor in the emitter of the driver transistor is there for good reason.
Is a Ft of the transistors of around 10MHZ sufficient? I think these are also very useful.
But test if the combination you choose is stable!
Play with the emitor resistor of the driver transistor and ALWAYS directly at the base of the driver transistor a resistor between 22 and 150 Ohm, test this well!
A lower Ft of the power section makes room for longer wires if necessary, but.... test, test, test.

If you use fast transistors like I use them, spaghetti wiring is out of the question!
When using fast components, you will always need to use short twisted (low induction) wiring techniques.
And of course not only short wiring but also an low induction current measurement resistor.
If the builder for example thinks he gets a nice fast regulating / stable power supply with long wiring around the fast transistors, then the builder fools himself.

I learned a lot when I played with radio and stations on different high frequencies in my youth.
Then you learn that 5mm wire always has resistance and induction!

Current source
kleinstein I've already understood you the first time, what you say about the current source for current limitation control, I don't ignore you. 
For now I leave it as it is, I detest that I give the current limit a star value of 50ma.
Lower I don't think it makes much sense, because of the temperature sensitivity you already indicated.
But even when I replace the LED with a transistor, it still remains the case that there is not really a good coupling between this power source and the current sensor transistor in the LM723.
In my opinion, the "compensation" varies when varying the set maximum current.
I want to think more about this, and I would like to hear from you what you think about my comments on this.

negative rail
I'm still testing to find the best way to make the negative power supply rail, the schematic is not fixt about that.

For anyone who wants to know what kleinstein meant by supplying a small part of the reference voltage to the inverting input of the LM723,
search with Google for the schedule of the Philips PE1542A power supply.
This schematic also provides a way for better current limitation control.
This schematic has a high value for the compensation capacitor, this makes the dynamic behavior somewhat slow.
The tests in the schematic show a current variation for the dynamic tests from 80 to 100%.
I usually choose for 10 to 90 a 100% which is a lot more difficult for the power supply to be tested.
I will continue with my own design but find this solution certainly interesting.

Cliff Matthews
The negative supply rail can reach a value of -30V together with the 21V output voltage, gives slightly more than 50V over the LM317.
So a "normal" LM317 is not suitable here.

Hi SK_Caterpilar_SK
Thanks for the schematic, but I show here how to use a single transformer winding and a LM723 to create a pretty good power supply.

If I want a power supply with very good voltage and current control then I build one according to the Harrison concept and forget the LM723.
This topic is for fun and i hope people wil learn from it.


Kind regards,
Bram

[Wolfgang]

I have different schematic. Its a old lab powersupply for school purposes so basically idiot proof. No voltage spikes on startups no anything unwanted. Tested because I have the original unit which is performing great. The whole thing is nearly flawless. In its original config it uses the metal can variant of the 723 but I rewrote the shematic so that the pinout is compatible with the DIP version. Also uses some obsoltete transistors and zenners and diodes. I made a PCB layout for it so I can have another one to my desk. The original goes from true 0 to 20V. The only issue with it so far is that it requires a dual winding tranformer. I got the voltages and components so I can give you a list if youre interested to test it at all. It does 0 to 1A but I modified mine so it can do 30V max and 3 amps..

Soo if  youre just interested at all in testing it. I will get you the infos so you can do science on it if you want. I will send you the full image if you want because the max filezise is tiny for this to work properly.

Hi,

good circuit, and expandable, too. In case you dont like the dual winding transformer you could use a small PCB transformer for the 723 and a larger one for the power part.

[SK_Caterpilar_SK]

Hi,

good circuit, and expandable, too. In case you dont like the dual winding transformer you could use a small PCB transformer for the 723 and a larger one for the power part.

Exactly. It doesnt have to be powerfull at all for the LM, the specified voltage for the LMs winding is 12VAC with like almost absolutely no current requirement except to drive the little transistor. You can stretch this circuit to anything realistic. I think I mentioned that the original design is meant for school application (20V 1A) and its idiot proof..no voltage spikes on startup nor anything. Some stuff on the schematic should be changed. The output capacitance should be just enough so that the supply is stable. Not so overdone as it is. In case youre testing zenners or something that 100uF charge will render the zenner into a smelly piece of a bridge. The transistors really do not matter. Keep in spec kind of stuff, the output transistor used originally is a 10A TO3 transistor. I bellieve one of the best packages to dissapate heat . The KF508 can be replaced by a BD 139. The KZ141 is a 4,8-5,1V zenner (very high tolerance components were used for maximum cheapness but still works great) 400mW..basically anything within that voltage range. The KA261 is a 50V silicon diode 500mA (anything will do really 50V diode or above almost all doing 1A these days). E16 E1-E8-E16 are alll just general purpose diodes rectifiers KY132/150 (voltage 150V and current I guess more than one amp.) I did replace the original ones with shottkys because I found them a bit out of comfort warm so I just replaced them with 5A shottkys because thats what I had but anything 5A will do. On my design I used a premade bridge rectifier so I dont have to solder a bilion diodes and its cheaper. On the right up side you see the volt/current meter. Its obsolete and unuseable so I just ignored it on my boards that I have designed based on that schematic. If you want I can send you the board files (either Eagle or GRBL). I only know this circuit by coencidence that I have the original PSU xD And I really like it because even tho its under specs for like regular standards. Only 20V instead of 30V that you would expect from a lab supply these days, and only 1A, But it does not really mater cause the supply is reallly stable electrically quiet, fits on the desk works no matter what and I just like nostalgic gear fight me  .

[SK_Caterpilar_SK]

Hi SK_Caterpilar_SK
Thanks for the schematic, but I show here how to use a single transformer winding and a LM723 to create a pretty good power supply.

If I want a power supply with very good voltage and current control then I build one according to the Harrison concept and forget the LM723.
This topic is for fun and i hope people wil learn from it.


Kind regards,
Bram

Yeah that I just toally forgot about 

[SK_Caterpilar_SK]

Tesla, right? That brings back memories...

Yeah it is. True nostalgy

[Kleinstein]

This thread started with intentionally not using the floating regulator type.

So the circuit with the auxiliary supply while still using the 723 is a little off topic.
The floating regulator has its good sides and it thus today kind of the standard for a lab-supply. The shown old implementation looks easy, but I am not so sure it behave that well under extreme conditions (e.g. with a larger low ESR cap at the output). It still has the classic old weakness of a set voltage dependent loop gain. So performance at the ends is not expected to be that good. New digitally set ones naturally adjust the reference input side and thus avoid this problem as a side effect. The auxiliary supply could also power a possible display.

The circuit chose from the start of the thread still has a few advantages:  even with a very slow compensation the load drops are not that large and the output capacitor can be small. In addition it does not need a second transformer winding. However it is kind of limited to some 25 V maybe 30 V.  At least with the crude current limit from the 723 I would not call it a lab-supply, at best a low end one.
The shown circuit with the shunt at the high side also makes it difficult to add a digital display of the current.

[blackdog]

Hi,

Kleinstein
It still has the classic old weakness of a set voltage dependent loop gain.
Did you not look at the last schematic?
It has now a fixt gain setting!


At the moment I am testing what the deviation is that occurs due to variation of the mains voltage of the -5V negative supply rail.
If I use an extra zener I can get the variation less than 10uV DC by large mains voltage variation of the -5V rail.
I use a LAB power supply to test this, I vary at the input of the LM337 from -22V to -32V.
22V is the minimum voor 20V output and 32V will be the if there is a very high net voltage.

And I also tested the attenuation with a modular power supply (HP 6823A) to measure the attenuation of the 50, 100, 150 and 200Hz frequencies.
This is from -60 to -75dB depending on the measurement frequency, this is sufficient for my setup if i chose the buffer capacitor.

But...
This does not provide the results I would like to have for this power supply.
That is low noise and the best possible DC stability power supply without using expensive parts and extra transformers or windings.
I took out the zener and resistor and tested with a -15V LM7915 which I had available, BINGO clean!
I'm probably going to replace it with an LM7912 to reduce the dissipation in the LM337 a bit more to reduce the DC drift of this voltage regular.
Later today I will put this morning's testing into the last schematic.

kleinstein, later today I wil make soms remarkt about the stability of large capacitor on the output en maybe also BIG inductors. :-)

Kind regards,
Bram

[Kleinstein]

I saw the voltage setting was changed in the 723 circuit.

The remark on the variable loop gain was for the floating regulator circuit that SK_Caterpilar_SK  brought up. The same with the stability.

[blackdog]

Hi 

Here is the last schematic with adjustments I did after some measurements and thinking about the schematic.

I had already explained in the beginning that one of the characteristics I like to see in a power supply is good DC stability and few interference signals.
These interference signals can be noise from the LM723 itself and the extra -5V rail, but also interference signals from the 230V grid here in Amsterdam.

The tests I did this week showed that the sensitivity to voltage variations due to the 230V mains voltage variations and the load variations at the output of the power supply, I found the variations too large.
This means that I will have to use more parts than I had intended.
But the parts used are just like the LM723 "Old School" and I allow myself to use these parts. 
What I have done now is not so much different than that you see more in well designed power supplies, the LM723 itself is powered by "clean" power lines.

The -5V power supply line.
I have done extensive tests on the LM337 to see how good the hum suppression is and what the sensitivity is for variation of the DC level at the input of the LM337.
The output of the LM337 which is about -5V is in series with the reference voltage of the LM723, so you see every deviation of this -5V as variation in the output voltage.
After some tests with zener diodes I switched to a LM7912, it costs almost nothing and has a much better performance than a resistor and zener diode.
Now there is no hum or DC variation at the output of the LM337 that is present at the input of the LM7912.
The ripple on C3 is suppressed by more than 120dB if you do the wiring properly.

For a power supply with good properties, the wiring technique is important!
If you look at the diagram below, you can see that the ground (which is on J2 output bus) is split into multiple connections to this node.
There is a thick horizontal line in the diagram to J2 connection, which is the return connection of the energy coming from the transformer and the rectifiers and capacitors.
The same DC level of the -5V power rail (The GND connection of the LM7912 and this has a thick green line) must be connected directly to the J2 socket, as I drew in the schematic.
A third important connection to J2 is the resistor R21 in the diagram below, this is the negative "sense" connection.
Ignoring this wiring technique means that you end up with a worse power supply than is possible with the used circuit.

+23.5V power rail
Also the DC level on the Vs and the Vc connections of the LM723 are sensitive to variation.
When loading the power supply from 100mA to 1.9-Ampere, the voltage at C8 can drop 1V. (This is due to the Ri of the transformer)
This drop in the voltage at C8 is then seen as an extra drop in the output voltage, making the DC Ri lower than necessary.

The level of 23.5V will probably still change, as it depends on the Power output section.
I'm going to make two connections on my test print as shown in the schematic, the normal output which is pin-10 and pin-13 which is the direct connection of the difference amplifier in the LM723.
This direct connection goes beyond the Darlington in the LM723 IC.
I hope that with my high gain of the Sziklai pair I will have both a low DC Ri and a better phase margin, but the test wil show if this use of pin-13 is usefull.

The DC level of 23.5V therefore depends on the Power stage configuration and whether it is possible with the used transformer to reach 20V at 2-Ampere,
if it becomes 1.8-Ampere at a low mains voltage, then I will settle for that, no problemo.
Because this power supply is for fun and teaching. 

I have chosen for the 23.5V regulator to use a TL431 shunt controller.
The measurements I did on my test setup, which supplies the voltage to LM723 and the current that supplies the reference output of the LM723 to P-1 and R11 is between 5 and 6mA.
This is not the only current, also the current source made with Q3 which is about 1mA together with the current through the red LED are about 2mA total,
which should be added to the current consumed by the LM723, in total this is about 8 to 9mA.
C8 is now also enlarged to 2200uF to make the ripple as small as possible, this gives some more space for the TL431 controller.
later I can replace D3 with a Schottky version for about 300mV extra margin.
How to calculate the resistors arroud the TL431b?
I use a TI speadsheet which you can download via the link below.

www.bramcam.nl/NA/NA-723-PSU/TL431-Calculator.xlsm

Resistor R9 and R10 are setting the voltage voor de LM723 IC.
R6 determines the current that can be drawn that still falls within the control range of the TL431b.
Don't forget that you don't exceed the maximum dissipation of the TL431b!

I'm going to test the resistor R6 which is now listed as 120 Ohm in the schematic with the power supply connected to a 2-Ampere load
and 20V output voltage, to see if it still works at 220V mains voltage, at the two Power section configurations.


LM723 output voltage and sensor lines
To keep the phase marging stable on the output voltage between "0" and +20V, I decided to test the regulating the +input of the LM723 after a remark of kleinstein.
The bottom side of the potentiometer P-1 has a DC level of "0V" compared to J2.
The reference voltage(pin-6 LM723) is also with respect to J2 about 2.1V.
Note, I am not saying that the reference output of the LM723 is now suddenly 2.1V instead of 7.1V!
This 2.1V is due to the reference shift of the -5V regulator relative to J2, this because pin-7 of the LM723 is connected to the -5V regulator.
The trim-1 potentiometer in the -5V regulator ensures that the output voltage can be adjusted to "0V" when P-1 is turned all the way counterclockwise.

It is not that important that the -5V rail is exactly 5.0000V.
Noise, hum and stability are important here for this power supply rail.
It can also be -4.2V, -4.85V or even 3.765V or as kleinstein suggested 2.5V.
2.5V I found a bit on the low side by some testing I did at this low commonmode voltages for the inputs.
It's no problem to sit in the "sweet spot" of the input commonmode range for the best performance.
By using two regulated supply lines in this diagram, the LM723 can never get too high a supply voltage again.

As shown in the diagram, P-1 controls between 0V and +2.1V, this value will have to be amplified by the differential amplifier in the LM723.
As with a normal opamp, the gain is controlled by two resistors connected to the -input.
In the schematic these are R17 of 10K and R21 of 1K, I have a small trim range applied so I can set the maximum output voltage, this is done with Trim-2 potentiometer.

A few remarks about the values of the resistors and potentiometers that are attached to the + and - inputs of the LM723.
With the chosen schematic it is not possible to precisely equalize the impedances.
The impedance at the -input is fixed but that of the +input depends on the position of the potmater P-1.
At the +input there is always a 1K resistor present that together with C12 of 4u7 form a low pass filter, witch filters the reference noise and when changing the output voltage.
At this 1K, in the worst case with the potentiometer on "0V" the parallel resistance of P-1 and R11 must be added up which is around 700 Ohm.
At this low value of impedances, the bias currents and noise are less important.

Schematic version 0.7



Resume
So what has changed in the schematic:

1e
The output voltage is now controlled by variation of the + input, phase marging remains stable over the entire output voltage range.

2e
The LM723 is powered by two clean power supplies that are much more stable and cleaner than the first setup.
This makes the final performance a lot better, I love to design clean power supplies, which is one of my abberations. 

3e
That there is now the possibility to choose the output to which the power section can be connected with D7 and capacitor C16
which give a DC shift for the output commonmode without adding phase margin.


Still to be done
Testing of the Power section connected to the compentation connection.
Select the current limitation components well, it involves R20 and R24 and perhaps replace the current source version as proposed by kleinstein.
If the phase margin and dynamic behavior are good, the timing of the Power On/Off circuits should be further tuned so that no abberations occur.

And some more explanation why I made certain decisions for this power supply.
But, to say it again, this is a power supply for me, and not meant to be a building project.
And the second reason I started this topic is that I learn to express myself better.

This is mainly intended as a learning project for those who do not yet have much experience in building or designing power supplies.
Many who start with electronics want to build/design a power supply without sufficient electronic knowledge. (buy The Art of Electronic)
I am no exception, I made many mistakes and have read very many hours in the first versions than The Art Of Electronic

Power supply are difficult circuits, and yes I know the Internet is full of "simple" schematics.
This almost always means that the "designer" himself does not know what he or she is working with at that moment.
A common mistake is that the person who is going to rebuild a schematic will replace parts without much thought because they have them in stock.
Please do this in you want to have a poorly functioning circuit. 

For your information.
This piece of text with the adjustment of the schematic took me almost 6 hours today.
Gaining knowledge and transferring it takes a lot of time and effort, if you want to learn you will have to make a real effort
and this effort for good learning can not only consist of just looking up something quickly via Google.
I see more and more that people want to know everything at the push of a button.
And I think, who are you fooling?

Like Louis Rossmann I want to say now, I hope you learned something from, I dit.

Kind regards,
Bram

[Cliff Matthews]

R7 and Trim-1 seem hard to get, for the avg. bloke with just e24 series, why not use 390 and 100 ohm trim?

Also, can I use 2SA1695 I mentioned as the final output? (another big PNP I have is TIP147..)

[Kleinstein]

The 2SA1695 should be OK. The TIP147 is a Darlingtion transistor and would thus behave different. AFIAK the PNP does not have to be super fast, its more like the NPN in the Sizlaki pair should be considerably faster to ease stability. Otherwise one might artificially slow down the PNP with an extra cap.

The circuit still has the weakness of using 2 references in a kind of difference mode. In the last plan the adjustment range for the the pot is only around 0-2V for the input. This is set as a fraction of the 7 V reference inside the 723 minus the 5 V from the LM337. I don't think this a really good idea.  I would expect it to get better if the negative voltage from the LM377 is reduced, e.g. by having the adjust pin directly at ground. With a higher reference level there could be less divider in the feedback and thus better DC precision.

7 V - 1.3 V sound a lot less odd than 7 V -5 V for the reference. It should also work having the pot directly towards the ground (= negative output) instead of the resistor towards the -5 V (negative supply of the 723).

[blackdog]

Hi,

I already showed that I want to keep the impedances around the inputs of the LM723 low.
That results in some more gain needed to reach 20V.
I also indicated that with earlier measurements the open loop gain became lower, when the inputs have to work at low commonmode voltages.
I do think there is still room to bring the -5V to e.g. -3V, this without too much performance loss.

Wait.... let's just say that I'll run the new test print setup with -3V power rail.
Tonight I will have change the resistance values for that in the schematic.
But.... Keep in mind that this is a schematic that is in the design phase.
If it suits me or the measurements dictate this, I will adjust the value again.

In the weekend I decided to bring the maximum output voltage of the power supply to 20.1V.
The reason is that I have analog meters on my order list of 20V and 2 or 3-Ampere.

2SA1695
I prefer the 2SA1943, the reason for this is that this transistor is cheap, and the most important reason is the large size of the housing.
This almost always results in a better (lower) thermal resistance.
And this, please, please, please, please, do not buy the cheap transistor junk on ebay!

Cliff
Even if you only think of the TIP147, your punishment is that you have to read this whole topic 5 times. 
Yes now I am joking!
As kleinstein indicates, the PNP Darlington is a PNP that you don't want to have in this circuit.
This Darlington can also be used in power supply's, but not in this one without the properties I want to be affected too much.

Resistance values and trimpots.
I calculate the values I find necessary for the best results.
As you can see here already in this topic quite a lot changes in the schematic by insight and measurement results.
If you would like to build this design, it is best to wait until the schematic is in its final stage.

I use values from the E96 series, but if possible I choose nice round values like I did with R17 which sets the gain of the LM723 and also R8 of the -5V regulator.

Trim-1 I choose in the final schematic so that this trimpotmeter has only a small range to compensate for the inaccuracy of the LM721 reference voltage,
Potmeter P-1 resistance deviation, the resistors R7 and R8 and the accuracy of the LM337 reference voltage.
Trim-2 will be chosen so that I can eventually set the max. output voltage to 2.1V.
This trimpot will also have a small range, to be determined when I have the schematic ready.
The trimpotmeter will therefore have as small a range as possible, so that they will only have a small part in terms of DC stability.

As far as I'm concerned, Trim-2 may also be in series with R17 if the right ratio is chosen.
Trim-2 may also be parallel to R21 with a series of resistors, which is even nicer.
Of course I calculated it in such a way that it only has the necessary control range.
Then there is this, a trimpotmeter is usually placed on the side, which is the least sensitive, so the lowest impedance possible.

In my next schematic I will apply that too.
So the trimpot can be placed at the top of R17 or at the bottom of R21.
In schematic version 0.7 Trim-2 and R21 have to change places.
Where the trimpot Trim-2 will be depends on the values I will calculate.
So later this evening I place version 0.8 of the schematic in this topic.

But now first I go to my mothers house, to reset her TV, because the cable company has changed some settings.
And then I am expected to participate in the evening meal with the lady of the house. 

Kind regards,
Bram

[blackdog]

Hi,

The schematic has been modified, so I get a slightly higher value over the voltage potentiometer, this one is now 4V instead of the 2.1V it was.
The gain of the LM723 now only needs to be 5x to get 20V output voltage.

Schematic version 0.8


I have made sure that where I can I use default values.
I have made R10 equal to R18 so there is no need to buy 10.5 and 10K in a 1% E96 value.
Furthermore, it is no longer possible to use a 1K type for P1, this gives a too high a current that has to come from the reference connection.
This results in a larger dissipation in the LM723 IC and a larger startup drift.

So the next logical step for the ones who want to build the schematic is a 5K potmeter.
A 2K potentiometer can also be applied when the value of R9 is adjusted.
The current through the P-1 potentiometer is then 2mA which I consider to be maximum, I know this may be 15mA, but that doesn't mean it's good.

25mA current source
Dissipation VR3, the LM317HV, the dissipation of this controller can be limited a bit by recording a 560 Ohm 2-Watt resistor.
This is then 14 a 15V voltage, and there remains enough voltage for a good control, even at a low output voltage of the power supply.
Keep in mind when applying this resistor and maybe a small capacitor of 1uF should come after the resistor to keep the LM317HV stable.
At 680 Ohm you could even use a "normal" LM317, but then I would put a 36V zener over it for protection.
I'll test that and if I'm okay, include it in the schematic later.
All in all, a single LM317HV is easier.

Testprint
I had already started rebuilding the test print, but due to the other value of the reference voltage I will now have to change it again in part., live sucks 



At the far left top you can see three parallel connected bridge rectifiers (BY164) These are replaced when the schematic is finished by a GBJ-1506U .
The top three capacitors is C5 in the schematic, only here the total value is 6700uF.
The two large capacitors on the row below are 2x1500uF and that are C1 and C3.
These are 1500uF on the PCB and 1000uF on the Schematic, the 1000uF will be on the later PCB when I finish measuring on it.
And the last capacitor, the black small one is C8 in the schematic.
The small blue trimpotmeter and two resistors on the right side of the LM723 will be changed in value because the reference voltage has changed.

It will probably be the end of the week before I can show measurement results from the last scheme.
My work doesn't allow me to spend more time there, no matter how much I like it.
Please be patient, I will certainly read your comments because I appreciate that.

Kind regards,
Bram

[iMo]

The TL431 as the shunt regulator for the 723 - hmm, would not be better to use a 7824 instead?
Also, what happens if the wiper of the potentiometer P-1 5K loses its contact?

[blackdog]

Hi imo,

Of course I thought of a 7824 or a LM317.
But also think about the noise, I would like a power supply as clean as possible for the LM723.
These are all considerations that go a bit further than just looking at e.g. the dropout voltage of a regulator.
The suppression of e.g. the ripple voltages present at the input of the regulator.
This suppression often gets less the closer you get to the dropout voltage.

If the input voltage of the TL431 becomes too low to regulate properly, there is still good extra ripple attenuation by R6 and C14.
But these are now the decisions you will have to take when you are designing a circuit.
If I had some more voltage from the transformer like a 22V model instead of the 20V, there would be a LM317 in that place. 
But i am designing with my 20V transformer and the minimum of 220V mains voltage. (230V is here nominal)

What wil happen if R17 or R18 break lose...
And I can think of even more disasters, you have to live with it.

But back to P-1, and there is a capacitor of 4.7uf parallel to the + input, this capacitor does not only filter the reference noise but also helps with dendering of the rider of the potentiometer.
And again, the same as my comment about the transistors, don't buy rubbish on ebay!
Order the good Vishay type model 534, 10 turn pots.
I have hardly ever had any problems with this type.

Kind regards,
Bram

[Wolfgang]

Bram,

if you are concerned to supply a cleaner voltage to the LM723 you could try to preregulate the V+ pin. This pin goes to the error amp, and this error amp inside the LM723 has a limited CMRR only. So filtering this improves output ripple. What you could also do is filter the reference voltage. This reduces output noise, too.

If interested

https://electronicprojectsforfun.wordpress.com/power-supplies/a-collection-of-proper-design-practices-using-the-lm723-ic-regulator/

[blackdog]

Hi Wolfgang, 


I am filtering the reference voltage, but only on a different point.
C12 in the last schematic has also this function.

Kind regards,
Bram

[Wolfgang]

Hi Wolfgang, 


I am filtering the reference voltage, but only on a different point.
C12 in the last schematic has also this function.

Kind regards,
Bram

Hi Bram,

I've seen C12, but you could add another C directly at the reference output too, thats what I wanted to say. Pedantry ? Yes. 
Your solution for current control works from low currents on, but its still dependent on the Vbe of the 723s CS/CL pins. So its temp  dependent and not very accurate. 723 current limiting is more for the prevention of desasters than for fine and accurate current regulation. What you could do is to keep the idea of the current source, but then use an op amp to compare the drop across the current sensing resistor. The op amp could then (via a diode voter circuit) pull down the base of your pass darlington.

Have fun
  Wolfgang

[blackdog]

Hi Wolfgang, 

I know that the current limitation is not optimal....
I don't need that for this power supply.
As you already explain it, it is this kind of simple solution only meant for "calamities".
I am fully aware of the control properties and their temperature sensitivity.

A large part of my life (> 50 years) I often use the current limitation differently than many other people do.
When I connect a D.U.T. to a Power Supply, I first think about the current that will or may be consumed.
Then I set this current and whether this is 50mA or 55mA does not matter so much to me, even if the current drifts from 48 to 53mA, so what...
When I need a very precise current I almost never use a power supply but a current source.

And I'll repeat it again, I won't charge a battery with power supplies and certainly not without using a diode.
I think it doesn't happen more than twice a year that a battery is charged from a power supply in my LAB, I use battery chargers for that. 
And a LED is always tested here via a series resistor, or via one of my multimeters (TEK DMM4050)
I have 15 power supplies here in my LAB and if I need precision current limitation then I do so with a power supply type that is suitable for that.

This design I show here puts more emphasis on the smallest possible abberations and as little noise as possible and a reasonable DC drift.
It is not a "One size fits all" power supply.
And again, every user of measuring equipment must know the characteristics of this equipment if you want to make as few mistakes as possible.

If I want a more precise current setting, I can always choose at least two power supply's that are on my workbench, that is the Rigol DP832 and the Agilent 6632B.
But as I said, that almost never happens especially for that purpose.

I Love/Hate my Rigol DP832, this has a high resolution for both current and voltage and this together with the power display.
But the 470uF over the output terminals don't make much sense if it is for example set to 24V and 20mA maximum current.
The charge of the 470uF capacitor is dumped into the DUT when something goes wrong.
What is the point of a precise 20mA current limit then? so... the user must know the characteristics of this equipment if you want to make as few mistakes as possible. 

But I would like to hear more good comments, we learn from that.

Kind regards,
Bram

[Wolfgang]

Hi Bram,

I never intended to keep somebody from experimenting. So having fun is most important, its clear that your
PSU is not an academic endevour. In my view it should work and is worth to be tried out.

Some comments about output caps. Output caps have two functions:
- to reduce output impedance at higher frequencies where the regulator gain goes down
- for LDOs, together with a suitable ESR, to ensure stability.

Its clear, as you write, that all current limiting is useless if you have a large output cap and the limiting happens *before* the cap, as it does in all power supplies I know (DP832 included). If you need precision limiting in nanoseconds *plus* a stiff regulator, then stability with arbitrary loads becomes a challenge. So, the solution most PSU manufacturers offer is a medium size output cap with no further limiting, but an all load stability and a reasonable PSU output impedance.

I have tested the DP832A output impedances. Results can be seen here:

https://electronicprojectsforfun.wordpress.com/power-supply-impedance-measurements-for-various-power-supplies/

and here:

https://electronicprojectsforfun.wordpress.com/power-supply-impedance-measurements-using-the-bode100-lf-vna/

regards
  Wolfgang

A last comment: did you know that the (obsolete) SG3532 did implement your current limiting idea ?
This chip made by SGS ATES was meant to be an improved LM723 with a lower voltage reference, better current
limiting, overtemp protection, ... Only it failed on the market because the LM723 was often "good enough"
plus it soon became dirt cheap.

Have fun
  Wolfgang

[blackdog]

Hi Wolfgang,

Maybe I even have an SG3532, but as you said it makes little sense to develop anything with it because it is no longer for sale.

Are you the owner of the measuring instrument, because this device is quite expensive.
A few years ago I already looked up how much it cost, I fell off my chair...

Rigol tests
This is the left channel, the yellow trace is the current pulses from 50 to 0.8mA.
The measuring cables are optimally connected to measure as precisely as possible with as few abberations as possible.
The blue trace is the deviation at the output of the power supply caused by the current pulse.



This is the middle channel with the same conditions as the left channel.
It is visible that this channel is not quite the same as the left channel, but the differences are small.
This may be because the 470uF over the output is slightly different, and the sense wiring in the power supply is different.



And this is the right channel, which goes up to 5.5V but tested under the same conditions.



You want more? 

OK
The analogue DELTA E 030V 1-Ampere power supply, famous here in the Netherlands for its very good quality.



And this is also a DELTA Power supply and the type is ES 030-5 and this is a fully digital power supply.
The blue output pulse shows that this power supply is not suitable for precision applications.
An advantage of this power supply is that although it is a digital model, it has little EMF.
I use this power supply regularly when I need higher powers and the pulse response and EMF are not that important.


The power supply I made according to the Harrison concept
is much better than the pictures I show here of the Rigol and the DELTA power supply's.
That's because I use modern parts with fast power sections.
Does everyone need that? No, of course not, many can use an average power supply.
But I also think that many people don't know how their power supply behaves when switching it on or off or sudden large load variations.

I don't need this power supply with the LM723 at all, I do this because I like to get the most performance out of the LM723 without using expensive parts, adding opamps and extra transformers.
Furthermore, I hope that readers who don't have so much experience with designing will learn something from it.

Kind regards,
Bram

[Wolfgang]

Hi Bram,

your curves look quite reasonable, you would expect something like that. It proves that the DP832 is lethargic and the control loop has enough phase margin. The deltas dont look good, and weakly damped oscillations should never appear in a lab PSU.

When injecting current its important that leads are kept short to minimize inductance. Some tests of commercial current injectors I have done can be found here:

https://electronicprojectsforfun.wordpress.com/measuring-a-picotest-j2111a-current-injector/

or you use injection transformers (not so practical for a commercial PSU where you cannot open the regulator loop):

https://electronicprojectsforfun.wordpress.com/injection-transformers/

The equipment I have is in my EEVBlog profile, if you are interested. In fact I have a VNA, but not for low frequency stability work (E5071C 8.5GHz). It starts at 9kHz so its not too useful. I agree that LF VNAs are surprisingly expensive. The absolute record holder is the Keysight E5061-3L5 which can do sweeps from 5Hz to 3GHz in one go and a 120dB dynamic range for a whopping 35K€ list price. Keysight did not want to give me reasonable discounts because they claim that the market is very small and they have no competitors. A Bode100 does 1Hz to 50MHz with a little less dynamic range for about 5K€. I might get my hands on a Bode100 again soon, maybe Omicron and myself team up and I can have a test unit.

Have fun
  Wolfgang

[IconicPCB]

Bram,

What is this Harrison concept You refer to?
I know Harrison were bought ou by HP i guess in the sixties may be seventies, but i do not know what this concept is You refer to.

[Cliff Matthews]

Bram,

What is this Harrison concept You refer to?
I know Harrison were bought ou by HP i guess in the sixties may be seventies, but i do not know what this concept is You refer to.

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1962-07.pdf

[IconicPCB]

Thanks for the link, interesting read.

[blackdog]

Hi Wolfgang, 

Only the digital 30V 5-Ampere has no good dynamic behavior.
The 30V 1-Ampere model is the best of the pictures shown.
It has a low Ri and fast respons.

Measuring phase (bode plot)
I found a small transformer that had a good frequency respons and used this transformer in a measuring device.
Bolow you see the schematic of this measuring device.
the design is by Wenzel but I adjusted it to get a larger frequency range and to lower output impedance.
In the drain I included a coil and replaced the buffer circuit at the output with a Fet and a transistor with a LH0002 buffer.
This increased the bandwidth to 0.1dB at 1MHz.
The large input capacity of the 2SK170 low noise Fets is not important here, because of the low impedance at the input of the Fets.
Nice preamp! 

Schematic



Front
The box is a old Cisco PIX firewall.



Back side



Inside de box



The two outputs at the back I then connect to two inputs of my HMO3004 scope for phase measurements or if the frequency is not too high,
I can also use my Audio Precesion One measurement set for phase measurements.
But usually I use one of my Dynamic Loads for stability testing.
For the smaller power I use a design by Jim Williams with a very high bandwidth, so of the top of my head about 3 to 4 Mhz bandwidth.
And my 200-Watt dynamic load with a much smaler bandwidth say 100Khz.

I am aware of the measurement conditions when doing dynamic testing.
If you already look at my pictures of the tests I show here, you can see that there is no HF overshoot visible anywhere.
But later more on this if i test this LM723 power supply.

Kind regards,
Bram

[xavier60]

Bram,

What is this Harrison concept You refer to?
I know Harrison were bought ou by HP i guess in the sixties may be seventies, but i do not know what this concept is You refer to.

http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1962-07.pdf

The reason that it uses a PNP series pass transistor in the negative rail is likely because mainly PNP power germanium transistors were available back then.

[Wolfgang]

Hi Bram,

what I can see you are using the classic injection transformer technique to measure the transfer function. Why not, a lot of other phase/gain/loop stability systems do the same.

Some comments:
You measure with the control loop opened up and your injection transformer inserted there. For a homebrew PSU, no problem, for a commercial DUT it is sometimes unpractical to cut PCB traces and so on. Some modern PSUs contain IC regulators where the output divider is *inside* the IC. In this case, you have no chance to open the loop and to insert your transformer. Another example of this are chips like 78XX.

In case you cannot or want not to "invade" your DUT you can work with current injectors to determine phase and gain margins. This method is called NISM (non invasive stability measurement) and is aggressively promoted. As all hyped stuff, it does have restrictions (exactly only works for second order systems), but still better than nothing.  Personally I think measuring step response is more "honest" because it also works with large transients that could drive the PSUs internal regulator out of its linear range. Even nonlinear systems *can* oscillate

[blackdog]

Hi,

I didn't feel like working today, so I worked on the schematic of the LM723 power supply.

Four parts were tackled, the current source with Q3, the extra rectifier with D3 and C8, L1 and R37 and finally the extra MOSfet Q7.
I would like to hear from you what I try to achieve with these changes. 

Version = 0.9 of the schematic.
The component values of the new sections are a starting point, these have not yet been optimally determined by testing.


I wil be happy if i hear some explanations from you.

Kind regards,
Bram

Sorry, i placed the wrong schematic, it is now corrected.
No, it is not the alchohol! 

[Kleinstein]

One could get away without R9, if the lower end of the pot is connected to the negative output terminal level. For the regulation I would expect a slight improvement from this, though not much. Independently C12 could also be better of towards the neg. terminal level - this way the filter would be active for the negative supply contribution too. 

After the charge pump through C1, I see no need for the very high voltage. So it could be possible to reduce C1, up to the point that the voltage before the 7912 gets reduced. My guess would be that a low smaller cap (e.g. 220 µF) could be sufficient. It better to drop the excess voltage at the capacitor than with the resistor and 7912.

I think one half of D3 should go to the other side of the AC, to use full wave rectification.

If pin 10 of the 723 is used as an output, I would expect an extra resistor from base to emitter for Q4, so that there kind of a minimum current flowing from pin 10.

With the added Q7 and R31, I would expect that a larger value for C20 might be better suited.

[blackdog]

Hi,

Kleinstein good input!
About R9, i feel so stupid... changed and also removed Trim-1 and changed R7 to 221 Ohm to make the LM337 a 3.1V regulator, trim is no longer necessary!

As for C20, you are right, it must be bigger to keep a sufficiently low impedance at this point, it is now 1000uF.

A quick explanation about R21, the 4.7 OHm 25-Watt resistor.
This resistor limits the dissipation in the 2SA1943 transistor at lower output voltages of the power supply.
The switching point will probably be somewhere around the 8V output voltage.
The value of the resistors around Q8, the 2N3904 and the zener D11 will be determined later.

D3, L1 and R37 ensure a less voltage drop compared to the single resistor and the 1N4007.
L1 is a high impedance for disturbances from the 230V grid and R47 helps to prevent the Q of the coil from being too high.
With this setup I hope to have a little more voltage for the TL431b regulator.

As for the negative rectifier, I will do some more research on this.
but I don't like to drop large AC voltages over electrolytic capacitors.

About Pin-10 of the LM723, You are probably right about the extra load resistance that is needed.
The Darlington in the LM723 probably can't work well with the very high Hfe (low base current) of my power stage.
I will come back on this point.

I will also make a test setup where both the extra positive power supply line and the negative power supply line will be built a bit different.

Here is version 9b of the schematic.


Kleinstein thanks!

Kind regards,
Bram

[blackdog]

Hi,

Yesterday and today I did measurements on the transformer and rectifier.
These measurements indicate that at 220V mains voltage, 20V at 2-Ampere output current is not possible with the first 20V transformer I tested with.
That in itself is not a big problem because it is a "test" power supply and a learning project.

This is the first transfomator I did all the tests with, is is a 20V at 70VA model.
The top is normally closed with a lid, but removed here to see if it was posible to adjust the windings.
While removing the lid I remembered that I also have another transformer of 20V and this transformer I started comparing with the first transformer.



This is the second transformer it is a toroidal core of 20V and 100VA.



Both transformers side by side.



And now the big differences between these transformers.
The mains voltage is now a bit higher because I did these measurements with a separation transformer that gives a higher voltage at low load and for my savety. 
This is the first transformer of the E-I type, tested at 231.6V



Holy Moly 42.3VA no-load!



Because of the lower zero load of the toroidal transformer, the voltage here is 237.8V.



This transformer is much better and not only in terms of the no-load. very nice value at 237.8V.
Thise toroidal transformer is much nicer for Mother Nature, the power plant does not need to generate 40VA extra with this model transformer.



I also measured the DC resistance values of the windings of both transformers.
The first transformer, that's the rectangular model.
The primary winding is 23.85 Ohm.
The secondary winding is 0.23 Ohm.
Transformer wattage = 70-Watt

The second transformer, that's the toroidal model.
The primary winding is 13.35 Ohm.
The secondary winding is 0.17 Ohm.
Transformer wattage = 100-Watt

It is tempting to increase the current to e.g. 3-Ampere when using the toroidal transformer, but then I have another problem, cooling!
I have not yet discussed the cooling of the power section, but I almost always keep a maximum of 50-Watt per power transistor.
Below a picture of a PC heat sink that I used for another power supply test.
The parts that are now mounted on it will be replaced by parts that I ordered this week.


I have many more of these PC heatsinks, also larger models.
From experience I now know that 50-Watt dissipation is possible at a not too high sound volume of the van.
I want to try to build the power supply in a not too large housing and the toroidal transformer and this small forced cooled heat sink should help.

Tomorrow I will continue measuring the toroidal transformer, the first measurements already showed that with this transformer 20V at 2-Ampere must be possible at my minimum mains voltage of 220V.

Kind regards,
Bram

[Doctorandus_P]

I just wanted to know you' all that I enjoyed reading this thread.
Too many newbies who have no clue about what a dynamic measurement of a power supply is, nor even have an oscilloscope.

The HP article was also fun. Stating that it was becoming possible to build such power supplies with new advances in electronics.
Gosh, that was only in the 60's. Just a few years before I was born in this ever more faster morphing strange world.

[blackdog]

Hi,

Some short extra remarks about the transformers.
The "zero load" of the first transformer also causes extra heat.
This also happens when you supply a light load like an Arduino or other microcontroller with this power supply.
And of course I am aware that the 25mA current source at the output also dissipates about 1.5-Watt.
Then there is also the current from the bleeder resistors, that do not help the efficiency.
This will not be a high efficiency power supply, but one that has little noise and hum and is kind to the load.

Some info about the mains voltage deviation in the Netherlands.
If you are going to measure the voltage (which comes from the socket or wall socket) at home, the voltage is not always exactly 230 volts. 
There is a minimum and maximum margin.  In the Netherlands standards have been drawn up to which the voltage must comply.
On the connection in the meter box (the connection point), the voltage must comply with the European standard NEN-EN 50160. 
For the voltage it has been agreed that the voltage to the bottom may not deviate more than 6% (207 volts) and to the top no more than 10% (253 volts). 
This data is a shortened summary and is not complete in detail, but gives a global picture, in the NEN-EN 50160 is exact way of measuring.

I am in the middle of the city of Amsterdam and there is almost always the mains voltage on the low side.
I haven't measured any mains voltage here yet in my LAB that was above 232V.
This power supply is intended for use here in my LAB.
If anyone is going to use this schematic or any other schematic and/or transformer, think about it.

If you design a power supply that has to work at multiple locations and cannot break down at 253V mains, you have to take that into account when designing!

Just think about the following, the first transformer already consumed over 42VA at 232V without load.
This almost always means that the transformer is designed so that the core is against saturation.
There is a chance that this is a 220V transformer, I can't check this further because I don't have enough information.

What kind of power consumption would this transformer have if you offer it 253V, the core will be fully saturated!
Offering these 253V for a longer period of time, even at low or no load would result in a too high transformer temperature.
One of the positive points of this transformer is the built-in temperature fuse witch will blow at 253V afther a while.

And then the second positive point of this transformer, the windings are well separated.
The capacity between the secondary and primare winding is only 30pF, and if I use the iron package as a shield it is only 20pF.
The toroidal transformer has a windings capacity of 760pF, which is many times higher than the normal transformer.
This toroidal core is not equipped with a shield, which do exist, and are often found in medical equipment and/or measuring instruments.

One of the last pieces of this power supply design, is to deal with are the parts on the mains side.
Now back to the bench to do more tests on the toroidal transformer and its rectifiers.

Kind regards,
Bram

[blackdog]

Hi

The schematic has been slightly modified again.
I have tested the rectifiers but I still need to do some tests with the minimum power needed to make my setup work properly.

The charche pump made with D1 and D2 in the schematic together with C1 and C3 require a certain load on the power supply rails.
So the next test is to apply the bleeder resistors and two current sources that will form the load for the extra positive and negative power lines.
The transformer loaded by one of my dummy loads with 2-Ampere for more than half an hour and the 10,000uF capacitor, became just noticeably warm, good!
R5 is this last schematic keeps the dissipation of the 7912 down en helps to filter fast pulses from the mains, it wil be around de value in the schematic (15 Ohm)

The measured values in the schematic are those of the toroidal transformer.


Laters more 

Kind regards,
Bram

[coromonadalix]

Im gona do some comments, i dont want to critize your design  but i find it too complicated for nothing

Ref1 r15 r16 c10 r12 are not needed if the vcc supply max is under 40 vdc for the lm723  as per datasheet

I understand the q8 dissipation circuit,  but it add  thermal loss and voltage drop before the q8 emittor, i would  not do that,  why ?

The 723 as i see this circuit is not a floating regulator, you try to do this with d3 dual diodes, but its not.  You must have dual secondaries x-former.

You have some negative -3vdc  rail with your power on/off and enable on/off, i checked lots of 723 based circuit  and never saw one applied liked this ??

To negate the 723 output to zero, you bring a negative voltage on the output potentiometer voltage ... like the attached photo  (it's not a floating design)

What is the purpose of the  q3 r21 c16 r22 rref2  on the 723 CL pin ?   is it to have a very low current adjustment ??

To have a very low current adjustment you add a power diode in serial of your current sense resistor, that way (with current sense pot adjusted value) you will have a minimal voltage who will permit to set a very low curent ...   i have a similar 723 design at home, i can drive a led with no resistor on it can go down to 10 ma ... if i turn down the current adjustment the supply kicks into shutdown.

As i wrote  i understand your design considerations, but it is over complicated in a way ... i know you have some parts laying around and do what you can with them.

If you want  say 20vdc at the output, be sure to have at least 10vdc "overrange" dc supply at the q8 emittor,  you could have used an all npn circuit for the output section ?


https://electronicprojectsforfun.wordpress.com/power-supplies/a-collection-of-proper-design-practices-using-the-lm723-ic-regulator/

The term floating is :  the 723 ic ircuit has its own supply isolated from the main output transistors,  the 723 supply will always be higher (on top) than the output power supply, that way  any loads variations on the output will not affect the 723 regulation.

You have some mastech psu designs schematics who will show you that.


Check this link : https://bama.edebris.com/manuals/b&k/1602/

Its an b&k precision 1602  with an 723 in it, this lab grade circuit is very well designed ...

I will try to find my 10 year 723 design loll, my current limitter is "before" the output transistors section, that way all the voltages loss aren't affecting the 723 regulation, i dont use the 723 internal current limitter circuit, and the 723 is used in a foldback design if i recall.  Have to ask my friend to give it back to me for a week ...     

Kinda look a like this second attached photo

[blackdog]

Hi coromonadalix, 

Have you read all my starting points for this power supply?


Ref1 r15 r16 c10 r12 are not needed if the vcc supply max is under 40 vdc for the lm723  as per datasheet
The point is not the maximum voltage of the power supply of the LM723 but the suppression of hum and other signals from the grid!
It is also important that the DC value is stable with varying mains voltage and/or load of the power supply.
This schematic is to show what it takes to get a low Ri and a noise-free powersupply and to learn, and to learn that power supply's are not "easy" if you want to do it right.
Does everyone need such a nice low noise power supply, I don't think so, but that is not important for this topic.
This topic is there to learn from, what is important when developing a power supply.
And I have set some starting points like a single transformer winding, good dynamic behavior and no on/off abberations, what is wrong with that?
This power suppy I will never take into production, the one I build is to show anyone who wants to know how to do it with good results.

I understand the q8 dissipation circuit,  but it add  thermal loss and voltage drop before the q8 emittor, i would  not do that,  why ?
This is fitted to limit the dissipation of Q8, R10 taking over part of it to dissipate, at output voltages of the power supply below 10V.
This helps with the SOA of Q8 at larger output currents at low output voltages.
This is not so strange, was also used in HP and Harrison power supplies.

The 723 as i see this circuit is not a floating regulator, you try to do this with d3 dual diodes, but its not.  You must have dual secondaries x-former.
Yes and? Have I talked about a floating regulator? D3 is intended to make a power supply as clean as possible, so little ripple. and Ref1 makes the power supply clean like a battery.

You have some negative -3vdc  rail with your power on/off and enable on/off, i checked lots of 723 based circuit  and never saw one applied liked this ??
Yes and? am I not alowd to do it my way? If Frank Sinatra is allowd to do it his way, i have the same rights 
Have you ever tested the schematic you show yourself for on/off abberations, sensitivity to mains voltage variation hum and noise at the output, dynamic behavior etc.?
I did, because the schematic you show, was my starting point. 

What is the purpose of the  q3 r21 c16 r22 rref2  on the 723 CL pin ?   is it to have a very low current adjustment ??
This is intended to create an adjustable current limit from about 50mA to about 2.2-Ampere.
Q3 injects a current into the LM723 to make the current limit more sensitive, say my 50mA.
And yes that's a bit sensitive to temperature, but I'm not going to explain it again. 

To have a very low current adjustment you add a power diode in serial of your current sense resistor, that way (with current sense pot adjusted value)
you will have a minimal voltage who will permit to set a very low curent ...   i have a similar 723 design at home, i can drive a led with no resistor on it can go down to 10 ma ...
My starting point is that everyone can think up and build schematics the way they want.
The diode in series with the sense resistor is a solution but also creates more drop out voltage.
And I just had too little voltage available to apply this properly.
Furthermore I find 50mA as a minimum for this power supply sufficient.
If you want to test LEDs without qa serial resistor be my guest, if you would work in my company and do this kind of tricks, I would fire you. 

If you want  say 20vdc at the output, be sure to have at least 10vdc "overrange" dc supply at the q8 emittor,  you could have used an all npn circuit for the output section ?
The point is precisely to show how you can achieve good results with a little extra attention even if you have less voltage available, see also my remarks about noise and hum.

Its an b&k precision 1602  with an 723 in it, this lab grade circuit is very well designed ...
You don't find BK's LM723 power supply complex?
If I'm going to use a transformer with multiple windings, I'll build something according to the Harrison concept and throw the LM723 in the garbage can. 

Kind regard,
Bram

[coromonadalix]

I read all of the points,  that was my 2 cents comments   I simply dont like your charge pump idea and some deviations route you take, it adds complexity many other parts  etc...

You do know for the "grid hum" you have very good EMI RFI line filters who will help a lot, you know if you build an electronic supercapacitor, there's a limit too to
what the lowest noise floor will or can be (IE:  like adding tons of filtering capacitor, in the end it will have an adverse effect)

It's a classic circuit with no pfc correction.


I still dont like your q3 r21 c16 r22 ref2 circuit to go down to 50ma ...  thermal drift and other problems will or may rise.   

This is not following datasheets design and for other who may be tempted to follow or build your design,
You're right its your design, you can do what you want, but some warning should be issued, some people take things for granted you know.


If your project works and is reliable and stable  etc ...  with the specs you need,  test with inductive / resistive / capacitive loads ... shorts, reversed voltages etc...


My design was an assemblage of 2 separate designs put together, To have best of both worlds.

I will wait your final design and hope to see some reports, o-scope screenshots etc .. i'm curious too to see where it goes. and i applaud all your efforts and design considerations.

I tried tons of time to build an good 723 circuit, there was always something missing for one need or the other  loll  There should have been a new 723 chip design like the old mc1466 was ( higher voltage input / output )  there was an L146 ic who was supposed to have higher voltages ?? but same 723 functions.

Kind regards
Martin L

[blackdog]

Hi,

You do know for the "grid hum" you have very good EMI RFI line filters who will help a lot
I have already explained that I will show the 230V side later, and yes that will also contain some filtering.

It's a classic circuit with no pfc correction.
Have you also read that the first transformer is against the saturation that you do not just solve it with a pfc circuit.
That quickly becomes too complex, it is better to use a good transformer and yes that is not perfect because of the double phase rectification.
Saturation and rectification, I do not think that is a good plan.
And why do I only come up with this now, because the transformer always felt warm and I have done more tests as shown here and naturally comparing it to the toroidal transformer.

This is not following datasheets design
Do you get insecure because I don't use the example circuits in the datasheet?
I have already indicated several times that this is a learning project and that it is not meant to be "build this kit too!" "Its all  purposes! power supply!" No it is not!
If the reader of this topic does not want to become aware of this, that is his problem, not mine.

Martin, please show us your design and some power on/of behavior, noise performance on the output and pictures of the inside of you power supply.
I have become intrigued. 

Kind regards,
Bram

[iMo]

As a big fan of 723 (I own 16 pcs of it) I've always been looking for an improvement of the Current Limiting circuitry. I see there the current source feeding the CL and the 1k pot. Could you briefly explain the benfit/improvement, plz?

PS: I built my first 723 source in the simplest form many many many decades back and used it for maybe 20y. And it worked fine, with an exception of the current limiting - I always had it set to a fixed max level, what was not good, as I smoked a lot of silicon stuff with it..

I want to build another 723 one, but with the option of CL setting of 10mA, 100mA, and full xxAmp. I think those are the typical ranges which cover most scenarios with hw experiments..
10mA - the smoke test
100mA - running my latest duino design
xxAmp - charging my batteries or running my drill machine

[coromonadalix]

Managed to re-find the link example circuit.

Here's what i tought for the current limitation by adding a powerdiode, just need to recalculate the sense resistor with the adjustment potentiometer for your needs.

Only miss an powerdiode between emitor collector of the power outputs transistors.

http://www.tonyvanroon.com/oldwebsite/circ/ps3010/ps3010a.html

And the unobtanium L146 regulator :
http://www.electroniccircuitsdesign.com/power-supplies-circuits/0-40v-lab-power-supply-circuit-diagram-electronic-project-using-lm723-l146.h

See the current limiter is before the collector output, that way all the voltage loss does not affect the output regulation.


But i'm not high jacking your thread, i just want to show you another current limitation design ... simpler, same functionality and less parts ?

[blackdog]

Hi coromonadalix, 

I think it is clear that I'm going for good performance and not "less" parts, or easy for someone els.
With all those designs that you and others show, dit you see any pictures of the dynamic behaviour under pulse load and or off/enabled behaviour.
More than 90% of all schematics or designs don't talk about that, ask yourself why?

All the points I have addressed so far relate to the best possible power supply to do with a 20V transformer and 20V output voltage at max 2-Ampere.
This means that I have to deal with a lot of points and some things I can't do as nicely as with a modern power supply.
And now I'm talking about current limitation again, and I don't want to do that anymore, it's good enough for many conditions and fast!

I work more than 50 years in electronics, almost every schematic shown about the 723 on this forum i already know.
That doesn't mean I know everything, No Way! but use these schematics to develop my own version and my specification, almost all of us are standing on the shoulders of our predecessors. 

Back to the schematic, my SPlan CAD software crasht, my own fault...
I had to draw all my changes again. 

OK,
The transformer and rectifiers are tested with the bleeder resistors fitted and the supply lines loaded with about the currents I expect to run there.
After testing the coil with the resistor to dampen the Q, I decided to remove it again, it brought too little improvement.
Lets do it "Old School" R11 22 Ohm and a second capacitor brings the ripple at this point more than 20dB down, and lose about 400mV DC



My scope does not have a 500uV/Div sensitivity, but I use one of my broadband preamplifiers.
This preamp is designed by me for measuring circuits that have a low output impedance.
Therefore the input impedance is only about 25K.
This is the schematic of the preamp.



Project page on: https://www.circuitsonline.net/forum/view/135863#highlight=preamp+40db


Rippple suppression of the LM7912.
This is the input "Yellow Trace" and the output "Blue Trace" of the LM7912 regulator under full load of the rectifiers, so there is also 2-Ampere from the main power line used.


Noisefloor preamp
The noise of the preamp with the scope in its most sensitive position and the largest bandwidth, this is about 2.5MHz at -3dB.
To show that I am measuring the noise from the regulator.


Later I will explain how the current limitation works, now no more time, I still have to do some work for my customers.

Kind regards,
Bram

[dardosordi]

Really looking forward to see your progress. I’ll would like to build it! Keep up the good work.

[blackdog]

Hi, 

Yes, there is a new version of the schematic, but i want to talk about some other stuf...

This is a picture of the box where the LM723 power supply will be built in.
I decided not to use the smaller heatsink and to choose a larger model.
This heatsink comes from a DELL computer, this brand used very nice cooling material, which I have already used for several projects.
The CD box is to better estimate the dimensions, it is a CD from New Order and Joy Division are English bands and one of my favorites.



Front view.



This will probably be the setup of the transformer, 10.000uF C5 fan and the heat sink.
Also note the bridge rectifier on the side of the heat sink, now still glued at the moment of the photo, but it will be screwed on later.


This is one of the fans I have in stock which I have tested for this power supply.
It has been tested up to 18V power supply voltage and I will offer it up to 15V.
Most fans have a fairly large voltage range, just look at the big brands in the datasheets.
But apply what you like yourself, this is a fan of not so high power that cools much better at 15V, this with little noise.
Of course, the fan wears faster when running at 15V for a long time, but this will not happen too often, only at low output voltages and at the maximum current this power supply can deliver.
It is not a production power supply, it is a power supply for my LAB Bench, for testing various circuits.
But tlater I tell you more about the fan and its control.



Here you see the bottom of the heatsink that is normally printed on the Intel processor.
First I mounted four nylon bushes to isolate the heat sink.
The Collector of the 2SA1943 is mounted directly on the heat sink for the lowest thermal resistance.
This helps with the efficiency of the cooling block and also helps to keep the chip temperature of the 2SA1943 as low as possible and eventually results in a lower fan speed and noise.
All parts that are screwed onto the cooling element are provided with a thin layer of silver compount just like a processor cooling.
This picture shows what I want to mount on the cooling block, that are many parts.
The white part is "clickson" of 50 Celsius which interrupts the 230V mains voltage when you exceed this temperature.
I have here the 50 and 60 degrees Celsius in stock.
There will be two of these "clicksons" in this power supply, I will also mount one on the transformer.


These are the parts that are most important and for that I drilled holes and tapped M3 thread.
The setup was chosen with care and in the end it took me two hours to come to this setup.
I wanted the wiring to be as short as possible and the two big heat sources, that is the 2SA1943 and the 4.7 Ohm resistor not too close to each other and not too close to the edge of the heat sink.
With this setup that I show here, I think I succeeded well.
Right next to R34 the 0.22 Ohm resistor, we assemble the driver transistor for the 2SA1943.
There is still room left for the MOSFet that will control the fan and another power resistor that will only dissipate 1-Wat maximum.
When I have determined the position of all the parts on the cooling element, I'll flatten the heat sink and also the 2SA1943 transistor for the best heat transfer.


This is the fan controller, I have used this schematic several times with good results.
I designed it so that the fan always runs at a minimum speed at which the fan runs reliably.
The noise level at this speed is so low that it can only be heard close to the fan.
I also chose the fan on Power On which runs at full speed for a short time.
I use a MOSFet to control the fan, the sensor is a small glass of NTC that is mounted in the Heatsink just below the chip of the 2SA1943.
R5 in the schematic sets the minimum fan speed.
If the NTC measures above a certain temperature, the IRF540 slowly starts to conduct.
 P2 sets the temperature at which the fan starts running at a higher speed.
Usually I set it in such a way, that around 35C heatsink temperature the speed increases.
Always running the fan ensures, among other things, that the temperature inside the housing and provides more DC stability.

Depending on the fan used and how well the NTC is connected to the heat source, the IRF540 can be replaced by an IRF840 for a low power fan.
The IRF840 is less Transconductance about 5S and the IRF540 is 9S.



This is the starting point for the 230V grid voltage side.
Here too I have to do some tests to determine the optimal component values.


This is it for today. 

Kind regards,
Bram

[xavier60]

It's a bit difficult to see, the wiring of the IRF4095 doesn't look right. Current flow would need to be into the Source pin.

[blackdog]

Hi xavier60,

Current flow would need to be into the Source pin Why?  i like "MagicSmoke!" 

That will be the third time I have to change this...
My brain is trolling me 

Thanks,
Bram

[blackdog]

Hi xavier60, 

It is incredible how blind I can be from time to time for errors I draw in the schematic and then also apply in building this power supply.
Of course, those mistakes come to light when I'm testing.
But of course they take a lot of time.
So I'm happy with your comments about the schedule and its setup.
To show even better how things go wrong, I'll show you here below how I first draw the schematic while thinking of the dissipation limiter.
No errors...
But it went wrong when I started turning the IRF4905 to fit it easy on the schematic sheet and forgot to mention the S, D and G for the connections.
I find it valuable to show the mistakes I make, because this is part of designing circuits. 



The last schematic, I hope I have drawn the dissipation limiter correctly now.



Then some info about the fan control.
On the breadboard just behind the heatsink is an IRF540/IRF840 MOSFet and the power comes from the Rigol DP832.
The middle output is set to 15V and the right output of the Rigol varies between 0 and +5V.



I like it when I slowly vary signals to read them via an analog meter.
I use a beautiful PHILIPS PM2505 meter here.
There was some dust on the meter scale because I hadn't used this measuring instrument for a while and I removed it with my finger and continued measuring...
Que! what is this, this is not possible. The meter indicated values that were not possible on the chosen measuring ranges, is it broken?
Of course I checked the batteries, which turned out to be good.
Measuring instrument switched off and then it indicates 5V!!!
Then I got an inspiration, this I have had more with analog meters with a plastic housing, static electricity..
I could vary the needle by keeping my finger closer or further away from the meter scale.
I cleaned the entire measuring instrument with screen cleaner and then the problem was over.
One of the reasons this happened is that the humidity value here is currently around 30%, it freezes a little here and the central heating is on.
These are bad conditions for electronics.



When I finished determining the control range I fed the fan from 15V and started experimenting with the distance between the fan and the heat sink.
If the fan is against the heat sink, the fan makes the most noise, I think this is because there is too much turbulence.
About 3mm distance is enough to get the minimum noise level with the fan I use and this heatsink.

Later this afternoon/evening I will test the heatsink and transistor with 60-Watt dissipation power to see what the cooling power of this setup is.
For this I drill an extra hole in the heatsink to mount a 5K measure NTC which will be connected to one of my Keysight 34461A measuring instruments to log it properly.
The energy I then let come from the Rigol DP832 Power Supply, then I can directly read the power consumption on the display.
I will show some pictures of that, just like the corrected wiring around the 4.7 Ohm resistor and the IRF4905 MOSFet.

Kind regards,
Bram

[iMo]

FYI - you may try with this precision current limit with log pot scale when keen on good quality
All you need is: dual opamp, 6 resistors, 1 diode, 1-2 caps, 1 linear potentiometer.

[blackdog]

Hi imo, 

I'm looking forward to your real hardware tests and scoop pictures of how your circuit is working.

Kind regards,
Bram

[blackdog]

Hi,

This time I show some measurement data from the tests I did on the heatsink.
This is the corrected heatsink with before testing some extra parts mounted which are removed after testing.
On top of the current measurement resistor is a 2N3904 mounted which sampled the voltage across this resistor.
The extra parts as they are applied here provide a current of about 2-Ampere.
I also hold a measure NTC here, it has a value of 5K and is connected to one of my 34461A multimeters where you can see the pictures of.
At the front of the heatsink you can just see the hole that is filled with thermal paste.
The 2SA1943 and also the cooling block have not been made extra flat before I started the measurements, but good thermal paste has been used.



NTC mounted in the hole I drilled for the fan control (100K NTC) which is exactly below the chip position of the 2SA1943.
After the thermal tests the 5K measure NTC is removed and then the 100K NTC is mounted in this hole.



You can see that I "care" about the measurement setup, a box of Qtips on the heat sink so that all air has to pass the fins.
It is also visible that the fan is about 3mm from the heat sink.



Here is visible that I now also measure the temperature of the heatsink finning.
This at 63.8-Watt and a fan voltage of 10V, desensor under 2SA1943 measured at that moment 49.7C
Another small improvement I got by the tape I used to close the 3mm space between the fan and the heat sink on two sides, this gave about 0.4C lower temperature under the 2SA1943.
I'm going to test if this is also advantageous when the heatsink is built in.



The K-Type sensor is tightly clamped between the heatsink fins.


The following measurements were taken at approximately 21C LAB temperature.
On this picture you can see that for a longer period of time there was a slightly lower temperature,
this was at the 15V fan voltage so this resulted in only a little extra cooling.
The final value of this measurement is with 12V fan voltage.



Bij 63.5-Watt, 8V Fan almost 31C higher temperatuur under the 2SA1943, the fluke meter gave here 41.2C as fin temeprature.



Here I have kept the fan voltage deliberately low at 6V and the power at 65-Watt to see what the temperature then becomes below the 2SA1943.
This is a fairly heavy test for the 2SA1943 which normally will not occur, because the dissipation limiter also helps to dissipate the power.
This also makes the heathsink's rendament better because then the dissipation power is better distributed over the heathsink.



40-Watt, 6V fan, 23c temperature increase and the Fluke measures 37C fin temperature.



30-Watt, 6V Fan, only about 17C Delta-C...



And the last one for this post.
30-Watt, 15V Fan, Delta-T =    and the Fluke measures 28.6C fin temperature.



My conclusion is that this heatsink is sufficient to use in this LM723 power supply.
I'm going to do some tests to see which "Clickson temperature" I need to mount and where.
I have to take into account that the temperature can be 38C in the box on a hot day.
So probably the mounting point will be on one of the outer fins of the heatsink.
During normal operation, the idle voltage (6V) of the fan control will be sufficient to keep the power supply cool and the fan virtually silent.

Shoot at it!

Kind regards,
Bram

[iMo]

Shoot at it!

[Wolfgang]

Off center. 

[blackdog]

Hi,

I do a lot of hard work and show it here, and the only thing i get is some form of humor?   

Kind regards,
Bram

[blackdog]

Hi,

Here two more pictures of measurements and then this time with the series resistor of 4.7 Ohm of the dissipation limiter enabled.
The dissipation in this resistor in these measurements is about 18-Watt.
The Fluke DMM that measures the fin temperature gave only small differences that, in my opinion, are covered by the measurement uncertainties.

The first picture I show now, is at the maximum power of around 65-Watt of the power section that occurs in fault situations such as short circuit.
The fan runs on 8V and the Delta-T is about 24C here and the dissipation limiter helps to reduce the temperature below 2SA1943 by about 7C.



And now the measurement with a normal fan voltage of 12V and the Delta-T is now 21.4C.
I like the performance of this part of the schematic and if the dissipation limiter also behaves well during the dynamic tests it stays in the schematic.



Also these measurements are not optimized yet, after leveling the 2SA1943 and the heatsink the chip temperature of the 2SA1943 will be lower.
This lower chip temperature of the 2SA1943 will increase the reliability of this power supply.
One of my starting points that I already indicated in the beginning of this topic, is that the power supply must be "kind" for its load.
And this is one of the ways I do this, a lot of power supplies go faulty with too slow current limits and too much dissipation in the power transistors.

Do not hesitate to ask questions or make comments.

Kind regards,
Bram

[Cliff Matthews]

Will you eventually post output stability graphs of when Q5 pulls 20w away from the output?
IMO this function at least deserves a front panel LED.

Q - Which components provide hysteresis for this?
Q2 - Does C4 provide a slow bypass "glide-in glide-out" function?
Q3 - Should this cct be off (LED on) briefly at start-up?

[blackdog]

Hi Cliff, 

Thanks for your good question.
It is my intention to show how well the disipation works more limited, if is works? 
As for the LED that shows when the dissipation limiter works, for me this is not directly necessary.
If it works well in the end, I can forget this part of the circuit under use.

But a red high efficiency LED over the 4.7 Ohm resistor with a series resistor gives at 1.8V already an indication that this function is working.
This will be around the 3-Watt disipation point in the 4.7 ohm resistor.

Q - Which components provide hysteresis for this?
The switching point of the dissipation limiter is determined by Q4 the 2N3904 and the zener at its base that sampled the output voltage.
Largely the threshold is determined by the Zener D1 which is provisionally set as 8.2V in the scheme.
Some values of resistors are not yet determined around 2N3904, this is determined during testing of this part of the circuit.

Q2 - Does C4 provide a slow bypass "glide-in glide-out" function?
Correct, the 2N3904 will have some gain and it is undesirable that "normal" variation to the output of the power supply (which will always be there,
especially with large current variations) that these variations modulate the Gate voltage of the IRF4905, and this is undesirable.

Don't forget C22 which will be mounted close to the 2SA1943.
This ensures that Q6 and Q8 see a low impedance, which is especially important when the Dissipation Limiter is working.
A high impedance at this point (Emittor Q8) deteriorates the dynamic behavior of the power supply, I tested this on another design that I measured extensively.

Q3 - Should this cct be off (LED on) briefly at start-up?
I do not understand this question very well...

But if you are talking about the current source with Q3, the RC time built up by R22 and C16, I still have to test how this finally works at boot time.
I want the "Power On Protection" to work for so long that Q3 can work normally when the set time is over.

PS,
Something went wrong in my browser again, this resulted in double last lines and a piece that was not posted.
For those who do not yet know, you can have a great advantage by mounting the power part directly on the heat sink without using insulation materials.
But beware, this means that the heat sink itself must be isolated!

I also took into account that the 4.7 Ohm resistor of the Dissipation Limiter is not mounted next to the 2SA1943.
As I said before, this is also done to improve the efficiency of the heatsink.

Kind regards,
Bram

[iMo]

FYI: I've done a simulation of your current limiter - with R21=2k2 in the current source emitter (I=860uA), shunt 0.22ohm as in your schematics v0.9 and the following resistances of your pot P-2, R32=0, output shorted with 0.1ohm:

Code: [Select]

P-2[ohm]   I_shunt
1000    0.65mA
900     0.72mA
800     0.81mA
700     200mA
600     588mA
500     976mA
400     1.36A
300     1.75A
200     2.14A
100     2.53A
With an 1N4148 in series with 560ohm wired in parallel to P-2 (anode to CL pin) and R32=100ohm, I get:

Code: [Select]

P-2[ohm]   I_shunt
1000     0.82mA
900     22mA
800     157mA
700     315mA
600     503mA
500     732mA
400     1.02A
300     1.37A
200     1.75A
100     2.14A
0       2.53A

[blackdog]

Hi imo,

Thanks for your input :-)
The first list of values you showed corresponds pretty well to the values I had with my first test print.

These values also show that the current source resistor R21 in the last schematic is chosen too low in value and therefore there is too much current for a nice control range.
This value was from the first version of this setup, in the last schematic you can also see that there is xxx at this resistor, which means that I will determine the value of those components later on.

Yesterday Farnell delivered some parts for me to test with.
This is another value for the current measurement resistor R34 which is 0.22 Ohm in the last schematic and I now have a 0.25 Ohm to test.
I hope to get closer to the Max 2.2-Ampere.

I will say something again today about the current limit and its temperature stability.
As I have already indicated several times, this is not a problem for my application.
I have already shown that I have taken a step to reduce the temperature drift of the power source, but this drift of this part is only a very small part of the whole temperature dependency.

Lets make a list!
The temperature sensitivity of the current limiting transistor in the LM723
The temperature sensitivity of the current limitation resistance R34.
The temperature sensitivity of the current source made with Q3 the BC557b, high Hfe PNP transistor.
The temperature sensitivity of the P-2 and R31 in the last schematic.
The temperature sensitivity of the copper wiring at a hot heat sink at low current settings.

The power source itself with Q3 could perhaps still be replaced by a type built with a TL431, which is more easily available for builders.
Here are two possibilities, a power source built according to an application note of the TL431, or LM336-2.5V replaced by a TL431 with a connected reference connection to the cathode.
I will use my modulable power supply to test what the differences are at a low power supply voltage to see if they have a high impedance in this situation.
Maybe you're asking yourself why?
The lower the impedance of the current source, the more the remaining ripple that is on the plus connection of C9 is injected into the current limiting transistor.
This is especially noticeable when the current limit is set to a low value.

Without using a clean reference and e.g. an opamp, it becomes difficult to provide an LM723 with a current limit that is reasonably stable over the entire current range,
such as a power supply design with two opamps.
I already knew this when I started this project, but there's nothing wrong with looking at it without using a extra opamp to make it so that it also meets your expectations a bit.

imo I will weigh/test your circuit that shows you with the extra diode and include it in the scheme if it gives an advantage.
But this will be a bit later, I want to determine the position of the Clickson on the heatsink first.
This Clickson helps with fault situations such as a defective fan and high dissipation in the cooling block.

On, off, abberations
I have some extra ideas to make this better.
This because in some circumstances this was not good enough as I have drawn it now.

What I'm going to do is the following: maintain the "control loop" as much as possible.
Q2 in the last schematic pulls pin-13 to the ground connection (pin-7) of the LM723.
And this works, but it does interrupt the control loop, and with some output voltages and currents this is not optimal.
I will place an extra resistor between the + of C14 and the plus input (pin-5) of the LM723.
And then there will be a MOSFet that goes from pin-5 (drain) to the minus of C14.
That is one of the ways to make sure that when switching on the power supply, the +input of the LM723 sees "0V".

The extra resistance that comes to the +input of the LM723 I will have to determine.
This resistance cannot be taken too high as this will cause problems with noise and bias currents.
If this resistor is taken too low, then the current from the reference output of the LM723 is too high around the maximum output voltage,
this results in a higher LM723 chip temperature which in turn results in more DC drift and this also affects the temperature stability of the current limitation.

Another way is to interrupt the top of the potentiometer P-1 with a MOSFet.
That there might still be a bit of DC on the output, is not important as this will only be a small voltage.

Electronics is fun and complex! 

Kind regards,
Bram

[blackdog]

Hi,

Yesterday and today I worked on the circuit to ensure that the power supply is switched on and off properly.
This is not all that is needed for an impeccable behavior of this power supply when switching it on and off and using the enable button.
If this part I'm showing here now works well, I'll look at what else might be needed so that no abberation occurs during a power cycle and enable use.

The first schematics were equipped with auxiliary circuits that used the compensation connection of the LM723.
If you pull this connection to Vs from the LM723, the power section will not get any current anymore and there will be no voltage on the output.
Well a little bit, this because the power source at the output brings the shootky diode over the output in alignment and there is max -0.2 a -0.3V on the output.
But I didn't think that was a big problem and I don't know if the circuit I'm showing here does a power off without abberations.
That is one of the things I still have to test.

OK, this circuit, which is visible below, is intended to prevent the switch-on abberations from occurring.
The second function is an enable switch which ensures that when the power supply is switched off with this switch, the output cannot become negative due to the 25mA current source.
Earlier I had already indicated that the "voltage loop" should be kept closed for good behaviour.

When using the compensation connection of the LM723, you interrupt this voltage loop and this always results in more abberations that you can not solve in all circumstances,
see my pictures earlier in this topic.

On the right a simple version of the LM723 schematic, I did this in order to make the changes more visible.
The power transistor is now drawn like a simple Darlington and can for know be ignored.




The diagram is drawn when the power supply is switched off and the Enable switch is on.
On none of the power lines is voltage at that moment.
If we now look at the right part of the schematic, we can see that the relay contacts are also drawn at rest.
The +input of the LM723 is connected via the 22 Ohm resistor to the "0" of the power supply.
The reference output is also connected to the power supply "0" by a 4K99 resistor, but there is still no voltage on the power supply lines.
This doesn't seem very useful, but I will explain the function of resistor (R7) later.

Now let's switch on the 230V mains voltage!
The LMM723 and the Blah-1 Power transistor get their power now quickly, I will measure later, how fast the buffer electrolytics have 90% of their charge.
While building up the voltage for the LM723 and the Power transistor, the +input of the LM723 is still at the "0".
At the moment that the LM723 starts working linear, this is from 9V supply voltage, then the +input of the LM723 is still connected to the "0" of the power supply and the output voltage is also "0 "volt.
But, where is the problem now? The point is if the LM723 is not yet linear, what does it do with the steering of the power transistor...
If I use the original uirgang of the LM723, then a resistor from connection 10 to ground will almost certainly help here.
If I am going to use the compensation connection to control the power section I will have to use an extra active component on pin-13 to get a nice behavior, we will see how it behaves.

If C1 in the left part is sufficiently charged and the SW-1 switch is closed, the relay contacts are switched over.
RE-1a switches from R7 to the potentiometer P1 and the 22 Ohm resistor is disconnected, If the potentiometer P-1 is not turned counterclockwise,
output voltage will appear at the output terminals of this power supply.
The output voltage will rise "slowly" due to the Potmeter P1 value and the R8 resistor.
C8 always filters the noise from the reference output and C8 has so two functions.
If the enable switch is turned off, the potentiometer P1 is switched off and the reference is now loaded with the resistor R7 which has the same resistance value as the potentiometer P1.
Also, C4 is now quickly discharged by the relay contact RE-1b via the resistor R9 of 22 Ohm.
This 22 Ohm resistor protects the contacts of the relay.

OK, why draw R7 in the schematic.
This is to keep the dissipation in the LM723 IC as Constant as possible, this reduces the DC drift.
The contact on the relay was free for this application, so why not! 
Now let's go back to the left part of the schematic, the enable switch has been removed and C3 shows that there is almost no bounching possible, this works together with the hysteresis of the relay.
I use the current through the relay to also control a red or green LED to see if the enable is on.

The relay used is a DIL-16 type of low power, that is completely closed.
This means that the service life will be very long.
The brands I usually use for these applications are: OMRON, Meisei and NEC in the 24V version.
To reduce the dissipation in the relay a series of resistors can be included, most of these relay types already work around 16V.
My minimum voltage at C9 is around 25V and maximum 34V as a resistor 4 a 5V reduction would be possible, but that is for later to see if this makes sense.
I also do not want to go too low in the relay coil voltage, as this also affects the speed at which the relay reacts.
Over the relay is a diode to counteract the EMF, but in series with the diode there is also a resistor that has about the value of the value of the coil itself, this resistor helps to accelerate the shutdown.

In the test shown below I took R2 loose and made it 1K, connected it to the output of one of my function generators and set it to 30Hz and a pulse width of 6mSec, 0 to 5V Top.
The blue trace is the pulse from the generator, the yellow trace is the 27V voltage which is also used for the relay coil and which goes through one of the relay contacts with a resistor to ground.
It takes almost 4mSec for the yellow trace to become positive, this delay is built up by the inertia of the relay and the 1K resistor together with the 0.1uF capacitor C3.
When the relay is switched off, it takes a little more than 2mSec if resistance R5 of 3K9 is included.


And this is the delay if the diode is switched directly over the coil, this delay is then 3.2mSec.
Furthermore, in this picture it is clearly visible that the contacts are bouncing.
This is especially the case in this picture at the rising edge.
The possible bouncing of the contacts has such a high frequency that it is well filtered by R8 and C4 at the +input of the LM723 and I don't expect any problems with that.



The trick with the series resistor is very useful here because the relay slows down a bit and is switched off by the C3 and R6 combination.
This means that no very high peak voltages are generated over the relay coil when the relay is switched off.

I want the power supply to go to "0V" as soon as possible when I set the enable switch to off.
A number of delays are added up to determine the final time.
That is this relay control and relay, the RC time of the potentiometer P1, R8 and C4 and at light loads, the 25mA current source, and the capacitors in the D.U.T that are located over the power lines.
And of course the biggest delay in "magic smoke" is my reaction speed... 

The only piece I haven't discussed yet, is the function around Q1 which is just like Q2 a small MOSFet with the type number BS170.
When the power supply is switched off, the voltage at the negative edge of C3 quickly goes towards the "0V", this lever gate control goes up for Q1 which then quickly releases RE-1.
With this I hope to get the output as good to "0V" for the power lines for the LM723 and the power transistor have dropped completely.
The timing which are determined for this function by R3, R4 and C2 I still have to test.
The 1N4148 should protect the Gate of Q1.
R3 and R4 are also chosen so that there is no more than 18V at maximum mains voltage at the gate.
Also the RC time C2 and R4 must be so long that C1 is emptied properly.

So the fine tuning of the parts values comes later, the power on delay as it is now, is about 1.5 seconds.
Which is determined by R1, R6 and the Vgs or Q2 (R2 is too small to have much influence).

Some of you may find it even more complex with the relay, but don't forget that this is a power supply for me
and I think it's important to pay attention to the detais to get the most out of this power supply regarding good behavior.

Now for the last parts I almost forgot to mention, that is an extra resistor to pin-2 of the LM723, I always thought it was already in the IC itself.
But after going through several datasheets and looking at the diagram of the structure, this resistor was never present...
This resistor is in the base of the transistor that regulates the current limitation in the LM723.

The last part is the diode1N4148 which goes from pin-4 to "0V", this protects the -input in case of a short circuit.
In case of a short circuit, the charge of C8 which 10nF in this schematic dumped into the -input.
At slightly higher output voltages, the -input may fall below the value of the -Vs connection of the LM723, this can damage this input, the 1N4148 prevent this situation.

Enough now, this was a long session to type this here, I hope you learned something.
And I'd love to hear your comments, additions and my stupid mistakes I made. 

Kind regards,
Bram

[Wolfgang]

Hi,

it probably works, but it looks a little bit too complicated for my taste.
The simplest (but extremely effective) method I have seen in an old GDR power supply was a DPDT mains switch. The "off" side drained the juice from the regulator and the output instantly via a small-value resistor and some diodes. If you dont like the switch, you could do the same via a line voltage powered relay.
After doing that, the problem left is safe startup. This could be done by the same relay as for switching off, but with a little delay for the "on" phase. While line voltage is already up but the delay is still running the resistors will shorten out any spike during this period.

[blackdog]

Hi Wolfgang, 

First, I don't find it complex, look inside a Rigol DP832 , HP6632a, DELTA Electronics ES030-5, HP6823 etc.
I have done extensive tests on "on and off" behavior of power supplies.
It is usually not easy to let the power supply behave well under different situations, as I have already explained a little and show some picture about my first setup.

Say you have for example the power supply on e.g. 3.3V and you switch it off and after two seconds it still dumps a 6V pulse in your load.   
And this one is also nice, which I have tested extensively with in another power supply, you plug the plug into the 230V socket, but you do not do that so precisely.
It is then possible that you give the power supply 5 to 10 times a power cylce within a second, and how does it behave then...
The behaviour is different for every power supply you test...
I have blown up a few circuits or damaged them by this unwanted behavior of the power supplys.

If you only use a power supply with loads that are not sensitive to voltage pulses during switch-on or switch-off, you do not need to investigate this in detail.
I do not want to have to think when I connect an LTZ1000a, or an LT1088, opamps that cost 30 euros etc, to the power supply that have on/off abberations that will damage these expensive parts.
But it's your party, you can connect every power supply to every load you test, I don't mind. 
I like to think about this kind of things en design it and when eventually 20% more parts are needed to be put into a circuit to make it perform exactly the way I want it to, who'll stop me?
I almost allway go for performance, if one of my low noise preamps need a 30€ opamp, I put it in the circuit, that gives me much more satisfaction than the behavior of "this is good enough".

Kind regards,
Bram

[Wolfgang]

... well, my "simple" solution of dumping all remaining charge after switching off prevents just that. For super sensitive stuff I always do it like this:
- power on the PSU and set params right, but let output disabled
- set limits, if paranoic install crowbars
- connect super sensitive circuit
- then enable output
- when turning off, same in reverse.

I had similar problems when trying to measure power supply output impedance with a VNA (see website).

[iMo]

A colleague of mine is testing his lab power supplies with the "red LED" method. He wires the red LED to the output terminals (none resistor in series), sets the current limit to 20mA and switches the PSU many times on/off randomly. When the LED survives he considers the PSU passed for further testing.

[blackdog]

Hi imo,

But the 10 million question is this one, on what voltage does he set the power supply?
2V, 5V, 13V, 35V?
The higher the voltage, the greater the risk that the LED will break or otherwise be damaged.
And don't forget that there are more and more LEDs made for small currents, 20mA can be too much.

What I mentioned earlier, if your colleague in my company would work and test power supplys this way, I would fire him because he draws too simple conclusions. 

Kind regards,
Bram

[iMo]

But the 10 million question is this one, on what voltage does he set the power supply?

An LED does not care about voltage but current..
What type of LED? - any one from your junkbox, even the most minuscule one, and "red"..
And sure, most PSU creators hate this test 

[blackdog]

Hi imo 

I'm afraid you didn't understand what I've written before...

The higher the voltage set on the power supply, the more energy is stored in the capacitor over the terminals of the power supply.
The energy in this capacitor is dumped into the LED and depending on the voltage and the solidity of the LED it will remain whole or will be damaged or will break down later.
Maybe it is better to understand if you see a LED as a zener diode,
above the threshold voltage of the LED, the current through the LED rises very quickly when it is connected to a voltage source (Low Ri).
This situation occurs for a certain time when the capacitor is discharged by the LED.

Even if the power supply is set to 5mA, there is still a good chance that the LED will break or get damaged.
The test your colleague does is not really intelligent, he doesn't seem to want to learn how things really work out, simple conclusions, bad technician. 

Kind regards,
Bram

[Cliff Matthews]

Nothing an always on, galvanically isolated $5 Arduino Nano and a few parts couldn't fix.. It has ADC's to quickly measure a few key points and can control fan and relays for both x-former power and output enable.

Optionally, it could be set to add brown-out/mains failure protection and have a hidden switch that turns off "auto power-up enable" if the voltage is not set below a certain threshold measured by the ADC (eg: the last guy to use the PSU forgot it was set high). This all can be done while still looking and acting like "old school".

[blackdog]

Hi Cliff,

It is not so easy to connect an Arduino to the power supply in a way that this microcontroller does not affect the operation of this power supply.
I also thought of the Arduino to control the timing of different parts in this power supply.
But don't forget that a number of things are beyond the control of the Arduino. (Think of a "brown out" lockt Arduino)

Of course I know that just like Wolfgang suggested with an extra switch contact or relay you can prevent many abberations in the different states.
But what happens, for example, when the Arduino software gets stuck?
Then I also have to think about building the electronics in such a way, that no thermo nuclear explosions wil occur. 

If you guys think think of PWM to control your fan, you don't know what a low noise power supply is...
bin there done that in one of my Electronic Dummy Loads i use a PWM fan controler.
Even with a full symetric amplifier stage that is well connected, I see disturbing pulses in my measurement signal from the PWM ventilator.
My analog control circuit in this power supply is a less efficient, but it is completely noise free and it is good.

I have already made many considerations for this power supply, such as an Arduino for the reading of voltage and current, and also as Cliff suggested to do the timing of different parts.
For the time being I will continue in the analogue way as I am currently working on, to make it as optimal as possible, nothing wrong with that...
And now and then I will process the remarks like those of Kleinstein and Wolfgang if these are good additions to the design for this power supply.

And don't forget, one transformer winding! low noise!, good dynamic behavior! low Ri!, No abberations!,  etc, etc.
The above requirement forces me to think very carefully, how can I solve it with a good 20V transformer 20VDC output voltage at 2-Ampere current
at very low noise and hum when using a LM723 and no expensive or complicated parts, without abberations in use occur, designing this, is brain candy for me.

Not only because it has to be a "old school" design, but because it has to be good, according to my specifications.
Because a measuring instrument is equipped with a processor, this does not mean that it is better. (yes the timing via a Arduino is more precies, but necessary in this power supply?)
Most of my power supply's are all analog, and 90% of the cases doing an excellent job.

And for other applications I regularly use the Rigol DP832 which is a modern power supply.
Usually these are applications where I also want to know the power consumption of the load.
I am always aware of the 470uF that is over the output terminals, even though I have set the DP832 to 3ma current limit.

Think of what Wolfgang told, in his remarks this morning, He shows that he is aware of what he is doing 
- power on the PSU and set params right, but let output disabled
- set limits, if paranoic install crowbars
- connect super sensitive circuit
- then enable output
- when turning off, same in reverse.

Kind regards,
Bram

[ZeTeX]

If you have X amount of capacitance at the output, and there is a switch from CV to CC, then I believe you could build a circuit to discharge the capacitor, so instead of the load discharging the capacitor, you would have something like a mini DC load inside the power supply. However, I believe it is not necessary and too complex for what you are trying to achieve.

I really like your design and I look forward to the finished project! you are very skillful and intelligent.
PS:
if you are using Chrome as your browser, you can add an addon called Grammarly: https://chrome.google.com/webstore/detail/grammarly-for-chrome/kbfnbcaeplbcioakkpcpgfkobkghlhen
it will fix your grammar on the spot!

[Cliff Matthews]

FWIW, the isolation I mentioned would be a small class-2 wall transformer with large enough capacitor to power the Nano, one relay coil, op-amp and maybe power an R2R driven 12-volt fan for 2 seconds (who said anything about PWM?). At least running the fan from the aux transformer would keep any fan BLDC pulses away from the main supply.

..you could opt for this second winding for that purpose, while still not invoking the Harrison curse 

[blackdog]

Hi Cliff, 

About the PWM Fan control, I don't mean that you made a comment about this, but on other forums one of the first things people start talking about is when a DC voltage needs to be regulated: PWM
In the case of power supplies, too, people immediately start talking about switching power supplies in order to improve efficiency.

Of course this is possible, but it is very difficult to get that noise and interference free.
That often results in a long design en testing time, and it is much simpler to use a transformer with multiple taps and a few relays.
You can then control the relays nicely by an Arduino, and if you solve it intelligently you can also measure the 230V mains voltage, mesaure the output voltage and the buffer capacitor voltage to determine the optimal switching point fot the relays. 

ZeTeX
Thank you for your kind words.
Chrome browser and also plugins that translate what I am typing wil not happen, but thanks for the link.
My Firewall does not allow this, more and more Microsoft, Google and social media websites are blockt here and i like it!

Dutch in my mother tongue and I have moderate dyslexia, in Dutch I can manage well with tricks, and my intelligence.
But all this quickly gets less when I get tired, or like the last few weeks that I often suffer from headaches.
So anyone who thinks there are a lot of mistakes in my English, bear with me, because posting in English is already extra difficult.
I can tell you that if I didn't already do my best now, only the schematics would be understandable.
These two things bother me daily, glasses and my dyslexia, live sucs

Kind regards,
Bram

[blackdog]

Hi!

Yesterday and today I was busy with the fan controler.
I made a little adjustments to the schematic and made it a little cleaner and some values have changed.
I decided to include the IRF540 in the schematic by default, because the differences with the IRF840 are small and a different heatsink, sensor position en the fanpower gives a bigger impact.

The modified fan controller.


I'll explain the circuit once more how it works.
First we have VR1 a 15V regulator, this is 3V higher than the rated voltage of the fan.
Most fans have no problem with that and there is only 15V present for a longer period of time on the fan in fault conditions on hot days.
Two Clickson temperature protections will disable the power supply under these extreme conditions.

Let's forget the MOSFet for a moment and only look at the fan and the resistor R3.
The fan I use is a "Low Noise" type and during testing, it already started up at 3V supply voltage.
By this I do not mean to connect the fan to 12V and then control the 12V back to the voltage that stops the fan.
I am talking about the voltage at which the fan starts reliably.

Because all the tests I did with the heatsink and transformer loose on the work bench, I have to take into account that the temperature in the closed box will be higher.
All heat dissipating parts are mounted on the heat sink, and the fan draws this heat directly out of the box.
The air coming in again partly flows through the slots in the box along the toroidal transformer.
At the top and bottom of the box are slots across the full width and the sides of the box are also heatsinks, which help to keep the temperature low.
So I chose the value of resistor R3 so that there is about 6V over the fan in the "idle" position en that is about 150 Ohm for my Fan!

As the temperature gets higher, the resistance of the NTC will get lower and lower.
With P1 and R2 you can set the point that the fan starts to run faster.
I usually measure this with a sensor close to the hottest point of the heatsink, and that is the hole I drilled, directly below the chip position of the 2SA1943.

For a good temperature control there are a number of things you need to think about to make it work properly.
Make sure you have a good coupling of your sensor, with your heat source.
Make sure your control electronics inherrent is stable.
Make sure that the heat sink and fan have sufficient capacity.
Just stick a sensor somewhere on the outside of a heat sink and assume it's good, that's not really wise.

Back to the diagram, R1 is there to keep the MOSFet HF stable, but I never really needed it, but it is there to be sure.
C3 is used when the power is switched on to ensure that the fan runs at full speed for 5 seconds and always starts.
C4 and C5 ensure that the fan always sees a low power supply impedance.

The 100K NTC next to a TO220 housing.
I put some plastic hobby glue on the white wires of the NTC, and then put the pieces of Teflon stocking on top of it and then I removed the excess glue.



Here the NTC is equipped with longer connecting wires and yellow heatshrink tubing for insulation.



In the heatsink I drilled one extra hole to mount both the NTC of the fan controler and the NTC connected to my 34461A DMM.



Yes, this extra hole is also filled with cooling paste.



Temporarily applied some PA tape to fix the wires for the measurements.
Later I do this in a neat way.



This is the fan control mounted on a piece of PCB.
Right on the vertical heatsink is the LM7815 mounted and in the middle is the IRF540.
Right next to the IRF540 is the 150 Ohm resistor mounted which determines the minimum fan speed.
Just above the IRF540 is the 50K trimpot for the set point when the fan starts to run faster.
I took some heatsinks from my storage boxes, they are almost unnecessary but to be sure I used small models.
The PCB is mounted in the cabinet on one of the walls, in a place where the air flows past.



Thats it for today, no more pictures 

Kind regards,
Bram

[ZeTeX]

Hi blackdog,
for the future, I can recommend you this fan:
https://www.arctic.ac/worldwide_en/arctic-f8-pro.html
as you see, it cost 6$ - very cheap, yet it is powerful (33 CFM / 56.1 m³/h (@ 2,000 RPM)) and very silent - 0.3sone - > 22.5dBA.
so you won't even need speed regulation since at full power it is already silent, and it is a reliable fan, with 'fluid dynamic bearing'. and in my case, it cost exactly like cheap fans I get in an electronics store that are much worse quality.

[blackdog]

Hi ZeTeX,


Thanks for the link, I think the Artic model wil not work well with my heatsink.
The fan i'am using know is also a low noise model and on 15V i kan hear it but at "normal" power levels, it is almost silent.
Other instruments in my lab have much more noise.

Lets not start talking about my Rigol DP832... 
I replaced the DP832 fan with a Low Noise type, because I found the noise level unbearable.
And when the DP832 was open, I also adjusted the wiring to improve the dynamic behavior.

A topic of mine about the Rigol DP832 on a Dutch forum: https://www.circuitsonline.net/forum/view/130740#highlight=dp832

Kind regards,
Bram

[ZeTeX]

Hi ZeTeX,


Thanks for the link, I think the Artic model wil not work well with my heatsink.
The fan i'am using know is also a low noise model and on 15V i kan hear it but at "normal" power levels, it is almost silent.
Other instruments in my lab have much more noise.

Lets not start talking about my Rigol DP832... 
I replaced the DP832 fan with a Low Noise type, because I found the noise level unbearable.
And when the DP832 was open, I also adjusted the wiring to improve the dynamic behavior.

A topic of mine about the Rigol DP832 on a Dutch forum: https://www.circuitsonline.net/forum/view/130740#highlight=dp832

Kind regards,
Bram

You are right, I sent the wrong link. Here is the correct one:
https://www.arctic.ac/worldwide_en/f8.html
or the same with 4-pin (pwn and tach output, just +1$):
https://www.arctic.ac/worldwide_en/arctic-f8-pwm.html

[2N3055]

...
And when the DP832 was open, I also adjusted the wiring to improve the dynamic behavior.

A topic of mine about the Rigol DP832 on a Dutch forum: https://www.circuitsonline.net/forum/view/130740#highlight=dp832

Hello Bram,
a question: did you ever perform dynamic testing to see if there is difference in DP832 behaviour after forming propper twisted pairs?
Thanks.

Regards,
Siniša

[blackdog]

Hi 2N3055,

I don't have the pictures anymore, sorry...
I only remember that by twisting the wiring in some other way the behaviour under pulse load became better.
So my modification consists of twisting the sense wires separately.
The coupling with the current-carrying wiring becomes less, originally the current-carrying wires are twisted together with the sense wires, this results in a "transformer".
The variation in the current-carrying wires induces an extra error signal in the sense wires.

In any case, I can tell you that tying wiring together is not always a good idea.
As a designer you will always have to think about what kind of coupling can occur between the bundled wires.
Rigol did one thing right at least, they have twisted the wires, it would have been best if they had done it as I have modified it.

But.... I wouldn't open the power supply to tackle the wiring now if you don't have problems with the power supply.
If you are going to do it, change also  the common connection between channel 2 and 3 to make the Ri much lower, as shown in one of my pictures. (I take no responsibility for your modifications!)

The only reason why I'll discuss this in more detail in this topic (It's not a Rigol mod topic ) 
is that I'm going to discuss the wiring of this power supply also to get best performance.

Back to the LM723!
And yesterday i dit some test on several LM723 to see how far the reference voltage deviates from the optimum.
And there is now a ST LM723 to test to see how much the drift is, at the first hours of operation a new LM723.
Like a good Reference IC, if you put it into operation to see what the first 1000H drift is.

For this I use a test circuit I made a few years ago, that delivers 20V output voltage.
See the picture below.
Left is the input en right is de 20V output


It is a bit modified LM723 circuit, because it has some extra noise filtering.
At 20KHz bandwidth the noise is less than 3uV <= This is not a typing error!
When I swap different LM723, I can easily see 1/F noise with this circuit, because broadband noise is no longer dominant.
The first LM723 I'm testing now, drifted almost 1mV down and 3.5mV up on the 20V output voltage in 14H.
This is as far as I can see now, it is not the temperature change in my LAB that causes this drift, but the settling of the reference in the IC.
The PCB itself with the resistors has been on for at least a month when I did the previous tests with it.

I have searched in my stock for better resisters to apply where it would make's some sense. 
That is with the -3.1V LM337 regulator and the resistors that set the gain of the LM723.
I advise nobody to do this, but I prefer these resistors in this circuit rather than that they are doing nothing in my stock.
So the 0.1% resistors in the next schematic that I'll show shortly, may be ordinary 1% resistors,
This because the LM723 and the LM337 will probably be dominant as far as the drift is concerned.

Times up...

Kind regards,
Bram

[2N3055]

@blackdog,
Thanks for explanation and sorry for offtopic!!
BTW, your posts are always interesting!
Regards,
Siniša

[blackdog]

Hi,

2N3055
It was a good question about the wiring and the advantage of the adjustments concerning the DP832, nothing to be ashamed of.
I was the one who suggested the DP832. 

I just try to make sure that others in this topic will now also discuss their DP832 or other power supplys which probably don't have any real value for this topic.
I would like to contribute to another topic, that for example is about the DP832, no problemo!

Too many topics on various forums get contaminated with remarks that have nothing to do with the original topic,
but that is normal human behavior and I to conrrol it a little, so that this will not happen here.
I spend a lot of time on this topic to make it as meaningful as possible, and my bad English is already disturbing enough in this topic.   

To show that I am still researching interesting LM723 schematics, I want to show an old one from a Dutch magazine "Radio Bulletin".
Here the designer "B. Th. Krol" shows a variation that Kleinstein or Wolfgang also brought forward.

This setup uses no extra negative voltage and uses the extra gain of the output stage in the LM723 to get more open loop gain.
This naturally requires more compensation to make it stable,  but no further data is visible how well this power supply performs.
I've never built a circuit with the LM723 in this way, so I have no experience with its dynamic behavior.

Anyone who wants to view this circuit can download the PDF, it is an "images" pdf, so difficult to get through a translation machine.
First you need OCR software to scan it to a text file, sorry...
www.bramcam.nl/NA/NA-723-PSU/Radio-Bulletin-Labvoeding.pdf

Kind regards,
Bram

[markce]

The 723 circuit enabeling adjustment to 0V and without negative supply is an active topic on circuitsonline.net
(in Dutch).

https://www.circuitsonline.net/forum/view/143967

[blackdog]

Hi Markce,

I know 

I would like to know how well the circuit in the Dutch magazine "Radio Bulletin" works.
One of its advantages is that no negative voltage is needed to control up to "0V".
But of course there are also a number of negative properties.
On this test board is now the "Radio Bulletin" circuit and later this week the circuit of this topic will be used for dynamic testing.
I do the tests on this circuit in order to gain knowledge which I can apply in the circuit of this topic.


"Radio Bulletin" 1e test circuit.



Building up the power stage, the yellow capaciors ar for decoupling.



Heat shrink tubing for insulation.



The transformer with which the first tests have been done, together with the bridge rectifier and two electrolytics.
You can also see that there are two bleeder resistors mounted over the electrolytics.



The first wiring has been done.



Here you can see that I keep the wiring of the potentiometers away from the thick red and black wire that carries the large currents.



Here is the wiring setup better visible.



There is also a 15V regulator attached to the buffer elco, whichts power the fan, in the red frame you can see the uA7815.
Do you also see the transformer? The connections are now wrapped with a number of layers of tape for safety during testing.




There is still some wiring to be done and the 0.47 Ohm resistor will also need to be mounted.
The capacitor over the output terminals is from another setup, but I use it for the first tests.
It is an ELNA of 33uF which is a bit low in value but I also have higher values of this good capacitor.
Also the BNC connector is visible which I use for dynamic testing and for Ri measurements of the power supply.
Always make sure that you are "safe"!!! Accidents happen faster than you think.


Maybe I have some time tonight to do the first tests.

Kind regards
Bram

[Kleinstein]

The  "Radio Bulletin" circuit is different in several aspects.
The way the negative supply is avoided is expected to work well - way better than with the extra negative supply. Though one might still need a negative supply to counteract leakage currents and provide some base load.

The second point is adjusting the FB divider instead of the reference side - this is old style and I would consider it more of a weak point.

The third is the output stage: this one looks a little odd with V_out (10) via a resistor to the positive side. So not sure this is right.

Current sense is at the low side. This may have a slight advantage when using a digital display, but other wise is not that different.

[blackdog]

Hi,

I tried to make a basic schematic of the version from "Radio Bulletin".
I am not really fresh, I have a good cold, Meh! 

But I hope this schemetic will make it a little clearer.
Did I draw any mistakes in the schematic, let me know!
The resistor connected to Vout injects a current into the zener so that this zener is always conductive.



Kind regards, snif, cough, pfff 
Bram

[Kleinstein]

Drawn this way the circuit makes sense. However the extra voltage gain could be a problem. The output stage is more like current controlling and thus needs some capacitance at the output as part of the compensation.
It might help to have some minimum current around the base of Q17.

[xavier60]

If necessary to reduce the gm, a resistor could be placed between Vz and ground as well as the resistor that needs to be between the Base of Q17 and Unreg. The 10Ω in series with the Emitter of Q17 will improve linearity.
The voltage loop will need type 2 compensation.

http://www.ti.com/lit/an/slva662/slva662.pdf

[blackdog]

Hi Kleinstein,

This is only a principle schematic no capacitors are in place.
I also find it strage that there is no BE resistor in the original schematic at Q17.
But there is this test setup for. :-)

In the original article it is explained that there is extra gain in the structure used and that this must be compensated.
But that is all that is being explained about it.
The problem is that the gain/phase response of the 2SC5200 may depend on the load.
I lowered the gain of the BD140 with the 10 ohm resistor in its emitter.

Some extra gain is welcome to lower the Ri of the circuit, but the phase and dynmic behavior should not suffer too much.
I look forward to seeing how the circuit works and measuring it.

Kind regards,
Bram

[Kleinstein]

To a first approximation the output stage is current controlling, a little like one has it in the floating regulators.  With a suitable capacitor at the output the regulator can still be stable.  The 10 Ohms resistor at the emitter looks like a good idea, especially with an extra resistor from Q17 base to the positive supply.

Alternative to to the shown output stage, one could use a very conventional NPN Darlington output stage. The extra voltage drop may be needed, as the output may not go low enough.

[xavier60]

This arrangement is known to work well. I recently found it to be stable with 1µF.
The gm starting at the Base of Q2 is 17. The transconductance is very linear above 200ma of load current.
Q2 would represent the output transistor in the LM723.
The faster 2SC5200 should improve performance further.

[xavier60]

Due to the simple construction of the op-amp in the LM723, the Comp pin voltage must never be required to drop below that of the Inverting input.

[blackdog]

Hi xavier60,

You hit the nail on its head!
After some adjustments of the voltage loop compensation, I connected the current limit loop.
What a mess that is in this schematic...
There is a latch point, and I think this is the effect you describe.

To find out some things, I took a separate transistor for the current limitation.
But one of the first adjustments I made to the original current limitation transistor was, to include a small emitter resistor of 22 Ohm, now the transmitter was at least quiet. 
Theni disconnect the transistor in the LM723 and uses a separate BC550 for this.
The same behavior! because I didn't realize then, that there was a form of latchup.

After some DC measurements without a current limitation loop was connected, I knew what the "normal" values in the circuit should be.
So I found out that the voltage of the compensation input can almost entirely go to "0V" with current limiting.
Then I put a 4.7V zener in the collector of the BC550 and after that the behavior became better.
But it is not yet as it should be...
Tomorrow I will increase the zener voltage a bit more and improve the values around the current potentiometer a bit.
I don't need 5mA as lowest curren at value 50mA is good enough for me.

The setup in this circuit is clearly different with regard to current limitation than the application note of the LM723.
In the application note you pull the current from the base to the output and not to V- so as you indicate a latch can occur.

Tomorrow more, the bad cold will terrorize me, time to go to bed. 

Kind regards,
Bram

[xavier60]

I wouldn't expect that forward biasing the C-B junction of Q12 in the LM723 to cause latch up. Hard to know what's really happening.
Page13 shows the Comp pin being pulled to ground via a 2K resistor.
http://www.ti.com/lit/ds/symlink/lm723.pdf

[IconicPCB]

Bram.

What would it take to increase current rating of the regulator, presumably  more 2SC5200 in parallel with appropriate emitter current sharing resistors.

How would this affect the stability of LM723 regulator?

Would it be reasonable to have a "universal LM723 controller" say with external add on power stages for increased current?

[Kleinstein]

For higher current it's mainly adding more power transistors. This would have only a slight effect on the loop gain (due to loading the stage before) and thus stability. The other effect is the changed emitter resistor - this may need to also increase the output capacitors to get a similar response.

However at much higher power it might be worth looking at things like transformer tap switching and maybe using a circuit with low minimum drop out.

[blackdog]

Hi IconicPCB, :-)

There are of course already many power supply circuits with one LM723 that can supply large currents.
See this topic as a learning project, for me and for those who do not have much experience in setting up and/or repairing power supplies.

At the moment I'm doing dynamic testing, and I could only do that well after repairing my Jim Williams Active Load which I had blown up a while back. 
That was about 5 hours of work to get it back in a good condition, it is almost finished now.
To get the HF behavior right, so nice flanks with few abberations you need a "Brick Wall" power supply, the circuit used by me is that of figure 6 in AN 104.
So I also did measurements on a test jig to make a very low impedance over a wide frequency range to calibrate the active load.
Jim explains this in application note 104 of LT about Active Loads,

So because I had to repair my measuring equipment, I didn't have that much time to work on the LM723 circuit.
I can say one thing, the current limitation of the Radio Bulletin circuit is not good.
This goes wrong because there is too much gain present and the circuit is therefore not very stable.

The problem is with increasing the current of power supply's is, that many features need to be adjusted.
With the larger currents, it becomes very important how your cabling runs for a nice dynamic behavior, how to apply effective cooling without long wiring.
You certainly need external sense wires that will also cause some problems with the dynamic behavior but also how to keep the sensor lines intact, with wrong wiring.
And then of course to limit the dissipation a pre-regulator or a transformer with taps.
Almost all pre-regulators ar EM transmittors,
Just look at the specifications of modern power supplys.
There is almost none that has less than 1mV noise, and let's not talk about the commonmode problem with the power supply's with a smps pre- controller.

It's all possible, but I make some more demands on my power supplies and that usually increases the development time and I don't have that time available.

I wouldn't use the LM723 for larger power supplys, then I take the Harrison concept, which has many more possibilities for a very good power supply than the LM723, which I still love as a chip. 

50 Amp power supply with an LM723 is of course possible, but I would never design that.

Kind regards,
Bram

[001]

sorry, i`m stupid
but what benefits of using LM723 in 2019? 

[HoracioDos]

sorry, i`m stupid
but what benefits of using LM723 in 2019? 

There was a quite big thread about that recently. I can't find it on my phone right now.

PS: This is it.
https://www.eevblog.com/forum/projects/ua723-voltage-regulator-viable-in-21st-century/

[blackdog]

Hi 001,

That's not a silly question, but did you read the whole topic, and what were my starting points?

Have you not seen that this is a learning topic, what all played a role in developing a power supply.
By taking an LM723 and doing some restrictions, you run into problems with what you need to find a solution for.

For those who think that a modern power supply should be equipped with microcontroller, they certainly have not understood it.
For those who want to start designing their own power supply, there is only one good starting point: voltage and current loop stability!
If you start thinking of digital meters, Arduino and other microcontrollers as a starting point, then you have not understood.

Good phase and gain margins are needed over the entire output range of the power supply.
If you can take care of that and that the power supply is still fast, then I mean a good dynamic behavior, only then can you start thinking about other parts of the power supply.
So I have started  a topic that I want to make a power supply with an LM723 that meets good specifications.

Kind regards,
Bram

[Cliff Matthews]

sorry, i`m stupid
but what benefits of using LM723 in 2019? 

Last year this uA723 thread was popular (it's 120 posts long.. and a very good reason why there is No other forum on the planet as good as EEVblog): https://www.eevblog.com/forum/projects/ua723-voltage-regulator-viable-in-21st-century/

[blackdog]

Hi,

"And Now for Something Completely Different"  Don't hit me with that fish!

A large part of the development of a power supply consists of testing how well it works under various loads.
For these tests I built a couple of measuring devices to test the phase and gain margin and two Active Loads.

This is my 200-Watt model



A look inside...



Schematic



The remark " troubles " with C9 in the schematic refers to the pulse behaviour during testing.
This is an MKP capacitor in the Active Load and in the case of pulses with fast edges this causes "ringing" because of the inductance of the connecting wires,
together with the capacitor which are over the output terminals of the power supply under test.
I have already received many comments when I showed pictures of dynamic tests done with this Active Load that my loop was not stable enough.
Even after I had already explained how things work out...

This Active Load I connect via 2x a Hirschman MLN banana cable of 50cm 32-Ampere and this is well twisted to make the induction as low as possible.
There is always a residue induction present and this with the steep flanks that why there is always resonance because of the 6.8uF and the residue induction.
This way of testing makes it extra difficult for the power supply to be tested.
The better the phase and gain margin the less ringing I see on the scope pictures.

But I also test with my other active load as well, it is connected as short as possible to the power supply to be tested.
This is a design by Jim Williams from the Application no 104, schematic-6.
www.bramcam.nl/NA/NA-723-PSU/an104f.pdf


This active load was broken by overheating, I forgot to turn it off... 
Here some pictures of repairing and building it.

That's how it looked 7 years ago when building.
Below the black connection is the 0.1 Ohm measuring resistor.
And below the red connection is the Power Mosfet mounted.
To the left of these connections is the Linear Technology LT1210 wide band opamp mounted.



This is the schematic with the addition of an adjustable voltage source so that the base current can be set between 0 and 1-Ampere.
I also use this current source for other types of measurements, because it is a fast current source for low impedances.



Seven years ago when he was just finished.



To calibrate the active load a very low impedance is needed, actually an AC short circuit up to 50MHz.
I used my capacitor test jig, which I use to test output networks for power suppies.
I had to expand the test jig a lot with various types of capacitors.
The left BNC connector is for the scope and the back BNC goes to the function generator.
In the middle you can see the 2x100 Ohm resistor which puts the generator signal on the test jig.
The generator is set to the maximum output voltage, which is 10Vtt at 50 Ohm terminated.



This side is connected to the active load.
The blue capacitors are of Nippon Chemi-con type 2A226 0N 26 and have a nice low ESR/ESL.



Here I connected the test jig to my spectrum analyzer, and after calibration the test jig gives this attenuation between 10KHz and 10MHZ.
There are three small resonance points that I don't find important because they are small and have a low Q.



Adjusting the active load, I will straighten the potentiometers when I am satisfied with the calibration.



This is a 100Khz pulse after the adjustment, better I can't get it better with the used MOSFet.
The first version of the MOSFet I used was a BUZ11 and when it went broke I looked up the specifications of the type Jim Williams used and looked for a more modern version.
I now use the 30N06L which has about the same capacities as the one Jim used but a bit more dissipation and current.
The problem with this MOSFet in the circuit is, that the properties (capacities) vary too much when adjusting Vds voltage.
This makes it difficult to set the three compensation potentiometers in such a way that the pulse behaviour is optimal over a large range.
Here a scope picture of a pulse frequency of 100KHZ and 10% Duty Cycle.
The current goes from about 700mA to 2.8-Ampere.
There is also a slight overshoot visible, this I can't get smaller without affecting other properties.



I used the double trigger of the scope when adjusting, this was handy because the calibration for the rising and falling edge are not the same.
This trick allowed me to see the edges at the same time.



The bandwidth measured over the current measuring resistor is as follows: 4MHz= -1dB, 5.2MHz = -2dB and 6.2MHz is -3dB which is more than sufficient for me.
I hardly ever use this bandwidth when testing power supply's.
Why not? This doesn't make much sense, as a few mm wire length already gives a fairly large error signal.
So a couple of years ago when I did a lot of testing, I decided to select a test signal and use it as a standard.
This is a 500HZ pulse with 10% Duty Cycle, and the edges have a rise and fall time of 5uSec
Look at the to pictures below for my standard test signal.



Now you can see the edge a little better.



And of course I also test with faster signals but I use the signal above most to get an impression how well the power supply performs.
If you have a fast regulating power supply then the left banana pins are no longer usable, they have too high induction at the maximum bandwidth of the active load...



Now I am ready to do dynamic tests and show them and to do phase margin tests on the LM723 power supply's.

Kind regards,
Bram

[Wolfgang]

Hi Bram,

I would say well done. As usual, some rants from my side:

- Linear MOSFETs please. Advantages: Less saturation voltage, no SOAR issues.
- What would be useful is an undervoltage detector. Had the same with other load pulsers.
- Another nice trick would be a SOAR check (idea: multiply sampled current and load voltage via analog multiplier, then check limit)
  Jim Williams had this in an early appnote, IIRC
- You could measure the current on the source side of the MOSFETs, so you have less voltage drop because you need the current sampling resistors only once.

Neveretheless, congrats !
  Wolfgang

[blackdog]

Hi Wolfgang,

No problemo, good rants are always helpfull! 

I used the 30N06L because Jim specified a logic MOSFet because of the smaller voltage compliance for the LT1210.

The power problem with the Jim Williams active load was my own dumb fault.
This active load is not made for high dissipation, I have a heatsink which I mount against the bottom of the box when I measure longer with large pulse currents.
I had forgotten this heatsink... 

Your trick with the four-quadrant multiplier is nice, I have here some Burr-Brown IC's and some models of AD.
But there is not much space available in the box, and making something outside the box is so messy again.
I have some low temperature Clicksons, I think I can find a place for them.

The Jim Williams active load has the measuring resistor in its source and I used it for the measurements.

Back to the MOSFet, I have some linear models here, but not in a housing (TO220) that fit in this design.
These are from IXYS and are TO-247-3 models.
Later maybe it is better to make the version with a BJT, it has no trim points, and the bandwidth is enough for most applications.

Kind regards,
Bram

[blackdog]

Hi,

Due to work and a severe cold I can't work so much on this LM723 power supply.
Today I had some time to relax and paid some attention to the temperature drift of the current limitation.
So my starting point today was to use a BC550C as a sense transistor and to keep it at a fixed temperature.
I have already built many types of ovens because I like this work. 
I wanted to keep the oven simple and compact, this means that the "gain" is less than a "normal" oven I usually build.

Of a good single oven that is well insulated, the gain will be between 150 and 200.
With some simple insulation this oven achieved a gain of over 40x.
This means that the variation of the transistor temperature will also be about 40x lower than a transistor that is not kept at a fixt temperature.

This schematic is new but based on something I built 6 to 7 years ago.
The LM317 is the heat source and the BS170 increases the current from the current source built up by the LM317.
R1 of 1K provides the minimum current through the LM317 and the resistor R2 and Rds of the BS170 determine the maximum current at which the current source thus provides the maximum dissipation.
This oven must be fed from a stable power supply, because the set point of the BS170 is dependent on this, I have in this power supply a -12V available hence the value you see in the diagram.
R3 is setting the temperatuur and the temperature also depends on the point at which the BS170 will conduct, this is a bit different for every BS170,
so R3 will give a different temperature for every oven that is built with this schematic.

The temperature drift of the Gate Voltage of the BS170 is no longer important because it is also kept at the right temperature.
The temperature of this oven does not have to be exactly 46.5C, only stable at a temperature between say between 42C and 52C.




Some pictures to show how I built this oven, this is the beginning.



Of course I built the first version with the wrong resistors. 



Now the BS170 is glued to the LM317.



I use some pieces of teflon sleeving here.



The NTC next to the BS170 is a 10K type, the NTC with some wire against the LM317 is a 5K model that is connected to one of my 34461A DMMs.
The BC550C transistor will be mounted at the location of the 5K measure NTC.



The two resistors on the left side are now 1K and 10 Ohm.
The yellow capacitor is for decoupling.
The wiring is very thin, this is done to keep energy loss as low as possible, which again gives a better gain.



The oven was wrapped in this piece of foam for the last test I did tonight.
This foam is from a sample IC packed from Analog Devices and it is isolated heat very well.



Fully wrapped in the foam for a number of tests.



Almost 1 hour of measuring time and only about 0.02C drift, lying open on the table.



A big problem with ovens is always the loop stability, with this oven this is no problem at all.
This is the current through the oven just after switching it on, no instability is visible.



And this is the temperature behaviour after switching on, also here a very nice behaviour without abberations.



Later I glued the BC550C on the spot of the measure NTC and I lay a layer of special thermal glue over the electronics, to improve performance.
Of course I'll show some pictures of that, when I'm done with it.


Times up...

Kind regards,
Bram

[blackdog]

Hi,

I would like to say a bit more about the mini oven I showed yesterday which is meant for the current limitation transistor in the LM723 power supply.
The transistor that performs this function in the LM723 is then no longer used.
By keeping the temperature stable, a reasonably stable minimum of 30 to 50mA output current can be used.

Why not put the whole LM723 in an oven, yes you can, but I had already decided not to do this.

This mini oven can also be used to keep other components at the right temperature.
Think of a TL431, LM329, LT1009, LM385, 1N829 series zener diode.
These parts made in a plastic housing and are the easiest to apply.
then there is no galvanic coupling with the oven electronics.

I have also checked whether if I use a more modern LDO or the number of parts will be reduced, this is not the case for what I have investigated.
I have looked at the LDO's of Linear, but then I miss the extra gain of the BS170 and more decoupling capacitors are needed, so that is not a good option.
It is possible to use one resistor less, that is the 10 Ohm resistor in the Drain of the BS170.
The peak current through the LM317 is then entirely dependent on the Rds of the BS170 and the current limiter in the LM317.

I have also looked at a larger housing for the LM317 or the LM338. The TO3 version is getting more and more difficult to buy and has a hefty price.
A TO3 housing does give some more building space if you're not used to building small electronics.

Some remark about ovens.
Use good isolation material.
Use when posible thin wiring (heat loss)
Use a tight coupling between the heat source and the NTC
Work neatly, ovens require the same way of working precisely as with voltage or current references.
Also build up your test circuits neatly, spaghetti wiring is absolutely undesirable.

The version of the oven that I showed here, uses at 12V supply voltage around 20mA.
That is about 250-mW, at 22C LAB temperature and about 45.5C oven temperature.

For those who would like to try this oven for their project, keep in mind the maximum LAB temperature.
In the very warm places on this world it is better to adjust R3 so that the oven temperature is above 50C.
If you lower R3 in value, the temperature of the oven will be higher.
It is also possible to take another value for the NTC, e.g. on ebay I bought glass  NTC's of 100K, which were very cheap.
Adjust the value of R3, with a 100K NTC you will use approximately 10 times as high a value for this resistance.

If a TO3 version of the regulator is used, then the larger surface area also increases the power required for the same oven temperature.

I hope you like this mini oven!

Kind regards,
Bram

[blackdog]

Hi, 

I have not forgotten you guys...

Work and the flu consumed all my spare time and energy.

But is is better now.
I emptied my workbench to do dynamic measurements on the LM723 circuits.
So a little patience please, it's coming!

Kind regards,
Bram

[blackdog]

Hi,

This is just the start of some measurements and how i do it to not fool myself or the readers of this topic.

Lets start whit the schematic/design i wil test first.
It is better to ignore the current limitation in this diagram, I'm going to solve this in a different way.

The two extra capacitors over the output are Ceramic and therefore have a low ESR and I chose this one especially here together with the Rubycon 100uF YXG capacitor.
This gave good results in the dynamic tests.
What I am now showing here is what is possible with the uA723 and a fast Power Stage regarding the dynamic behavior.

This is the starting point for me, the voltage loop should be good, before I continue with the other parts of the design.



And now some info about the current pulse I use for the dynamic tests.
I had already shown my Jim Williams Fast Current Source that I repaired.
This has now also been used in the tests below.

This is the voltage measured with a battery scope over the current measurement sensor in the Jimwilliams Current Source.
I use a battery scope for these measurements because of commonmode errors that often occur during these measurements.

This is the pulse measured at the maximum bandwidth and largest sample memory settings.
I show this to indicate that the used pulse is very clean and no signal will be masked if I filter this pulse in the scope to get rid of some noise.
The yellow line at the bottom represents the "0" current.
I modulate the Jim Williams Current Source with a 500Hz pulse from one of my generators and this pulse has a 10% Duty Cycle.
The level I use is from 200mA to peak 1.6-Ampere.
This pulse already has a much greater variation than most manufacturers use in their specifications.
And I also specify the edge steepness of the pulse, which here is 5uSec.

This is the pulse measured over the current measurement sensor in the Jim Williams Current Source at maximum bandwidth (200MHz) and memory depth.
Incredibly clean in my eyes.



And this is the pulse when I make the memory depth 1K and 16x average.
Now that the noise is gone, there are still no abberations to be seen.
I am now sure that what I measure is not influenced by a pulse current that would have abberations itself.



And this is the result, some of you might think, Bram that's not good at all...
And then I say, look at the scale settings of the scope, 2mV/Div. the pulse is about 7mV!!



But let's zoom in a bit to get a better picture of the pulse.
First I want to let you know, that when measuring with the Hameg scope I first looked again with a large bandwidth without averaging,
to see if there is something to generate at a high frequency, just to be save!
Yes, this power supply is FAST and stable. 
Of course the negative pulse is a bit faster than the positive pulse.
This is a 1 quadrant design that can only supply current and not draw current from the load or the output capacitors.


And now the scope is at 1mV/Div. to better measure the voltage drop in the middle of the picture.
The cursor measurement indicates that the voltage in this piece has dropped 720uV.
My Delta-i is 1600mA -200mA = 1400mA and that gives an Ri of about 0.5 mOhm of this power supply.


The attentive reader might think, He Bram, your scoop clips! that wil gives measurements errors!
Because of that, I also measured on the 2mV/Div. mode, this gave the same results only a bit worse to read.

Setup on the bench.


The grey wire in the foreground is the "0" sense wire.
The yellow thin wire is the +Sense wire.
The BNC connector is located at the points where the sense wires are.
The BNC cable is placed at right angles to the current-carrying wiring, this is done to minimize measurement errors through the magnetic field of the current-carrying wiring.



The measurements are done at the point where the sense wires are mounted on the output terminals.
Why not measured at the output terminals themselves?
Because I will measure the induction and resistance of the output terminals and not the properties of the circuit itself.

And what happens then when I connect 1M cables to my D.U.T. to provide energy.
Go and measure that, but prepare yourself for some horror.

You always need decoupling of the D.U.T. of sufficient size, so that the connection cables do not receive fast signals to process.
It can also help to use a power supply with external sense wires.
But really fast power supplus with sense wires are difficult to make, the wiring acts like inductors, and the compensation can become rather complex.
The first step for all currents larger than 1-Ampere, is the wiring twisting between the power supply and the D.U.T.

Now is the time to enjoy the first spring sun!

Kind regards,
Bram

[iMo]

The dynamic params of that 723 PSU - compared to various 2xopapms based PSUs people elaborate here today - should be better, indeed, as the opamp inside the 723 is with much lower gain, probably higher BW and consists of maybe 6 transistors..

Your SlowCC comparator consists of a single npn transistor.

Compared to 2xLM324/LM358/TLC072 the 723 is a pretty "empty box"

[blackdog]

Hi imo,

I do not onderstand your remarks...

Kind regards,
Bram

[iMo]

"Dynamic parameters of the above 723 based PSU will always be much better than the dynamic parameters of PSUs based on 2xopamps".

[blackdog]

Hi imo, :-)

"Dynamic parameters of the above 723 based PSU will always be much better than the dynamic parameters of PSUs based on 2xopamps".

This is only correct if you use slow opamps and slow power stages...

My "big" power supply design is even faster than this uA723 design.
But that design is equipped with fast opamps and the best result I had with opamps in that circuit that were 1000V/uSec.
That resulted in only a few mV variations and within a few uSec corrected again at a 10A pulse.
It was not possible to measure the Ri of that power supply in the same way as I did with this uA723 design, this is because the Ri was very low.

Back to this design, I had forgotten to mention that the hum and noise level at 2-Ampere output current is around 10uV RMS, yes uV... in a 22Hz to 22KHz bandwith.

Do these values make this a good design? It's a start...
Power On/Off may not create abberation that can hurt the load.
Current limiting must be fast enough.

I will say it again, current limitation is for me a function of a power supply that protects the power supply and the load.
I have a digital power supply where I can set the current per mA, I have never needed this in three years of use.
Of course the current was set but whether it was 50, 53, 46 or 60mA was never important.
I choose when I power a device the current slightly above the expected current that will be used, so with me a power supply is a voltage source with a current limiter on it for protection.
When I need a current source, I do use a current source device for that.
Power supply's are generally bad current sources with a very low AC impedance and often also a high hum current.

However, everyone can use their measuring equipment as they like it.
I have known for a long time that I regularly look at this in a different way 

The next measurements I will do are the ones that show how the switch on and switch off behaviour is of this circuit.
And if the behavior of the circuit is good enough for me, then I will do the measurement to the current limitation part.
And yes, the current limitation will be done with 1 transistor and this transistor will have an oven to reduce the temperature sensitivity of the current setting a lot.

Stay tuned 

Kind regards,
Bram

[m3vuv]

so where are these supposed schematics?,nothing showing here!!!!.

[m3vuv]

What schematics?,none show up for me,did they vanish in the server fire?,can you repost,im following this but all the op attachments have dissapeared from  this thread,73 m3vuv.

[Kartika]

Hi I  looking for Dual power supply with lm723?