Need tips on building a variable power supply

[jah118]

Hi fellows

I am trying to build a variable power supply, with an adjustable Current Limiting feature.

I have an old schematic on a design of a variable power supply(first attach), using a lm301 as the ic

with that schematic in mind i tried to make my own, but i have to replace some components, because i will be using a transformer  with an output of 30v 10A, i and i am afraid when the
parts arrive, it will not work because of my inexperience.

so my first draft of converting the schematic to fit my requirements(because of the one i am basing my design off, is only up to 2A):

2 boards in one unit, which can handle up to 5A each and adjust its current and voltage. So i have 2 small power supplies in one box.

with that in mind, and i have no idea how get current limiting part to work as intended. i decide to just make a schematic for the voltage regulating part,
but i have kept the current limiting part, by adding a screw terminal with the output needed to wire that section to voltage schematic. while i figure out to get right values and parts for the  current limiting part.

first draft of a schematic (second attach)

when i made the first schematic i did not take into account, how to cool the 3055 transistor, so with that in mind i have changed it for a screw terminal, and brought a large heat plate for it to be mounted in.

so redid that part an have now made a new schematic, with a board designed that is maybe ready to be ordered ordered.(last 3 attach )

is there any thing i can improve on, or does anyone have an idea to what part i need for the current limiting part, my main problem with it, is that Resistor6 and resistor 7, is the hard part, beacuse
how big does the effect Resistor(6) need to be to handle the full load. And how would i make a circuit, that short its self so i can set the max current there can be drained? after it has been set.   


Thanks

[mvs]

Do you wish to built your own design, or it is ok for you to use some DIY Kit?
On ebay, aliexpress and some other chinese shops you can find 30V 3A kit for linear power supply (pcb + components) for around $5. It is based on TL-081 opamps, schematic is in attachment.

https://www.aliexpress.com/item/DIY-Kit-0-30V-2mA-3A-DC-Regulated-Power-Supply-Continuously-Adjustable-Current-Limiting-Protection/32903085230.html

[David Hess]

10 amps at 30 volts is beyond what a single power transistor can handle reliability without fold-back current limiting; 20 volts at 1.5 amps would be more realistic so I suggest building a lower power supply to gain more experience first.

The design as shown is certainly workable although the current limit is pretty basic.

[mvs]

i will be using a transformer  with an output of 30v 10A

You mean 30V AC RMS? It is too high for both designs. After rectifier it will be >40V. Power supply limit of LM301 and TL-081 is 36V, and you need also acount for negative supply rail (2.7V or 5.1V) and some safety margin.

[wasyoungonce]

Jah...do you have a link to the original PSU page?

Thanks

[spec]

Hi jah118

Just in case: Having a transformer with a much higher current capacity than the PSU circuit can handle is no problem. In fact, from a circuit performance point of view, it is a positively good thing.

But, as has been mentioned, the raw supply rail will be 40V peak, which is more than the LM301 can handle, but the solution to fix this is quite simple. The most basic method is to put a 6.8V zener diode in series with the opamp positive supply pin, but there are better ways.

You could completely eliminate the negative supply and improve performance, by changing to a more modern rail to rail input/output (RRIO) opamp, like the OPA192. http://www.ti.com/lit/ds/symlink/opa192.pdf

At 1A current drain the ripple voltage across the reservoir capacitor will be 5V peak to peak. It would be better to at least half the ripple voltage by doubling the value of the reservoir capacitor to 4700uf.

Generally speaking, without going for fancy cooling and expensive transistors, you can only dissipate 20W in a power transistor in a TO3 or TO247 case mounted to a large heatsink. This means that your PSU current should be limited to 0.5A (500mA) (40V * 0.5A) . Of course, you can always use two power transistors in parallel. The maximum junction temperature is the limiting characteristic and the 2N3055 is particularity good in this respect, with a maximum junction temperature of 200 degC, but the thermal resistance junction to case is poor at 1.52 degC/W. https://www.onsemi.com/pub/Collateral/2N3055-D.PDF

The other thing you need to heed is the safe operating area (SOA) graphs on the power transistor data sheet. Look for the DC curve and check what current the power transistor can candle with 40V between its collector and emitter. The 2N3055 can handle 3A at 40V which is excellent (SOA is completely separate from maximum junction temperature).

It is advisable to connect a 100 Ohm resistor between the base and emitter pins of any 2N3055 output transistors. The resistors only dissipate 100mW so high-power types are not required.

You will also need a few decoupling capacitors to help ensure that the PSU does not oscillate. Most important is a 100nF X7R dialectic ceramic capacitor from each opamp supply pin to 0V as close to the opamp pins as possible.

You can also replace the MOSFET constant current generator, by a 33k resistor.

If you will be using a basic zener diode for the voltage reference, it would be better to change to a 5.6V or 6.2V zenner. That is the voltage where zener diodes operate best. But rather than a bare zener diode it would be far better to use a proper voltage reference diode/chip for a slight increase in cost. The voltage reference can be any voltage you like. But the feedback resistors will need to be altered to suit. Calculating the new values for the feedback resistors is simple.

And finally Keep wires as short as possible, especially the wires to the power transistors. Once again, this is to aid frequency stability.

The two biggest fails with home-build PSUs are frequency instability #1 and #2, over-stressing the output power transistors.

[not1xor1]

It looks like every day there is a new thread about a beginner willing to build an unrealistically featured PSU based on various dubious schematic diagrams, at best outdated, at worst with no chance of working.

I think I've found a good solution for a beginner PSU. It doesn't require any obsolete or hard to find parts, just a single winding transformer (for low power version), while providing good regulation of both current and voltage.

It is based on just one LT1013 which is not too expensive and works up to the negative rail, both on input and output and 1/2 cheap bandgap voltage references.
A charge pump is used to get a higher positive voltage rail to let the LT1013 go up to the positive ground.
Optionally the higher positive voltage rail, regulated by a 78L05 or LM317 or similar, might be used as voltage reference.

I do not have much time at the moment to explain all the details and the possible modifications to adapt it to various voltages and powers, so here is just an over simplified circuit to show the principle of operation.

Later or during the next few days I'll attach all the .asc needed for both voltage and current loop gain analysis, etc.
BTW would a new thread in the project section would be more appropriate?

[cdev]

It seems to me that for a beginner, having a good basic power supply with the common voltages - (even fixed voltages, and the top voltage doesnt need to be that high, even 20 volts is quite adequate, similarly for current, 3 amps is more than enough for most beginner projects).

Where the value is most added is in variable current limiting, and there the more control the better.

A nice 10 turn pot for setting that limit accurately, without difficulty, and a meter plus lights (plus buzzer, even), visual indication if the current limit had been hit previously, even just momentarily, since it was turned on, or was currently active (light plus switchable beeper or similar, not loud, just audible), would be super useful.

[mvs]

BTW would a new thread in the project section would be more appropriate?

If you plan to implement and verify your design in hardware... maybe yes. Otherweise it makes not much sense.

There is almost nothing special in your design... emitter folower and 2 error amplifiers. LT1013 is also strange choise, since it is quite slow and has odd pinout for dual opamp.
The only thing i liked, is usage of Sziklai pair instead of Darlington.

just my 2 cents

[spec]

It looks like every day there is a new thread about a beginner willing to build an unrealistically featured PSU based on various dubious schematic diagrams, at best outdated, at worst with no chance of working.

I don't see how this sweeping statement applies. Neither can I see any reason for it. The OPs circuit is quite satisfactory and simpler than the circuit you have outlined. It is these kinds of statements that confuse and discourage newbies and forces then to seek help via PM to avoid the flac. On one forum I was doing more PM advice than forum advice because of this kind of statement.

[spec]

BTW would a new thread in the project section would be more appropriate?

If you plan to implement and verify your design in hardware... maybe yes. Otherweise it makes not much sense.

There is almost nothing special in your design... emitter folower and 2 error amplifiers. LT1013 is also strange choise, since it is quite slow and has odd pinout for dual opamp.
The only thing i liked, is usage of Sziklai pair instead of Darlington.

just my 2 cents

The Sziklai pair is not advisable. While its DC performance is superior on paper, it suffers from a poor frequency response and very poor turn off characteristics, which can lead to odd behavior and frequency instability. The Darlington pair although less good from a DC point of view is much better from a ton/toff and frequency point of view.

And that is my point of view

 

[David Hess]

The Sziklai pair is not advisable. While its DC performance is superior on paper, it suffers from a poor frequency response and very poor turn off characteristics, which can lead to odd behavior and frequency instability. The Darlington pair although less good from a DC point of view is much better from a ton/toff and frequency point of view.

In my experience, the Sziklai pair is more prone to parasitic oscillation with difficult loads unless local frequency compensation is used so the design and testing requirements are more stringent.  (1) Other than that, the difference in performance between the Sziklai and Darlington configurations is insignificant in most cases.

Where this would matter the most is a power supply design which includes the minimum of output capacitance for good current limiting performance.  The low output capacitance makes frequency compensation of the feedback loop and Sziklai pair more difficult.

In the example below from the Tektronix PS503A bipolar output power supply, the top Sziklai pair has an extra 0.01 microfarad capacitor to ground presumably to prevent local oscillation and the 50 microfarad output capacitors are integral to maintain stability.

(1) Essentially a transient response or network measurement should be made on the Sziklai pair to verify stability independently of the tests made on the error amplifier's frequency compensation.  With some experience, one can get an idea of what is required to stabilize the Sziklai pair; usually just a small capacitor is required for local feedback.  This is a good place to use an RC substitution box during testing.

[spec]

The Sziklai pair is not advisable. While its DC performance is superior on paper, it suffers from a poor frequency response and very poor turn off characteristics, which can lead to odd behavior and frequency instability. The Darlington pair although less good from a DC point of view is much better from a ton/toff and frequency point of view.

In my experience, the Sziklai pair is more prone to parasitic oscillation with difficult loads unless local frequency compensation is used so the design and testing requirements are more stringent.  (1) Other than that, the difference in performance between the Sziklai and Darlington configurations is insignificant in most cases.

Where this would matter the most is a power supply design which includes the minimum of output capacitance for good current limiting performance.  The low output capacitance makes frequency compensation of the feedback loop and Sziklai pair more difficult.

In the example below from the Tektronix PS503A bipolar output power supply, the top Sziklai pair has an extra 0.01 microfarad capacitor to ground presumably to prevent local oscillation and the 50 microfarad output capacitors are integral to maintain stability.

(1) Essentially a transient response or network measurement should be made on the Sziklai pair to verify stability independently of the tests made on the error amplifier's frequency compensation.  With some experience, one can get an idea of what is required to stabilize the Sziklai pair; usually just a small capacitor is required for local feedback.  This is a good place to use an RC substitution box during testing.

Very informative

[not1xor1]

BTW would a new thread in the project section would be more appropriate?

If you plan to implement and verify your design in hardware... maybe yes. Otherweise it makes not much sense.

There is almost nothing special in your design... emitter folower and 2 error amplifiers. LT1013 is also strange choise, since it is quite slow and has odd pinout for dual opamp.
The only thing i liked, is usage of Sziklai pair instead of Darlington.

just my 2 cents

I do plan to build the real circuit.
I showed just the simplified one, but I've been running thousands of simulations on this kind of circuit (with models of real parts) in order to check all the possible conditions (voltage/current control loop gain stability, start-up, switch-on/off, remote sense, fine/coarse regulation, short-circuits, master/slave tracking, transformer-tap switching, etc.)

LT1013 speed is more than adequate for a power supply, and in particular mode for a beginners' one.

In any case I'll be glad to know about any other faster opamp with the following features:
- cheap and easy to find as LT1013
- possibly in DIP package (some old guy like me cannot struggle with those small SMD arthropods )
- low voltage offset/drift
- maximum supply voltage >= 44V
- ability to work with input/output at least close at one of the supply rails without phase reversal
- and yet I might have forgotten something 
So please suggest one.

The most funny thing is the hint to the "odd pinout" of LT1013... I just checked... most dual opamp I got here are the same:
AD8066, LM358, TL0*2, MC3**02, microchip ones, NE5532... and so on... You must be kidding 

BTW I just noticed you suggested the usual bad ebay kit.
That is one of those ugly circuit design I was referring to.

While in most cases TL071/4-TL081/4 do work with input voltages close to the positive rail they are not granted to do that.
E.G. for ±15V supply the datasheet grants only ±11V of input voltage range and AFAIK there is no mention about phase reversal protection.
Besides that TL0* output range is much more limited than that of LT1013 (3V less than supply rails) and so is the maximum supply voltage.
I would not comment further on that other weak design choices as that was already discussed here in past.

Advantages of the circuit I propose (some are evident only in the completee circuit I show later):
- it is cheap and yet offers much better performance than many other circuits you find on the net
- it is simple to understand and to build for a beginner
- needs only one transformer winding
- no need for an additional differential amplifier (only one opamp in the loop of the current control)
- the current control reference is supplied by a basic constant current circuit that also works as down-programmer without loading the current sense resistor
- the ability of LT1013 to work with voltages close to the negative rail allows a really low dropout
- probably the cheap dual V/I panel meters sold on ebay would fit easily and might even be used directly for the current sense (unfortunately I do not have one here to ckeck)

Of course I'm not pretending that it is the best circuit.
There are much better circuit using more parts and separate rails for the control circuit (Harrison's design) widely discussed, for instance, in the HP's DC power supply handbook.

[not1xor1]

It looks like every day there is a new thread about a beginner willing to build an unrealistically featured PSU based on various dubious schematic diagrams, at best outdated, at worst with no chance of working.

I don't see how this sweeping statement applies. Neither can I see any reason for it. The OPs circuit is quite satisfactory and simpler than the circuit you have outlined. It is these kinds of statements that confuse and discourage newbies and forces then to seek help via PM to avoid the flac. On one forum I was doing more PM advice than forum advice because of this kind of statement.

What I wrote was meant as a general statement.
In any case the OP circuit is outdated, not much simpler than mine and with the usual coarse single BJT current limiter rather than a real current control.

[not1xor1]

The Sziklai pair is not advisable. While its DC performance is superior on paper, it suffers from a poor frequency response and very poor turn off characteristics, which can lead to odd behavior and frequency instability. The Darlington pair although less good from a DC point of view is much better from a ton/toff and frequency point of view.

In my experience, the Sziklai pair is more prone to parasitic oscillation with difficult loads unless local frequency compensation is used so the design and testing requirements are more stringent.  (1) Other than that, the difference in performance between the Sziklai and Darlington configurations is insignificant in most cases.

The sziklai pair is not relevant.

The circuit works and has been simulated for loop gain stability and transient response, both with sziklai and darlington pass elements and with an assortment of power transistor models.

Real circuits of course are affected by parasitic capacitance/inductance and are another matter of fact.

Nothing so difficult to handle, but in any case that is the reason why I'll show a complete circuit with various possible options only after I'll have built and tested a prototype.

[mvs]

I do plan to build the real circuit.
I showed just the simplified one, but I've been running thousands of simulations on this kind of circuit (with models of real parts) in order to check all the possible conditions (voltage/current control loop gain stability, start-up, switch-on/off, remote sense, fine/coarse regulation, short-circuits, master/slave tracking, transformer-tap switching, etc.)

LT1013 speed is more than adequate for a power supply, and in particular mode for a beginners' one.

Perhaps we have just different expectations and needs. I have simulated similar circuit with LT1013 and CC-CV transient response was really bad.

In any case I'll be glad to know about any other faster opamp with the following features:
- cheap and easy to find as LT1013
- possibly in DIP package (some old guy like me cannot struggle with those small SMD arthropods )
- low voltage offset/drift
- maximum supply voltage >= 44V
- ability to work with input/output at least close at one of the supply rails without phase reversal
- and yet I might have forgotten something 
So please suggest one.

To get good DIP parts is difficult nowadays... Look at MC34072, perhaps it will fit your needs...

The most funny thing is the hint to the "odd pinout" of LT1013... I just checked... most dual opamp I got here are the same:
AD8066, LM358, TL0*2, MC3**02, microchip ones, NE5532... and so on... You must be kidding 

Pinout of LT1013 is normal in DIP-8 package, but in SO-8 it is very unusual.

BTW I just noticed you suggested the usual bad ebay kit.
That is one of those ugly circuit design I was referring to.

Design is quite old and it was not made to be beautiful... Voltage error amplifier connected after current error amplifier do not saturate, like in your simplified design. This improves transient response in CC->CV transition. TL081 are also a bit faster, then LT1013.

While in most cases TL071/4-TL081/4 do work with input voltages close to the positive rail they are not granted to do that.
E.G. for ±15V supply the datasheet grants only ±11V of input voltage range and AFAIK there is no mention about phase reversal protection

Circuit design limits input range of voltage err. amplifier to Vee+5.1V | Vcc-4.2V, other amps are even more limited, so no problems hier.

Besides that TL0* output range is much more limited than that of LT1013 (3V less than supply rails) and so is the maximum supply voltage.

Nobody is perfect, cheap TL0xx opamps are not always better then expensive LT parts.

[David Hess]

Pinout of LT1013 is normal in DIP-8 package, but in SO-8 it is very unusual.

It took me 20+ years to learn where that weird SO-8 dual operational amplifier pinout came from and I had to figure it out myself.  The large LT1013 die is rectangular and has to be rotated 90 degrees to fit inside the SO-8 package.

[spec]

Pinout of LT1013 is normal in DIP-8 package, but in SO-8 it is very unusual.

It took me 20+ years to learn where that weird SO-8 dual operational amplifier pinout came from and I had to figure it out myself.  The large LT1013 die is rectangular and has to be rotated 90 degrees to fit inside the SO-8 package.

You learn something new everyday

[not1xor1]

Perhaps we have just different expectations and needs. I have simulated similar circuit with LT1013 and CC-CV transient response was really bad.

Thanks for having highlighted the problem. 
At 7.5V output, on short-circuit recovery, I got about 22% overshot with LT1013 and 19% with faster opamp (tl072, lt1056).
I doesn't look like a so great improvement IMHO.
Of course the overshot as % is much worse for low output voltages
On the other side it is less when instead of short-circuit (output voltage to 0V) is just CV/CC switch.

A great improvement comes instead from reducing the opamp saturation.
Just a schottky diode (BAT46) connecting the inverting input to the positive ground reduces LT1013 overshot to about 20%.
More complex solutions are beyond the purpose of keeping the circuit as simple as possible.

In any case I'll be glad to know about any other faster opamp with the following features:
   [...]

To get good DIP parts is difficult nowadays... Look at MC34072, perhaps it will fit your needs...

I have it. But it is only just a bit faster.

The most funny thing is the hint to the "odd pinout" of LT1013... I just checked... most dual opamp I got here are the same:
AD8066, LM358, TL0*2, MC3**02, microchip ones, NE5532... and so on... You must be kidding 

Pinout of LT1013 is normal in DIP-8 package, but in SO-8 it is very unusual.

I apologize. I just forgot to check the smd case. 

BTW I just noticed you suggested the usual bad ebay kit.
That is one of those ugly circuit design I was referring to.

Design is quite old and it was not made to be beautiful... Voltage error amplifier connected after current error amplifier do not saturate, like in your simplified design. This improves transient response in CC->CV transition. TL081 are also a bit faster, then LT1013.

I'll try to simulate it to check how much better (in any) it is. 

Thanks for your suggestions.

[not1xor1]

OK
I realized that there was a sensible overshot in switching off from CC back to CV.

So I ran multiple simulations comparing the classical CV/CC switch by diode topology with that where the VC reference is decreased as needed.
I also compared LT1013 with LT1056 that with its 14V/µs of slew rate and 5.5MHz GBP is quite faster than LT1013.

The circuit topology where current control works on the voltage reference (like that ebay kit) does indeed completely avoid the overshot. Unfortunately it does that at the price of a much lower recovery time.
That is because both opamps are in the loop and must be overcompensated to avoid oscillations.

A countercheck of that is the fact that by lowering the speed of the CV feedback network of the CV/CC diode switch topology (by putting an RC in parallel with the DC feedback resistor) it is possible to completely avoid the overshot while still preserving a reasonable recovery time, even with the slow LT1013.

Another proof is that even a zener diode in the CV amp local feedback, while preventing saturation, does nothing on that overshot.

Of course if the same is true in the real world is a completely different matter  .

If anybody is interested I can clean the circuits to make them more readable and attach screenshots and or the asc files.
It would be better to check... just in case I made some coarse mistake... 

[wasyoungonce]

[quote author=not1xor1 link=topic=152870.msg19928

If anybody is interested I can clean the circuits to make them more readable and attach screenshots and or the asc files.
It would be better to check... just in case I made some coarse mistake... 
[/quote]

Actually that would be great and helpful.  Thanks for your efforts for others
Brendan



Sent from my iPhone using Tapatalk

[David Hess]

A countercheck of that is the fact that by lowering the speed of the CV feedback network of the CV/CC diode switch topology (by putting an RC in parallel with the DC feedback resistor) it is possible to completely avoid the overshot while still preserving a reasonable recovery time, even with the slow LT1013.

This is why it is seldom advantageous to use a faster operational amplifier in a power supply; extra frequency compensation is needed anyway to maintain stability slowing it down.  The one reason I have done so in the past is for lower noise.

Another proof is that even a zener diode in the CV amp local feedback, while preventing saturation, does nothing on that overshot.

Saturation should be handled separately.  Ideally the transconductance node in the operational amplifier is clamped but only some amplifiers which support external compensation make this point available which is a good argument for using old LM301A or LM308 type operational amplifiers in high performance regulators.  The 723 voltage regulator can be used this way also but its transconductance stage is pretty low performance although better than most applications require.

If enough output capacitance is used, then recovery from saturation is not a problem and most power supplies take this route.  Saturation and overload recovery become a problem in a design with a minimum of output capacitance like that shown below where the current control loop is clamped like I described.

[not1xor1]

A countercheck of that is the fact that by lowering the speed of the CV feedback network of the CV/CC diode switch topology (by putting an RC in parallel with the DC feedback resistor) it is possible to completely avoid the overshot while still preserving a reasonable recovery time, even with the slow LT1013.

This is why it is seldom advantageous to use a faster operational amplifier in a power supply; extra frequency compensation is needed anyway to maintain stability slowing it down.  The one reason I have done so in the past is for lower noise.

Another proof is that even a zener diode in the CV amp local feedback, while preventing saturation, does nothing on that overshot.

Saturation should be handled separately.  Ideally the transconductance node in the operational amplifier is clamped but only some amplifiers which support external compensation make this point available which is a good argument for using old LM301A or LM308 type operational amplifiers in high performance regulators.  The 723 voltage regulator can be used this way also but its transconductance stage is pretty low performance although better than most applications require.

If enough output capacitance is used, then recovery from saturation is not a problem and most power supplies take this route.  Saturation and overload recovery become a problem in a design with a minimum of output capacitance like that shown below where the current control loop is clamped like I described.

In the simulations, the highest output voltage spikes appear when switching from minimum load (i.e. mA) to short circuit and get much lower when the switching is between an highest load and short.

In the real world they would be even lower due to the parasitic inductance of the PSU cables.
Besides that, we are in the order of the tenths of nJ (nano jouls on CV-CC switch) or µJ (micro jouls - CC-CV switch) of energy.
Even recovering from 100A to 1A yelds around 20µJ of extra energy. Nothing to be scared by. 
OK... I've just realized the screenshots were relative to a 100mA/100A pulse rather than a 1/100A pulse which instead produces about 200µJ of extra energy.

Here is the circuit I used for the tests:



Here are the full simulation traces:



Here is the CV-CC mode switch zoomed:



Here is the CC-CV mode switch zoomed:



This screenshot shows the steady state energy (same time interval of the above one).
The extra energy is given by the difference of the two.

[jah118]

10 amps at 30 volts is beyond what a single power transistor can handle reliability without fold-back current limiting; 20 volts at 1.5 amps would be more realistic so I suggest building a lower power supply to gain more experience first.

The design as shown is certainly workable although the current limit is pretty basic.

hmm. i think i then will have to run 3 or 4 of them(2n3055) in Parallel,(and i have som massiv coolers from and old radio)