Question about blackdog power supply

[ElectronSurf]

This forum is hopeless...

[strawberry]

only when voltage is constant

[ElectronSurf]

only when voltage is constant

It is constant! it's a voltage regulator with very low load on it...

[ElectronSurf]

150 view and no replies? 

[ledtester]

150 view and no replies? 

blackdog has recently logged into the forum so how about sending him a PM informing him about your question and this thread?

[Zero999]

This forum is hopeless...

It doesn't help when the original poster deletes the first post.

I noticed this thread and was going to provide an answer, but was busy at the time and thought I'd leave it, until I had time to spare.

[Simon]

This forum is hopeless...

Let me know when you have had enough and I will gladly ban your account! People tend to feel rather insulted when people behave like that. No one owes you anything.

[EEVblog]

This forum is hopeless...

It doesn't help when the original poster deletes the first post.
I noticed this thread and was going to provide an answer, but was busy at the time and thought I'd leave it, until I had time to spare.

I can't even read it now, no idea what the question even was.

[Zero999]

This forum is hopeless...

It doesn't help when the original poster deletes the first post.
I noticed this thread and was going to provide an answer, but was busy at the time and thought I'd leave it, until I had time to spare.

I can't even read it now, no idea what the question even was.

Exactly, I don't understand why he deleted it. It might have had 150 views, but if most of those were guests and beginners, which is hardly surprising, given the section it's posted in, it's unlikely any of them would be able to help. He only gave it a day before giving up. People here are busy and have jobs to do. If you don't get any replies, add more information and bear in might there's no guarantee of a helpful response, as you've not paid anything.

If I remember rightly he was asking why a transistor was driven with a constant current source, rather than a resistor, which he thought was odd, given the voltage is regulated. I can't be bothered to look at the schematic. I think it might be to do with the fact the base voltage varied significantly and with a resistor the base current would be lower, at higher output voltages, which is when more base drive is necessary. Anyway, I'm not going to waste any more time on it.

[MK14]

What a pity, we can't hack, reconstruct the original opening post.  Let's try anyway:



Now using my memory banks (so take it very approximately, indeed):

Given that the (LM)TL317 is a voltage regulator, why is the LM334 constant current device (source) being used, instead of just a resistor, since the LM317 (LT) is already a fixed voltage device (part/output/regulator) ?

(The bits in brackets, show lack of certainty on my part, as to the exact wording, they originally used)

[ledtester]

What a pity, we can't hack, reconstruct the original opening post.  Let's try anyway:

For some reason your img tag is not working.

You can find the schematic in the first post of this thread:

https://www.eevblog.com/forum/beginners/how-does-blackdog_s-psu-work/

and that thread has an extensive discussion of the PSU as well as comments from Bram himself.

Given that the (LM)TL317 is a voltage regulator, why is the LM334 constant current device (source) being used, instead of just a resistor, since the LM317 (LT) is already a fixed voltage device (part/output/regulator) ?

That was basically the question as I understood it.

[MK14]

For some reason your img tag is not working.

Maybe it was cached on my computer or something.  Here is what I see:

[blackdog]

Hi,

This week I had a private conversation with ElectronSurf and I gave him some tips on his version of the schematic I showed on this forum....

Questions about this design
I get 3 to 4 questions a year about this linear Power Supply.
The basic circuit is a development from the end of 1950 by Harrison that later went over to HP.
Yes, that design is old and still very good, especially with modern components.

Modern opamp
There are now opamps for sale at reasonable prices that have around 90 degrees Pase Margin like the ADA4625-1.
This one also has a low offset voltage, low bias current, low noise and 48v/uSec.
And yes, you will have to tune the frequency compensation around the opamps a bit for that for best results for the ADA4625-1.

Powerstage
As ElectronSurf has also done and others also seem to do regularly, is to adjust the Power Stage configuration, that is exactly the thing you should not do.

The "Sziklay Pair" is specially chosen for this Power Supply.
You can use the following transistors for this purpose:
2SA1943, Fast! low thermal resistance
MJ15025 average speed, low thermal resistance
MJ15004 moderate speed, low thermal resistance
NJL1302D Fast! extremely low thermal resistance, hFE linear and additional diode that can be used as a sensor to drive a fan.
Buy good, real transistors, no China junk.

Reference stage
I have put a lot of effort into making a Linear Power Supply that is very stable, the reference section is built so that there is no hum or noise to be found.
I also applied filtering on the 230V(120V) side so that interference signals cannot reach the reference section(LT1021-5V).
The supply voltage for the LT1021 is chosen to give it at least the voltage at which it works well and low enough to keep the dissipation in the LT1021 as low as possible.
This helps in limiting the warm-up drift.

And what do you think most builders do?
Kill the reference circuit.... 

Construction engineering and measurements
Really every part of the circuit I showed was measured extensively, since many measurements are about hum and noise I use my Audio Precision Portable One Plus measurement set for that.
I also use a measurement computer with a good Audio sound card 24Bit, 192KHz to do broadband measurements.
Most DC measurements were done with the TEK DMM4050 and the KeySight 34461A and some with my 3458A DMM.
And several ways to test stability with Dynamic Loads and a insertion transformer in de feedback loop.

Build the power supply as if it were an HF transmitter in terms of wiring techniques.
Use Short twisted cables etc, etc.

Transformer section
In a few versions of the schematic I have shown how to keep the dissipation as low as possible.
I myself eventually went for two times a toroidal transformer (2x toroidal 2x9V = 4x 9V) of which I put the windings in series and switch them with relays controlled by comparators or intelligently via an Arduino.
So that you can measure on the output and on the buffer capacitor so that you measure the difference over the power section and on the basis of that control the relays with some hysteresis.

Good lineair power supply's are complex if you weant high performance.
You need at least a good scope and a good function generator as an example a function generator that generates almost perfect pulses is the Siglent SDG 1032x.
Use coax connections if you are going to do dynamic tests to go to your scope.
Let the coax cable not parallel to the output wiring, this is due to coupling between these wires, preferably at right angles.

Often I do pulstests to 15-Ampere at a 5-ampere power supply, this to see if the peak current limit and the normally set current loop reacts quickly.
So parallel wiring with these large pulse currents thus distorts your measurement.

Nice, enough info for now, can the rebuilders get back to the work/measurement bench! 

Kind regards,
Bram

[ebastler]

This week I had a private conversation with ElectronSurf

I hope your conversation also touched upon the intricacies of forum etiquette...

Seriously though, thank you for the background on the PSU design! For the avoidance of doubt -- is the schematic which MK14 has copied above the one you refer to?

[blackdog]

Hi ebastler,

Contact with ElectronSurf was normal, no complaints.
I did not see this topic until after I had contact with ElectronSurf.

This is one of the last schematics of he NA-01 design in high resolution.
https://www.bramcam.nl/NA/NA-01-PSU/NA-PSU-45.GIF

Kind regards,
Bram

[sorin]

Thank you very much blackdog for your work, but the original question is still unanswered .
Why did you use a constant current instead of constant voltage + resistor?
What will happen if we use just constant voltage + resistor?
I calculated that in your circuit the max voltage swing across LM334 will be less than 1V, so if we use 8V + resistor we will have a Delta I of max 15% (0-5A Load).
Why HP never (at least in my knowledge) used CC to drive the pass Transistor?
In my modest opinion is better to have to have a resistor because it will increase the response time , in case of a short circuit or current spike.

[blackdog]

Hi sorin,

Are you sure about this comment? => but the original question is still unanswered 

I answered the question in the email conversation with ElectronSurf
But of course I want to make you and the other readers happy too! 

Current Source
The reason for the power source is to remove the last bit of hum/noise from the design.
It was one of the last "tuning" steps I did to get the hum/noise a lot below the 5uV RMS at 20KHz bandwidth.
Of course the opamp noise also plays a part in this and of course the wiring technique you use, which becomes very dominant if you are not careful.

Wiring technique
Twist the wires to and from the transformer, use a toroidal transformer, these almost always give the best results, exceptions aside.

Keep the transformer as far away from the other electronics as possible.

Twist the wires to and from the bridge cell, twist the wires to and from the buffer capacitor.

As best you can, make sure the wires around the Power Stage are as short as possible, wideband power transistors are used,
and it works best when the wiring inductances are as low as possible.
Some of the tips are a bit contradictory, it is also not easy to get the best out of test equipment. 

Current Source-2
Back to the current source, I chose this one around 6mA, which is 10x higher than minimum necessary for 5-Ampere output current.
I had applied a small selection to the indicated power transistors and arrived at a hFE > 30000.
But it in never good to assume the maximum of my selected transistors, so even with an hFE of 10000 of the Sziklay Pair, there is 12x more current available for achieving 5-Ampere output current.
I don't see the speed of the current source at 6mA as a problem with the capacities present in the node to which the current source is attached.

The datasheet of the LM334 shows that at 1mA the LM334 only needs about 25uSec to stay within 0.5% of the set value, at the large current I chose this gets a little better.

And as you yourself point out, the voltage variation is small, so slew rate problems will not occur here.

That HP never chose a current source to drive the Power Stage I can not say for sure, but I have studied many of their schematics.
Just because HP didn't do it doesn't mean that what I chose wouldn't be good.
I have therefore chosen to keep the noise level at this point as low as possible by using the LM334.

And yes, they are only small differences, but what do I care about the 0.5$ of the LM334. 
This schematic was/is a test how well I could make it, far below the hum/noise level of commercial products.
And of course it is no problem for the operation to take a resistor instead of the LM334.

Some background
I design for myself, no one is standing here breathing down my neck whether it is ready yet or whether the parts chosen are too expensive.
I want quality for me and I want to learn to do it as well as I can.
On this and especially on the Dutch forum CircuitsOnline I show the path I take to achieve a good result, but also the mistakes and the assumptions that were not correct.

Also I don't see the schematic as Heathkit kit, this is because I know almost everyone wants to do it their own way and often ignore the decisions I chose after many hours of precision measurements.
Bram can I also use LM741, Bram can I use 2x a 2N3055, I hope you guys understand what I mean.
So I usually give a handful of tips on what to look out for and otherwise it's your journey to achieving good results.

I too have had to learn what it takes to make a good linear Power Supply!

One of my first ones I blew up at least five times due to a short circuit.
Another schematic I got from a friend of my father's when I was about 16 years old was more of an oscillator than a power supply.
Little did I know what phase and gain margin was.

Only at that time there was hardly anyone who could help me, in this day and age there are several Fora where very many people are available who have many years of knowledge available on many subjects.
We as a human species are very rich in available knowledge!
This is totally different from 1970 when I really started with electronics and by that I mean that I moved away from building radios from booklets and started experimenting with Power Supply's and transmitters.

Above all, keep building and experimenting, be very careful with assumptions, they are so easy to make and end up costing you a lot of time.

Kind regards,
Bram