Linear power supply project

[Therm Mr.]

Since I just signed up I thought I would say hello. So hi, let me introduce my self. I am a semi retired A/C Contractor. When I was younger, BC (Before Children) I was an electronics hobbyist. Then life got in the way for about forty years and here I am.

So now just an old man with too much time on my hands. Thus I decided to get back onto the “bicycle” so to speak, and take on some electronics projects. Currently I am planning to build a professional quality linear power supply. I know I could buy one cheaper but I want the challenge.

I found a hp/Agilent 3610a schematic and parts list (I think on this site). It’s a 8v@3a and 15v@2a dual supply. It also has CC and CV capability. So, lets get to the questions.

1) It uses a darling pair of 2N6036 power transistors. I believe they are now obsolete. Any suggestion of what to use in their place? Also any suggestion for a fan cooled heat sink for them?

2) Also there is a U3 listed I can not reference and I have no specs for. It is a LM336BZ any idea on what it is.

3). The transformer shows 8 wires feeding three different bridge rectifiers how would I sort that all out? One controls high low range to pass trans, one to display power and one reference and bias supply.  How can I spec the correct transformer.

4) There also is no schematic for the LED display driver board. How can I find an alternative for that?

5) Do you know of a more modern alternative that you think may be a better power supply project? A schematic, parts list and maybe PCB design would be helpful?

I will attach the schematic file. Thanks in advance for any help or advice offered. I know I may let out some smoke and $$ but an education costs money sometimes.

Therm

[Kleinstein]

For the start I would consider only a signle channel, so just 1 voltage. The others are kind of copies of the same / similar (e.g. reuse the same PCB tye).

The power transistors in the plan look like MOSFETs. AFAIK HP used IRFP150 - todayone would use the slight higher voltage IRFP250 as an alternative, though the new ones are not as suitable as the old ones.
The circuit is made in a way that it could also use Darlington transistors instead (relatively little changes needed).
One could use integrated darlingtons or separate transistors. For the start it helps to keep the voltage / power small. Scaling is later not that difficult.

For the transformer the control part and display only needs relatively little current. For a better estimate it needs a final circuit, especially the display part. The is a chance to use the same supply for the control and display and this way simplify the transformer selection.  For of the shelf transformers one may use 2 separate transformers for the main power and the control part.
For the display part there are several options: old style ICL7106 panel meters, modern cheap panel meter modules from China (usually ADC+µC based) and your own ADC+µC version.
Some need a separate transformers. Some could use the control supply.

More modern PSUs use digital control of the set point. This needs a slightly different circuit at the ref. side / voltage adjustment. The old way with an adjustable divider has its draw backs anyway. There are also plans around for some of the cheap Chinese supplies - largely similar principle, but could be simpler parts. It is definitely worth also looking at 1 or 2 of those.
There are 2 design here in the formum base around LM324 and using 1 supply only.It is a little less flexible, but a real possibilty.

[pqass]

1) TIP36CTIP147 (PNP darlington) is a readily available substitute in a TO-247 package instead of the original TO-3. It's only the E3612A that uses the MOSFETs so the schematic is a bit misleading with additional, unpopulated, or substituted components based on model. See service manual page A-6 to A9.

EDIT: according to the service manual the series pass transistors should be 2N6056 not "36"; typo?  This is an NPN darlington part. As such use the TIP35CTIP142 instead.

2) Instead of the LM336BZ5.0 you can use a common TL431 with two 10K resistors to make a 5V zener: anode-10K-adjust pin-10K-cathode.  I did just this in my E3611A.

3) You could use two transformers; one for the main power source and another for the floating Ref+Bias supply. Then use an isolated DC-DC for the display supply off of the Ref+Bias supply. The Ref+Bias supply transformer can be low power (< 5VA) as it only supports the op amps and display.

4) The display is just two separate ICL7107 ADC + LED driver boards.  You can buy generic 0-200mV 3-1/2 digit voltmeters and add appropriate divider resistors.  The important bit to remember is that most of them require their power source to be isolated; that is, the negative sense input (to the voltmeter) should not be connected to the negative power supply lead (to the voltmeter).

5) You do not necessarily need to implement exactly what you see.  You can strip away some of the circuit and simplify to better understand how it works (and this one too).  ie. Rewrite the schematic to show its core parts by removing/substituting the components for a specific supply like the E3610A, removing the the CV/CC indicator parts (comparators), display divider resistors and supply parts. You can add them back in later.  Important hint: notice the (+S) symbol connects the output +(red) post with the GND of the Ref+Bias supply.

[Kleinstein]

The TIP35 / TIP36 are normal transistors, not darlingtons. They are still a reasonable choice, but may need an extra transistor to make it a darlington circuit.
Standard power darlingtons are TIP140/145.

Depending on the skill level it makes sense to not just build a 1:1 copy, but only take the principle configuration and configure / trim the circuit yourself. Here it can really help to simulate the circuit to get a good understanding.

[pqass]

The TIP35 / TIP36 are normal transistors, not darlingtons. They are still a reasonable choice, but may need an extra transistor to make it a darlington circuit.
Standard power darlingtons are TIP140/145.

Yes, of course.  I don't know why I wrote that.
Brain probably thought TIP35/36 were darlingtons to TIP31/32.
Corrections were made to my post.
Thank you.

[Therm Mr.]

Thank you both for the replies. Very informative and helpful.

pqass don't be too hard on yourself mistakes just prove your human. The Component and tech info you provided will be very helpful.

Kleinstein your advice will too be considered and appreciated. When I was doing this 30-40 years ago the World Wide Web was a baby. Now I am getting help from people all over the world. This is great.

I am trying to stay as true to this design as possible due to my engineering skills being limited. I understand what things do but the why they do it is is where I am most lacking. My craftsmanship is is good and I learn pretty fast. I would consider my skill level Not beginner, not expert but intermediate. I am definitely open to changing the design as long as my end product is a quality device. I have the time to take my time and learn from the experience.

A few more questions if you don't mind.

1) I can have a good design and good craftsmanship but if I have poor quality components I will not have a good device. How do I assure quality components? Which manufacturers do I look to?

2) The attached section of the schematic shows all the components that are deleted in the 3610a. It seems as though the section needs a complete redesign with the TIP147. Is there a reference somewhere I can use to help with that?

3) Also in this section Q1,2 will be the Darlingtons. It shows two components next to them D6,7 that look identical to Q1,2. These are not listed in the parts list and Have me a bit confused. What are they? Edit: I have uploaded a PDF of the schematic in the original post. It is much more clear and D6,7 are actually Q6,7. Still not listed parts, so still confused.

I am told stupid questions are the ones you don't ask.

Thanks again

Therm

[andrewtaylor]

To Therm Mr. some remarks which might help:

2) Also there is a U3 listed I can not reference and I have no specs for. It is a LM336BZ any idea on what it is.

pls .see datasheet (TI.com etc.). It'S a shunt refetce (similar to a zener diode, but better specs).


3). The transformer shows 8 wires feeding three different bridge rectifiers how would I sort that all out? One controls high low range to pass trans, one to display power and one reference and bias supply.  How can I spec the correct transformer.

The HP is already a quite clever design with some good engeneering practices used. For your projecte I'd recommend a 2x12V Transfomrer with 4A capabilty for the main power (based on if you go for the specs HP has in.
It is close to the ratings you need, but of course you have to keep in mind: HP has custom wound their transfomrer, which makes sense if you bulid 1000+ power supplies. for a single project, it's econimic to go for the parts RS etc. sell from shelf.

The other windings: well, you have the +/- reference - I'd take a 2x15V 200mA Transformer for my power supply projects for this part.

And you need power for the DVM/DAM displayd - I'd recommend you buy the digitla modules from shelf (available as 200mV and 20V Fullscale models).
And after the buy, get some transformer that powers these seprately - remeber to keep the auxiliary power separted, no common ground.

4) There also is no schematic for the LED display driver board. How can I find an alternative for that?

pls. take a set of 2 DVM moduels as above mentioned.


5) Do you know of a more modern alternative that you think may be a better power supply project? A schematic, parts list and maybe PCB design would be helpful?

Well, honestly spoken: The performance you will get out of the HP36 series is hard to beat (low ripple, low dissipation, low PARD) and still quite impressive.

I would not call it a "beginners project" to bult a clone (I buld lab bench supplies since more than 3 decades on my own) -- but you can learn a lot.
And, you can download the complete HP36 sevice manual for free -- it has a lot description on how to ansd why they build it the way HP build it.

If you like a more indrotion, I recommend the classic:
DC Power Supply Handbook (Agilent Technologies   Application Note 90B)

It's stone old, but for the basic it shows and teaches all the does and don't on how to build an use a power supply .

Have fun!

PS: As you might have noticed, I did not give a remark like you oftne see in forum as "building makes no sense, buy is dead cheaper" etc.

Yes, you get these power supplies from Philips, Delta elektonika, Kepco, etc for 50 bucks on the used market/flea market.
But I think that's not at all you intention - build one is not cheap, but can be very intuitive .-)

[Kleinstein]

The part with Q4 and Q5 and R44/R45 is there to get better balance in the current with MOSFETs. With BJTs one can usually skip that part if the emittter resistors are large enough.
Q4/Q5 may drawn the wrong way around (E and C swapped). With 4 transistors one should have 4 emitter resistors, not just 2 as in the plan.
What looks like D6 D7 should be Q6 Q7, just a poor scan / copy of the plan.

A lab supply looks simple at first. If done right and with all the details it can be quite envolved and a good learning experience. It is a basic feedback amplifier or regulator problem.  A point that makes it a bit tricky is that one wants stablitly / fast regulation even with a changing load (system). In control theory terms this is calling for a robust regulator.

It is not only about getting quality parts, but also about the right regulator tuning (frequency compensation in amplifier terms). There are mainly 2 ways such supplies can go wrong:
one is the thermal design with overheating / overloading parts so they will eventually fail. The other is borderline compensation which results in large overshoot or turn on spikes or oscillation with certain loads (especially capacitive).

[Solder_Junkie]

Back in the 1970s, I built the ETI 131 power supply, it is so good that I still use it today. I modified it to use a digital Volts/Amps display as the old analogue meter had become prone to sticking.

With a print board design package you could update the board layouts to have a set manufactured in China for very low cost, JLCPCB charge approx $2 + postage for a set of 5 boards!

The original article is on this page in two zip files:
https://www.thebackshed.com/forum/ViewTopic.php?TID=4422

SJ

[xavier60]

Back in the 1970s, I built the ETI 131 power supply, it is so good that I still use it today. I modified it to use a digital Volts/Amps display as the old analogue meter had become prone to sticking.

With a print board design package you could update the board layouts to have a set manufactured in China for very low cost, JLCPCB charge approx $2 + postage for a set of 5 boards!

The original article is on this page in two zip files:
https://www.thebackshed.com/forum/ViewTopic.php?TID=4422

SJ

That ETI design is similar to what I used for my last bench PSU build, https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664
Slightly simpler than Harrison topology with good performance.

[Therm Mr.]

Thank you andrewtaylor. Good info and references, I will look at all of them. Your post pretty well describes my motivations and intents.

Kleinstein, I must have been editing while you were posting. I added a much clearer PDF of the schematic from the manual to the original post. It shows Q6,7 just like you posted. I will pay close attention to those things you mention.

Thanks for the info Solder_Junkie I have a long way to go in design and testing before I am ready for PCB.

Therm

[Solder_Junkie]

Therm, back in the 1970s I used a Dalo etch resist pen and ferric chloride to hand draw and home etch PCBs. I have been using heat transfer paper and ferric chloride until recently.

Only in the past few months have I started to have boards made by JLCPCB, the quality is incredible and they are really low cost. My last batch was delivered in under 2 weeks from uploading on a Saturday, the cost was the equivalent of $4.93 for 5 boards, inc postage and tax. Well worth looking into.

SJ

[mariush]

If you want some other schematics, a few of Circuit Specialist power supply schematics and a TTi psu schematic / service manual type of pdf and the hp e3620a schematic

[pqass]

A few more questions if you don't mind.

1) I can have a good design and good craftsmanship but if I have poor quality components I will not have a good device. How do I assure quality components? Which manufacturers do I look to?

2) The attached section of the schematic shows all the components that are deleted in the 3610a. It seems as though the section needs a complete redesign with the TIP147. Is there a reference somewhere I can use to help with that?

3) Also in this section Q1,2 will be the Darlingtons. It shows two components next to them D6,7 that look identical to Q1,2. These are not listed in the parts list and Have me a bit confused. What are they? Edit: I have uploaded a PDF of the schematic in the original post. It is much more clear and D6,7 are actually Q6,7. Still not listed parts, so still confused.

See attached refactor of the Series Pass and Driver transistors for an E3610A implementation.

The non-applicable components (those for the E3612A) were removed or replaced with a short if they were in-series.  Also, only one series pass transistor + emitter resistor is shown since it's the minimum required.  You can add as many as you need depending on the output current desired; needing more current means hotter transistor so double,quadruple-up to keep any one from exceeding its spec.

1) Quality of components:
If obtained from a reliable source any manufacturer equivalent part would do.  The design isn't demanding; no high-speed switching requiring low ESR caps, Q1, Q2, CR4,5,8 can have many alternatives, etc.  However, there is some nuance in the design (notice the the ferrite bead L1).

Mini rant: HP seems to use odd valued components (non E12 series) and precision (< 5%) in the most mundane of places like pullup resistors; no BOM consolidation at all.  No wonder even their simplest products are so expensive.

2) Redesign schematic for darlingtons:
The HP schematic is generic for three implementations (E3610,11,12A) but as is, it's hard to see the core within the clutter.   Just re-writing it without the clutter helps (see attached).

3) Diodes across MOSFETs (I think you're referring to):
These are the "body" diodes inside every MOSFET.

Sometimes the schematics are not accurate (C2 on page A6 vs A9) or intentionally omit.  In the context, Q6,7 are the doubling-up of Q1,3 so should be the same parts.

[coromonadalix]

you have many projects on the web   for linear supplies,  CV, CC    etc ...  with simple pass transistors and op amps   

why choosing the complicated route ?

You ask for under 50 watts supplies  8v 3a,  15v 2a    ....       even cheap ones based on lm723  are good ?? 

and for the displays, you have  low cost but good  3 1/2 digit  led or lcd panels, 4 1/2  too  if you want to get fancy


even kits like this :
https://www.ebay.com/itm/155160183261?
https://www.ebay.com/itm/355584454179?


i understand  you want to get back on a "bicycle"   but start at a proper level,    mosfets based psu's are somewhat trickier to build.  and avoid self oscillations,  pcb's designs too     etc ...

[Doctorandus_P]

It does not have to be expensive. I don't know what the original 3610A cost but I just (re) viewed Dave's old teardown

https://www.eevblog.com/2011/04/22/eevblog-166-hp-agilent-e3610a-lab-power-supply/

and there is nothing special in it, and the posted schematic also confirms this. So I wonder why it's considered "so good".

Also, you may have a suitable transformer or other parts laying around, a box to put it all in maybe?
For myself, I would put a microcontroller in it. (With ADC and DAC). I much rather have encoders instead of the 10 turn pots, and it can also handle the rest of the front panel (Display / buttons). On top of that, an uC makes it easy to add remote control or to link multiple power supplies.

[Kleinstein]

The HP supplies were good at the time when they came out. it looks like a good thermal design (no overheating) and the trim of the regulation is likely good (so not much overshoot / ringing). Many of the early supplies from cheaper sources (and new ones from China) had there troubles with overheating / overloading parts to cause a high failure rate.  Also some have a tendency to ring or even oscillate with a difficult loads. In fairness one must note that the tricky very low ESR electrolytic capacitors are relatively new.  It is not that the HP supply is especially good. It is more than many others were so poor.

The LM723 is not making things much easier and cheaper. A version with LM358 and TL431 can get you better current regulation with only slightly more reference noise. The LM723 still needs the right compensation and needs extra complications to work down to 0 V.

[iMo]

This is a typical schematics of a cheapo 18V/3A PSU (or 15V/3A, etc) you may find under dozens of various names, with LED or LCD 710x based meters (Voltage/Current)..
I got one as a gift and it works fine.. I wanted to mod it heavily, but finally I resigned on that effort and I only added a front panel switch and a resistor with the current shunt R22 split into 3A and 0.2A ranges.

PS: there are 2 additional simple 7809 floating (isolated) power sources inside on the pcb (the transformer has got 2 additional windings + diode bridges + caps + 7809s), powering the two (current and voltage) front panel 7106 LCD meters. They are sensing the current and voltage at the shunt R15.

[pqass]

It does not have to be expensive. I don't know what the original 3610A cost but I just (re) viewed Dave's old teardown

and there is nothing special in it, and the posted schematic also confirms this. So I wonder why it's considered "so good".

Also, you may have a suitable transformer or other parts laying around, a box to put it all in maybe?
For myself, I would put a microcontroller in it. (With ADC and DAC). I much rather have encoders instead of the 10 turn pots, and it can also handle the rest of the front panel (Display / buttons). On top of that, an uC makes it easy to add remote control or to link multiple power supplies.

HP Catalog for 1993 has it for $300.  That's $661 in today's money. 

I like the aesthetics and UI simplicity. Although, I wish it had an output on/off.

Sometimes it's good to go back to basics - to build a minimalist implementation in order to fully understand the core. 
See here and attached.  You can always add on to it later.

Furthermore,
the attached implementation (below) of the circuit (link above) does work.  However, it does have some issues that I'm working through.  This is when learning happens.   
That is: Understand theory > simulate it > prototype it > test/characterize > revise until all goals are met > product-ize it.

[dietert1]

We have a similar E3630A and it is pretty stable. If i manage to adjust it for 10 V and come back some hours later it will be within 1 mV. It's well built, except the meter push buttons sometimes jump out. Maybe i should use glue to fix them once and forever.

Regards, Dieter

[Kleinstein]

The HY1803 uses the same basic principle, but it still misses out on details of the compensation that can make the difference between good and prone to oscillation (or awfully slow). The resistors in series to the feedback cap at the OP-amp are the step of going from a crude I-regulator to a better PI regulator. The additional RC element at the voltage divider makes it a full PID.

There is quite some though / calculation (or try and error) in these small details. It is a shame those cheap supplies failed in this detail that adds very little costs, but can make quite some difference in the performance.  Especially with modern simulation tools available there is no good excuse for this in a series product.

Adding at least a simple output on / off is relatively easy and you are free to do this for a DIY build.

[Therm Mr.]

This thread is giving me one heck of an education. Thank you all.

That power supply handbook is chock full of info.

Lots of variety in all those schematics posted also.

pqass is the pic I uploaded in this post the proper way to add another trans?

Do you guys have any preferred schematic drawing software? I did that one on one of the online programs. It was a bit clunky. I am fairly good with CAD programs. I use BricsCad because AutoCad only rents now you can't buy it. But Brics still let's you buy.

It will take me a bit time to absorb all the info in this thread. I likely wont be able to absorb it all. But I will get a lot of it. This place is great.

Therm

[pqass]

pqass is the pic I uploaded in this post the proper way to add another trans?

Do you guys have any preferred schematic drawing software? I did that one on one of the online programs. It was a bit clunky. I am fairly good with CAD programs. I use BricsCad because AutoCad only rents now you can't buy it. But Brics still let's you buy.

See attached for revised schematic.  Changes/corrections are in red.

As I'd said, HP loves odd values but you can substitute:
1K for 1.2K, 4.7K for 5K, 1N400x (x=3-7) for CR8, 1N4148 for CR4,5, and
don't bother with L1 for now (it's actually a tiny ferrite donut on the leg of Q2).

Yes that's how to add another transistor. 
The emitter resistor is very low value of about 1R or less.
It's used to equally spread the current load across all parallel transistors due to manufacturing differences. 

I like kicad. It's actively being developed and provides much more than what I need (schematic capture and layout of simple circuits).

[andrewtaylor]

That power supply handbook is chock full of info.

completely agreed .-)
the detailed description in a HP E36xx service manual gives you even more "brain food"

Lots of variety in all those schematics posted also.

Here is where the pain in brain begins: Question is where to start? Which one to choose/take?
If I would be in your situation, my recommendation: Start with any one, and finish it to the very end. So stay fully focussed. Once your through this very time consuming 1st build, the next ones will be easier.

Next ones? Yes, one rarely stays with only a single power supply when you are into this hobby .-)



Do you guys have any preferred schematic drawing software?

I like KiCad. Fairly easy to learn, a lot of tutorials,
and good for PCB (BTW: v.7 has an autorouter extension, I guess v.8 will soon follow to get one).

It will take me a bit time to absorb all the info in this thread. I likely wont be able to absorb it all. But I will get a lot of it. This place is great.

Take your time, as previously Kleinstein has mentioned:  Build is simple, but build it RIGHT to work without oscillation etc. is the key that divides a design from poor to "excellent for use every day".
Go for the last.

my 2cts of opinion .-)

[Kleinstein]

For relatively simple schematics one can also use the a simulation tool like Tina or LTspice. For getting suitable values for the tuning (e.g. the small capacitors) the simulation tool can be a big help.