Fixing old 723 variable power supply. Should I replace the guts?

[coppercone2]

it is capable of slaying capacitance related problems

[coppercone2]

Another feature can be a output isolator switch so you can plug things in and adjust it before turning it on (possibly with a lowish value power resistor to drain the circuit it will connect to)

sometimes when the power supply is too basic it feels like its a ghetto ass hydraulic system where it starts spilling fluid on you when you disconnect it (without like a check valve or something). Its nice to have a switch for this

A really good supply IMO would come with a variable load, so you can power it up on a load to get it warmed up the correct amount and then transfer it over to the circuit you want to run (if its precision work on stable devices), so an hour after you want to plug something in a green light comes up (like a txco)

[David Hess]

Another feature can be a output isolator switch so you can plug things in and adjust it before turning it on (possibly with a lowish value power resistor to drain the circuit it will connect to)

The output switch on the Tektronix PS503A power supply shorts out the capacitor in an RC filter after the reference.  So when enabled, the output ramps up and when disabled, the output is instantly removed and pulled down with a few milliamps.

Some power supplies have a class-b or class-ab output stage so will actively discharge their output capacitance.  Others may include a low current pull-down.  The PS503A does the later, but with an added twist that the operational amplifier used for the error amplifier has an external offset applied which is always greater than its possible input offset voltage.  This means that the voltage control loop will actually control the output to zero volts, and a couple millivolts negative.  And no trimming of the operational amplifier is needed to allow it to regulate down to zero volts.  I have seen some old designs where this was done by deliberately pulling the offset null adjustment one way to induce an offset, but in some operational amplifiers this can ruin input offset voltage drift.

A really good supply IMO would come with a variable load, so you can power it up on a load to get it warmed up the correct amount and then transfer it over to the circuit you want to run (if its precision work on stable devices), so an hour after you want to plug something in a green light comes up (like a txco)

The voltage reference dominates any warm up errors, and should be settling within milliseconds.  After that the error amplifier is the next largest error source.  Unless you are expecting microvolt stability, I would not worry about warming up unless you have an ovenized reference.  Some old designs had an ovenized error amplifier to reduce input offset voltage drift.

[PCB.Wiz]

I have an old 723 based variable power supply that I built when I was about 13.  Finding myself in need of another bench supply, I dug it up and fixed the enclosure.  I may change the big input capacitors, but what about designing and building one of my super precision regulators, adapted to variable operation, to replace the insides, with noise, load, and line regulation under a microvolt?

With just an analog meter, and a standard pot, I presume you jest there.

The first upgrade I would do to any bench supply is to add decent V.I displays.
Aliexpress these days has 5 digits ones like this V.A.W
https://www.aliexpress.com/item/1005001830298410.html
or this one for A
https://www.aliexpress.com/item/1005005855666549.html

Claims
Measurement accuracy: 0.01%（+/-2）*2
Refresh rate: 3 times/second
Display mode:5 bit 0.36 inch LED

Maybe it uses NAU7802 or HX711 ADC or similar?

Adjustable current limit, and Adj V to 0V, are also useful on bench supplies.


 

[coromonadalix]

any 723 based psu's can be improved. mostly on the current limiting section

the "comp" capacitor  is supposed to prevent any oscillations from occurring

[RoGeorge]

The two 1100 microfarad 50 volt capacitors were trashed with 25 ohms of series resistance.  I replaced each one with a pair of NOS (New Old Stock) 2200 microfarad 35 volt capacitors in series which are "only" 30 years old.

If you still have those 1100uF, it would be interesting to re-form them by progressively applying up to 60...70Vdc for a few days, or at least 50V.  Then measure C and ESR again, to compare that with the values measured before forming, just out of curiosity (not as an attempt to keep using them).

I've fooled around last year with some random electrolytics fabricated during 70's - 90's, and most of them were able to regrow the oxide layer after keeping them for a few days at 1-1.2 Vnominal.  At some point, there were horrible spikes in the plotted current before reforming, then after only a few hours of oxide layer regrowing, the leakage dropped to 1uA or so, and not much noise:  https://www.eevblog.com/forum/projects/reforming-long-unused-electrolytic-capacitors/msg5680727/#msg5680727

I wonder if other brands of many decades old electrolytics can be reformed with similar success.

[PGPG]

I have an old 723 based variable power supply that I built when I was about 13.

Being about 13 I have started to do my power supply (in a plywood casing), but I had no access to any ICs so it was whole made by discrete elements.
I had no regular radiator but had an aluminum block 10x10x1cm so 2N3055 is mounted to it.
Knowing that it is not good in heat dissipation my concern was to allow for dissipate as small heat as possible (output current limit was regulated in a range 0...4A).
Because of this I wounded at 100W trafo coils giving me 6+6+6+6+6+1+1+1+1+1V and using bana plugs I can select any AC voltage from 1V to 35V used for next supply stages.
Few years later I have added separate small transformer to get negative voltage and replaced the regulating circuit with LM723.
At home I don't need anything more from very rarely used supply

[schmitt trigger]

It’s safe to say that a significant amount of electronic hobbyists built their own linear PSU with a 723 and a 2N3055, I certainly did.

If you rebuild the unit, in addition to the output enable switch mentioned earlier, a bright “OUTPUT ON” Led.
But most importantly, replace the potentiometer with a multiturn type, if you can find one at a reasonable price. Otherwise use two pots in a “COARSE/FINE” arrangement.
A further refinement, if you have a transformer with a dual winding secondary, to include a “HIGH/LOW” switch, to reduce the power dissipation while outputting low voltages.

[David Hess]

It’s safe to say that a significant amount of electronic hobbyists built their own linear PSU with a 723 and a 2N3055, I certainly did.

This one was a local kit as I described, with just a printed circuit board and the parts.

After a few 7805/7815/7915 fixed voltage power supplies using the same transformer type, and when I knew enough electronics to be dangerous, I designed and built a reference and operational amplifier power supply with adjustable current limiting and 10 turn poteniometers.  It worked, but was a disaster because of the needed frequency compensation to deal with the cascaded operational difference amplifier used for current sensing.  After that, I went back to boosted current 317/337 designs for a while.

[floobydust]

There's big jumps in what the single chip regs can do.
For all the three-terminal regs the regulation and drift is not that great. A few % maybe. I think designing for really tight regulation is not practical unless you go 4-wire.

The LM317 has no adjustable current-limiting and the 1-1.5A output limit can be a bit less than you need.
Went to the LM350K and it's too high though at 3-4.5A
Current-limit is not great on the 723 I think it's more intended for short-circuit protection.

I need to go below 1.2V for single-battery circuits.
I need 13.8V for charging batteries so the 7815 was out.

This rules out most of the simple circuits out there. I think it leads to using several op-amps?

The ubiquitous chinese kit 0-30V 2mA-3A (copy from 1978 Practical Electronics) does not perform well but is really in the zone of what is commonly needed, I think. If the PCB layout was fixed, ratings made conservative and a few other quirks fixed, I think it would be popular.
They also have a Hiland LM317 one (pic) with added current-limit (hey great idea) but it's kinda hilarious. I was going to draw the schematic as it sets a new world record for the number of 1N4007 diodes (12) used lol.
Both of these need that -ve rail which adds complexity.

[David Hess]

Good load regulation with 3-terminal regulators requires careful layout.  The top of the output divider should be attached as close as possible to the package, otherwise the error from the voltage drop between the package and top of the divider gets multiplied.  Of course the load regulation will never be as good as a regulator with remote sense, but there is no need to make it worse with poor layout.

With a current boosted design, a significant or majority of the output current may be diverted around the regulator, improving load regulation.

Remote sense can be added with a single operational amplifier, improving load regulation considerably.

I know a clever mind bending circuit for allowing a 3-terminal regulator to work down to zero volts without an additional reference, shown below.  The operational amplifier operating as a different amplifier samples the internal reference and uses that to generate a negative reference for the adjustment divider.  Of course it still requires a negative bias supply.  I could do the same thing for my 723 power supply, and have a transformer winding to generate the negative bias supply, however I question whether the effort is worth it because the same operational amplifier plus a reference would make a better regulator whether the 723 was included or not.

[David Hess]

the "comp" capacitor  is supposed to prevent any oscillations from occurring

I know from previous experience with this power supply that without any output capacitance, it will oscillate with some unbypassed loads.  I think I will add a small output capacitor to improve the stability.

One thing I do not like about these designs is that the output divider being within the feedback loop alters the frequency response, so the frequency compensation becomes a compromise.  The Tektronix PS503A works this way also, but it was deliberate to provide remote operation, so at least its design has an excuse.

I have an old 723 based variable power supply that I built when I was about 13.  Finding myself in need of another bench supply, I dug it up and fixed the enclosure.  I may change the big input capacitors, but what about designing and building one of my super precision regulators, adapted to variable operation, to replace the insides, with noise, load, and line regulation under a microvolt?

With just an analog meter, and a standard pot, I presume you jest there.

The first upgrade I would do to any bench supply is to add decent V.I displays.
Aliexpress these days has 5 digits ones like this V.A.W
https://www.aliexpress.com/item/1005001830298410.html
or this one for A
https://www.aliexpress.com/item/1005005855666549.html

Claims
Measurement accuracy: 0.01%（+/-2）*2
Refresh rate: 3 times/second
Display mode:5 bit 0.36 inch LED

Maybe it uses NAU7802 or HX711 ADC or similar?

Adjustable current limit, and Adj V to 0V, are also useful on bench supplies.

I do not disagree with what you suggest, but it would be outside the scope of this project because I would instead just design and build another power supply from scratch.  I will likely to that eventually, but not at the cost of only reusing an enclosure that I can replace in a couple hours.

One of the things I would like to investigate is improving common mode isolation and noise.  Modern power supplies seem to suffer even more from this problem, and many of my old designs did as well.

[David Hess]

I put it all back together, measured the ripple, and non-unexpected found the problem shown below, which I believe is snap recovery in the rectifiers.  So now I have another problem to fix.

I have found this problem in lots of power supplies, including my own, and it is why I have a keen interest in a design with better isolation.

[coppercone2]

that is why they put those ring of capacitors around the rectifier I think? Hp always has a few capacitors, not always four, sometimes 3 (muntzing?). I am almost certain once I thought I had to replace four but then there was only three.

Something like 1 to 10nF.

Sometimes they have a cap across the transformer (ceramic) that likes to make acoustic noise, sometimes they also have caps across the diodes. Their old RF things IIRC have 4 caps right around the rectifier, but on things like power supply I have seen that arrangement but the caps are like a inch away on the rectifier/capacitor PCB. I wondered why this was, because they could have put the capacitor right there, but instead it was farther away. I wonder if it is using the inductance of the trace or something. And I think it had a separate trace back to the transformer pin (like a star connection) rather then looping back across just the diode...? I mean it felt like they had to go out of their way to make it like that instead of just putting it across the diode. I need to try to remember which unit that is. Could it have been the HP current source 6177.. possibly the A version which got dispatched for being a nuisance  (the remains are in a 'canopic jar'. )

Maybe just start with 4x 1nF 500V SLCC caps around the rectifier and see if it goes away ?

Those tiny ceramic capacitor on the rectifier/capacitor board are the ones that sing, they always give me a problem, making me replace them with foils. But do they do their job, I am not sure, because the SLCC capacitor "bandwidth' is huge




also it could be related to shortwave RF
http://www.ll-leasing.com/pritch/shortwav.htm

[KE5FX]

Another effect worth considering is described on page 410 of AoE X Chapters, in which the transformer's leakage inductance causes a voltage spike to appear when the diodes turn off. 

That was a revelation when I came across it, because I've often used the Signal Transformer split-bobbin parts that they talk about in various homebrew projects.  I had seen the phenomenon happen (and been annoyed by it) several times without understanding what was going on.

[coppercone2]

I wish I had a time machine so I can go back 60 years and plug the same box into the outlet to see if its the outlet or if they just did not care  about the singing noise 

fyi related to this, putting random IEC filters does not really help eliminate the noise. The only thing that did it is replacing the caps with foil. I think the foil cap is just heavy and it dampens itself.

[coppercone2]

Is the solution for the leakage inductance to put a capacitor across the transformer?

So then you need a cap across the transformer and caps across the diodes to deal with both problems? Though I am thinking if you use 4 caps across a bridge, their basically across the supply (but with reduced capacitance because their series capacitors)... then you still might need a 5th capacitor if it needs a larger value then the one created by the caps across the diodes?

I see some creative solutions to transformer problems, this guy is thinking to use dampened capacitors
https://www.diyaudio.com/community/threads/rc-snubbers-for-diode-recovery-noise.190638/page-2

it seems that it would be fun to try to find a bad transformer and see if these solutions could have a measurable effect on a scope.. I only saw that kind of arrangement specified for EMI filters for 'military specification' DC/DC converter solutions (R+C dampener)... they looked more complicated then the ones you commonly see for inlet filters.

[KE5FX]

Exactly, they recommend a snubber cap in the 1 uF range directly across the secondary, with a few ohms in series to cut down on ringing.

[coppercone2]

The 6227B has a cap across the transformer (with no R) and then cap's across the diodes, I think that is where I saw this.

[PCB.Wiz]

I put it all back together, measured the ripple, and non-unexpected found the problem shown below, which I believe is snap recovery in the rectifiers.

That looks a bit long for recovery times?

What load was that, and can you include the pre-regulator ripple waveform ?

[floobydust]

There's 4 diodes (bridge rectifier) on the top of the PCB and then more on the bottom side like a center-tapped rectifier so I'm confused why two?

I use an RC snubber, or just a cap across a transformer's secondary to lower leakage inductance ringing (actually it just lowers the frequency), especially if slow recovery diodes are there causing the problem.

[Kleinstein]

The shown waveform has too long drops to be the diode recovery. Diode recovery times for 1N400x is at some 3 µs. A capacitor of some 1 µF at the transformer secondary could help a little against voltage spikes and as a side effect it would take over part of the magnetizing current. So the AC loading is more a beneficial and not an issue.

The ripple shown could be the voltage doping too far. Another possibility would be a poor layout so that the charging current spike couples to the output. One chould check this, by looking at the ripple at the filter capacitors too. The charging current effect comes a little later.

[David Hess]

So it was not the reverse recovery, but there is a problem.  Shown below are the ripple at the output and the input.  The output ripple occurs during the time time that the input capacitors are charging, and not after reverse recovery.  I am going to check the layout for return current from charging the input capacitors getting into the control loop.

I added 20 milliamps of output load at 10 volts and it had no effect.  The headroom is like 15 volts.

Note that the power supply has a high output noise level not shown here.  I think I need to filter the reference.

[floobydust]

I've seen that waveform - filter cap ripple current make a mess when the PCB layout/PSU wiring is not right. You need a star ground to the filter cap for the bridge rectifier, filter caps, PSU output and 723, output pot etc. It can be just a takeoff or trace routed wrong.

[David Hess]

Apparently the 723 has horrid line rejection on its V+ pin, so I was setting up to power V+ separately and accidentally applied reverse voltage, damaging the only TO-99-10 723 that I have, so the improve ancient power supply project is over.  I am sure not going to buy a replacement metal can 723 for $20.

I will likely hack together an operational amplifier based reference and regulator using an LM329 reference and LM350 regulator as a pass element.