Power supply design (2)

[liquibyte]

We've solved it.  The new 2141's don't suffer from this, you are correct.  Audioguru over at e-lab was working with me on this so pin 4 stays at the revised ground point  with R10 but adding in Q1 again with a voltage divider across its base of 12K and 160R to the negative ground solved the transient issue.  I was getting 7.6V until he suggested I may have the collector and emitter reversed.  Apparently the datasheet I referred to for the pinouts was different than the 2222A I was using.  Radio Shack parts, go figure.  I don't think I've been this happy to have figured something out in my life.  I've also learned that I hate working with surface mount components.

[David Hess]

We've solved it.  The new 2141's don't suffer from this, you are correct.  Audioguru over at e-lab was working with me on this so pin 4 stays at the revised ground point  with R10 ...

That is right.  I did not examine the offset voltage correction circuit carefully because it is not needed; I would have left it out.  On the TLE2141, it is suppose to connect to the negative supply.

From the TLE2141 datasheet:

Inputs can operate between VCC? ? 0.3 to VCC+ ? 1.8 V without inducing phase reversal

It is sometimes a good idea to include external glitch suppression even if the operational amplifier does not suffer from phase reversal because the operational amplifier may be changed in the future by someone unaware of the issue.

I was getting 7.6V until he suggested I may have the collector and emitter reversed.  Apparently the datasheet I referred to for the pinouts was different than the 2222A I was using.  Radio Shack parts, go figure.  I don't think I've been this happy to have figured something out in my life.  I've also learned that I hate working with surface mount components.

Just wait until you reverse the collect and emitter of a bipolar transistor and the circuit works better.

[liquibyte]

Just wait until you reverse the collect and emitter of a bipolar transistor and the circuit works better.

I can't even begin to imagine a scenario where this would be apparent to me in any way.

I'm good at following directions and I'm desperate to learn the intricacies of how all this works but I still need to go to the various calculators online to help work through circuits and even then some things a lot of you guys take for granted stump me.  One of these days it'll click I'm sure but for now I still struggle at times, especially when trying to understand op amps.  I solder better than anyone I've ever met though, so there's that.

Edit:  I've attached a modified version of the parts list and schematic for posterity.  From what I've seen so far, the power supply now performs admirably for what it's worth.

[Kevin.D]

That is right.  I did not examine the offset voltage correction circuit carefully because it is not needed; I would have left it out.  On the TLE2141, it is suppose to connect to the negative supply.

He did have it connected correctly in his revised circuit with the glitches .
I think also the schottky ( maybe a schottky Vf drop might be just a slight bit to much for the  -0.3 this opamp gives as max)  may have solved it for him without Q1  .
 Dont forget D10 and that R15 (or some scheme) for your Q2 Veb(max)protection liquib .

[liquibyte]

Dont forget D10 and that R15 (or some scheme) for your Q2 Veb(max)protection liquid .

The guy that did the revision seemed to think that these parts really didn't add anything to the circuit.

R15 was removed because it didn't do anything useful and it wasted valuable output drive voltage. D10 also never did anything and still doesn't do anything.

I've posted over there about it but I can't imagine that it was in the original circuit for no reason, someone thought it important enough to include.

[Kevin.D]

Dont forget D10 and that R15 (or some scheme) for your Q2 Veb(max)protection liquid .

The guy that did the revision seemed to think that these parts really didn't add anything to the circuit.

Heres a little experiment for you .
Go  get a known good   bjt comparable to Q2 and measure it's HFE then pass a ~10ma  reverse current through the Veb junction for a just a few seconds then remeasure it's gain afterwards .(Vbe breakdown at ~ 7V- so in your supply you would get about ~ 7mA through 1k R16 and Q2 junction if you connect a ~13.8V battery to your supply output and turn your V setting down ) .
I think the gain stops reducing when it reduces to a certain point   but how low it goes is probably variable between devices and depends on reverse current.

A little extra math shows that once more than 7mA is flowing through R16(1k) and Q2 then you will have enough drop between the emitters and bases of  Q5 and Q4 for them to also start breaking down (so a voltage of >~ 15V applied to the output will be enough to damage all you pass bjt's ).

[David Hess]

Go  get a known good   bjt comparable to Q2 and measure it's HFE then pass a ~10ma  reverse current through the Veb junction for a just a few seconds then remeasure it's gain afterwards .(Vbe breakdown at ~ 7V- so in your supply you would get about ~ 7mA through 1k R16 and Q2 junction if you connect a ~13.8V battery to your supply output and turn your V setting down ) .
I think the gain stops reducing when it reduces to a certain point   but how low it goes is probably variable between devices and depends on reverse current.

A little extra math shows that once more than 7mA is flowing through R16(1k) and Q2 then you will have enough drop between the emitters and bases of  Q5 and Q4 for them to also start breaking down (so a voltage of >~ 15V applied to the output will be enough to damage all you pass bjt's ).

Hfe degradation because of base-emitter reverse breakdown is well known (although maybe not here) but only a significant problem at low collector currents like you would find in a differential input stage which is why precision and low input bias current bipolar operational amplifiers have protection against excessive differential input voltages.  Even a little bit of reverse breakdown will ruin them.  It can be ignored with transistors operating at high collector currents which will have low gain anyway.

In a power supply, I would worry more about reverse breakdown destroying the output transistors if the input supply shorts and the output capacitance discharges through the pass transistors.  Integrated regulators sometimes die from that also.

[David Hess]

Just wait until you reverse the collect and emitter of a bipolar transistor and the circuit works better.

I can't even begin to imagine a scenario where this would be apparent to me in any way.

It is not widely advertised but bipolar transistors will work with their collector and emitter reversed.  They have low gain (maybe even be below 1) and low breakdown voltage when operated this way but also have a very low saturation voltage.

They used to manufacture bipolar transistors with a more symmetrical emitter and collector for special applications like choppers and switches but JFETs and MOSFETs took over that role.  I still occasionally see a circuit design that takes advantage of a reversed bipolar transistor as a low saturation voltage reset switch.  The advantage over a JFET or MOSFET is that it can be turned on with only 0.6 volts so it is handy in low voltage circuitry.

[liquibyte]

On the suggestion of another member I tested moving the collector to the input of U2.  Still getting a voltage spike but not near as bad as it was.  I'm starting to think the way to mitigate this may be some sort of isolation scheme such as a relay but I could also be way off too.  The scope I have is old but it does agree with what I'm seeing with my meter which is also not very good.  One of these days I intend on getting proper equipment but for now this is what I can offer in terms of data.  What follows is what I posted over there.

I've tested with the collector of Q1 at pin 3 and pin 6 of U2.  Given my meters error rate (and that of my eyes), the results are roughly the same.  There is definitely a voltage rise happening on power on.  Once I get above 1V the auto resolution of the meter probably hides the rise I get on power off but starting at 800mV and under I can see a definitive rise to around 1.25V - 1.5V as the caps drain out over many seconds.  I wouldn't want to keep something that required very low power levels hooked up on power off because all control seems to be lost.

To test consistently, I waited to power on once my meter settled to under 200mV as the caps drained and then turned on.  I did this several times to make sure it was consistent and to compensate for my brain not registering the number the first go around.

         Pin 3         Pin 6

30V      36.75         36.6
25V      33.06         32
20V      29            28
15V      23.56         23
10V      16.23         16.69
5V       7.95          8.56
3.3V     5.49          5.6
2V       3.6           3.9
1.5V     3.2           3.4
1.2V     3             2.8
1V       2.7           2.6
.8V      2.2           2.2
.5V      1.6           1.5
.2V      .9            .3
.1V      .36           .3
0V      Too small and quick to see (39.2mV == 0)

[David Hess]

I think you are looking at the wrong thing or at least making the wrong comparison.  The control group would be the design using the original TL081 and no Q1.

And if Q1 was wired to pull pin 3 of a TL081 down to ground, it could cause the very problem that it was added to solve.

The TLE2141 should make Q1 superfluous anyway because it is not suppose to suffer from phase reversal.

[liquibyte]

I don't understand op amps enough to know what phase reversal really is.  Q1 definitely has had an effect on the transient spikes which is what we're trying to mitigate along with some ringing that was observed.  I'm just beginning to understand some of this and, most assuredly, don't understand all of it.

So far, there's only been two people that have been testing this design to try and work out the bugs, the third guy has never built one but made all the changes from the TL081's to the TLE2141's etc.  The other guy just got a DSO and has been working on the noise and ringing issue and sent me one of his boards so I could look at the issue along side the transient problem after replacing U1 with a voltage regulator.

For me, the transients are the thing my scope can see and what I noticed most, the noise is non-existent or has been dealt with in the design with the regulator in place of U1 although the ones I built aren't bad.

When you posted the model of the PS you were discussing, I got to looking at the circuit compared to this one and the original and made the leap to Q1's removal contributing somewhat to the issue.  I haven't tried this with the version that has a U1 in it yet, I have to build a third board up to do that.  Having said all that, the guy that hasn't built one is the one that suggested the pin 3 swap to avoid having to put D10 and R15 back in, which I haven't done yet.  I haven't tested this loaded either, just unloaded at the output and have been very careful to avoid shorts and such.

At this point I kind of wish that I'd just bought a supply.  I could have gotten something decent to use and done this one later.  As I said before, it's still been fun and I am learning, albeit slowly.  I'm going to take one of my spare boards and build it up and start testing it as well to see if I get the same results as redwire's version.  I suppose that I should make a version of the original as well just to be thorough.  It was my understanding that the original suffered some serious issues due to underrated components etc.

[David Hess]

I don't understand op amps enough to know what phase reversal really is.  Q1 definitely has had an effect on the transient spikes which is what we're trying to mitigate along with some ringing that was observed.  I'm just beginning to understand some of this and, most assuredly, don't understand all of it.

Phase reversal means that when the input voltage range is exceeded, the output reverses.  In extreme cases, this can cause the operational amplifier to latch up and even fail.  In modern operational amplifiers which suffer this problem, it usually comes about when either the inverting or non-inverting input is pulled too close to ground or below ground causing the output to shift to the positive rail.

Page 11 of the LT1055 datasheet has a pretty good explanation:

http://www.linear.com/docs/2885

The ubiquitous LM324, LM358, and the TL071 and TL081 series all suffer from this problem.

At this point I kind of wish that I'd just bought a supply.  I could have gotten something decent to use and done this one later.

You would not have learned as much though.  Maybe we need a simplified modern lab power supply design that has been thoroughly vetted for problems.

[liquibyte]

I don't understand op amps enough to know what phase reversal really is.  Q1 definitely has had an effect on the transient spikes which is what we're trying to mitigate along with some ringing that was observed.  I'm just beginning to understand some of this and, most assuredly, don't understand all of it.

Phase reversal means that when the input voltage range is exceeded, the output reverses.  In extreme cases, this can cause the operational amplifier to latch up and even fail.  In modern operational amplifiers which suffer this problem, it usually comes about when either the inverting or non-inverting input is pulled too close to ground or below ground causing the output to shift to the positive rail.

Page 11 of the LT1055 datasheet has a pretty good explanation:

http://www.linear.com/docs/2885

The ubiquitous LM324, LM358, and the TL071 and TL081 series all suffer from this problem.

Thanks for the explaination.  The reason for not pulling pin 3 below ground through Q1 though.  That actually makes more sense now.  I'm going to find reading material now.

At this point I kind of wish that I'd just bought a supply.  I could have gotten something decent to use and done this one later.

You would not have learned as much though.  Maybe we need a simplified modern lab power supply design that has been thoroughly vetted for problems.

I have begun to see patterns in circuits even when they're arranged differently so I guess that's a good thing.  I just wish my math were better.

I hope to have a better scope within a year or so but to be honest it took me a couple of years to get around to a temp controlled iron which I just got on Friday.  I can solder up a mean board with a firestick though.  I'm not sure if it'll help though.  I seem to do fine without even using one but they are handy for helping to mitigate noise issues.

I picked this supply out of the n^th versions for its simplicity figuring it had already been thoroughly vetted given its age.  I was wrong on that note but am now determined to see it through to completion.  To be honest, the versions I built vs. the one I've been using to test Q1 are more stable on shutdown though they do suffer the same spikes as the one with Q1 on power on but don't have it.  There's something odd going on with the surface mount version with the voltage regulator in place of U1.  I have a dual version that I built and I'm going to pull one of the boards out of the case and hack in a Q1 and test side by side and see how things behave.

[David Hess]

Thanks for the explaination.  The reason for not pulling pin 3 below ground through Q1 though.  That actually makes more sense now.  I'm going to find reading material now.

It does not necessarily need to be pulled below the negative supply.  With some operational amplifiers, it just needs to be pulled below (or above) the common mode input voltage range and that may be a volt or more away from either supply voltage.

The ancient LM709 was the one that would latch up internally under these conditions and self destruct although their may have been others.  This could happen if the power supplies were not sequenced correctly so when the LM741 and LM301 came out advertising "no latch-up when the common mode range is exceeded" to replace it, there was much rejoicing.

Operational amplifiers which suffer from phase inversion may latch up depending on their external circuit configuration.

[paul18fr]

Dear All contributors

I'm quite interested by this "project" and I hope that a consensus will be found to freeze a design;  for hobbyists (that my point of view at least), it's interesting, rewarding and motivating to build some materials (PSU, basic function generators and so on).

I cannot hide the economical aspect as well 

Paul

[liquibyte]

Maybe we need a simplified modern lab power supply design that has been thoroughly vetted for problems.

While I am trying to learn, these guys are light years ahead of me on this subject so I'll defer to them.

I can add that I did add Q1 with the voltage divider across the base into one of my supplies and that it is working very well except that I'd recommend not having anything sensitive hooked up to it on power on.  There is still a P1 voltage dependent spike happening when the power is switched on.

There is some question as to RV1 even being needed as well.  I can state that on my version it does nothing measurable even though it's supposed to zero the voltage and I'll probably remove it going forward based on the testing this guy did with his version which also has Q1 I might add.

I can tell you this does work, I have two of them in one case and use them.  I never turn them on hooked up to a circuit unless I'm absolutely sure the components are rated above what the spike produces.  At 30V, you'll get a 36-37V spike.  At 12V, you'll get a 19-20V spike.  At 3.3V, you'll get a 5-6V spike.  If you can live with this, build yourself one.  The nice part is that the components are reusable if you decide to go with something else.  The transformer, pots, and pass transistors doubly so.

[David Hess]

I can add that I did add Q1 with the voltage divider across the base into one of my supplies and that it is working very well except that I'd recommend not having anything sensitive hooked up to it on power on.  There is still a P1 voltage dependent spike happening when the power is switched on.

If the voltage spike is above the set voltage, then something is wrong.  I would find such behavior in a power supply unacceptable.

There is some question as to RV1 even being needed as well.  I can state that on my version it does nothing measurable even though it's supposed to zero the voltage and I'll probably remove it going forward based on the testing this guy did with his version which also has Q1 I might add.

RV1 works through the operational amplifiers offset adjustment pins.  In most power supplies, the effect from the error amplifier's offset voltage is insignificant.  Further, if the offset voltage adjustment is used to remove other sources of offset, it will likely degrade the error amplifier's offset voltage drift although again, in most power supplies that will be an insignificant effect.

Who cares if the 0 to 30 volt power supply can only go down to 15 millivolts instead of 0 millivolts?

I can tell you this does work, I have two of them in one case and use them.  I never turn them on hooked up to a circuit unless I'm absolutely sure the components are rated above what the spike produces.  At 30V, you'll get a 36-37V spike.  At 12V, you'll get a 19-20V spike.  At 3.3V, you'll get a 5-6V spike.  If you can live with this, build yourself one.  The nice part is that the components are reusable if you decide to go with something else.  The transformer, pots, and pass transistors doubly so.

That is outrageous.  I would consider the design broken if it produced a start-up or shut-down output spike like that.

[liquibyte]

I took the liberty of posting the parts and costs etc. below sourced from Digikey.  I chose fairly good quality parts so you could probably end up a fair bit cheaper but don't forget to factor in shipping as it adds up quick across multiple orders.

Code: [Select]

Value Qty Part Number Each Total

0.47 16W =1 (A102127-ND) $3.06 $3.06
2.2K 2W =1 (RSMF2JT2K20CT-ND) $0.43 $0.43
82 2W =1 (RSMF2JT82R0CT-ND) $0.43 $0.43
0.33 2W =2 (RSF2JTR330CT-ND) $0.56 $1.12
1.5K 1W =1 (RSF1JT1K50CT-ND) $0.35 $0.10
220 1/2W =1 (CF12JT220RCT-ND) $0.14 $0.14
68 1/4W =1 (CF14JT68R0CT-ND) $0.10 $0.10
150 1/4W =1 (RNF14FTD150RCT-ND) $0.15 $0.15
1K 1/4W =3 (CF14JT1K00CT-ND) $0.10 $0.30
2.2K 1/4W =2 (CF14JT2K20CT-ND) $0.10 $0.20
10K 1/4W =5 (CF14JT10K0CT-ND) $0.10 $0.50
27K 1/4W =2 (CF14JT27K0CT-ND) $0.10 $0.20
33K 1/4W =1 (CF14JT33K0CT-ND) $0.10 $0.10
56K 1/4W =1 (CF14JT56K0CT-ND) $0.10 $0.10
Total = $7.18

5K trimmer =1 (3296W-502LF-ND) $2.41 $2.41
20K trimmer =1 (3296W-203LF-ND) $2.41 $2.41
100K trimmer =1 (3296W-104LF-ND) $2.41 $2.41
10K 10 turn pot =2 (987-1523-ND) $11.99 $23.98
Total = $31.21

15000uF 63V =1 (399-5658-ND) $16.04 $16.04
47uF 50V =2 (ESW476M100AH2AA) $0.48 $0.96
10uF 50V Film =1 (399-5999-ND) $2.70 $2.70
220nF film =1 (399-6037-ND) $0.30 $0.30
100nF film =2 (399-5861-ND) $0.22 $0.44
330pF ceramic =1 (399-4173-ND) $0.40 $0.40
100pF ceramic =2 (399-9707-ND) $0.44 $0.88
Total = $21.72

6-10A 50V =1 (GBPC15005FS-ND) $2.86 $2.86
1N4148 =6 (1N4148TACT-ND) $0.10 $0.60
1N4001 =1 (1N4001FSCT-ND) $0.18 $0.18
BZX79C5V6 =1 (BZX79C5V6-ND) $0.13 $0.13
BZX85C10 =1 (BZX85C10-ND) $0.22 $0.22
LED =1 (365-1189-ND) $0.14 $0.14
Total = $4.13

2N4401 =1 (2N4401-ND) $0.22 $0.22
BC557 =1 (BC557BGOS-ND) $0.38 $0.38
BD139 =1 (BD13910S-ND) $0.46 $0.46
2N3055 =2 (2N3055GOS-ND) $2.34 $4.68
TLE2141 =3 (296-10456-5-ND) $1.81 $5.43
Total = $11.17

28V 4.6A 130VA =1 (237-1281-ND) $27.34 $27.34

Grand Total = $102.75$102.75, HOLY CRAP!!!

Let's look at this though.  Source your own transformer, 2N3055's, big filter cap, trimmers and pots, bridge rectifier, and 0.47R resistor.

Transformers can be had many places and used isn't usually an issue as long as it can supply 28V @ 4A.  You can get 2N3055's elsewhere cheaper if you look, just avoid the counterfeit ones.  I'd recommend not skimping on the filter cap but the value I got was 15,000uF, the design only calls for 12,000uF so you could probably save something there.  The trimmers and pots I used are all 10 turn versions and rather expensive.  You can get away with single turn versions if you'd like for much, much cheaper.  I used 10 turn trimmers so that the PS can be calibrated more accurately and 10 turn pots for better resolution.  You can get away with a much cheaper bridge rectifier, I over engineered here by a lot.  You can also get a non-aluminum resistor for this, the design calls for 0.47R @ 10W.  I couldn't find a 10W aluminum resistor at 0.47R so I went up.  I'm glad I did though, it gets hot under load.

You ready for this?   $17.56 if you take out everything I just said you can source elsewhere.  You can get even cheaper if you order the more common values in bigger quantities but cheaper is relative here because you're still spending money.  This cost is for only the quantity of components required.

[liquibyte]

I can add that I did add Q1 with the voltage divider across the base into one of my supplies and that it is working very well except that I'd recommend not having anything sensitive hooked up to it on power on.  There is still a P1 voltage dependent spike happening when the power is switched on.

If the voltage spike is above the set voltage, then something is wrong.  I would find such behavior in a power supply unacceptable.

That's the thing I'm trying to fix.

There is some question as to RV1 even being needed as well.  I can state that on my version it does nothing measurable even though it's supposed to zero the voltage and I'll probably remove it going forward based on the testing this guy did with his version which also has Q1 I might add.

RV1 works through the operational amplifiers offset adjustment pins.  In most power supplies, the effect from the error amplifier's offset voltage is insignificant.  Further, if the offset voltage adjustment is used to remove other sources of offset, it will likely degrade the error amplifier's offset voltage drift although again, in most power supplies that will be an insignificant effect.

Who cares if the 0 to 30 volt power supply can only go down to 15 millivolts instead of 0 millivolts?

It's around 35 millivolts actually, not that it ever bothered me.  RV1, according to audioguru over there, was supposed to be able to take the output to true zero.  It doesn't.

I can tell you this does work, I have two of them in one case and use them.  I never turn them on hooked up to a circuit unless I'm absolutely sure the components are rated above what the spike produces.  At 30V, you'll get a 36-37V spike.  At 12V, you'll get a 19-20V spike.  At 3.3V, you'll get a 5-6V spike.  If you can live with this, build yourself one.  The nice part is that the components are reusable if you decide to go with something else.  The transformer, pots, and pass transistors doubly so.

That is outrageous.  I would consider the design broken if it produced a start-up or shut-down output spike like that.

I agree, it's unacceptable as it stands.  I'm attempting to learn enough to fix it though.  To be honest, I'm not even sure where it's coming from.  I have a hunch that it's got something to do with the transformer inrush but I always thought that the filter caps helped there.  I could be way off though, I'm a newbie.  What I don't get is that people have been talking about this design since 2003 and no one has noticed this enough to fix it.  They've blown up plenty of TL081's though, which is why it was redesigned with the 2141's from what I understand.

I'd like to sim the circut but I suck at spice because I get absolutely lost.  That and I could never find a model for the 2141.

[David Hess]

RV1 works through the operational amplifiers offset adjustment pins.  In most power supplies, the effect from the error amplifier's offset voltage is insignificant.  Further, if the offset voltage adjustment is used to remove other sources of offset, it will likely degrade the error amplifier's offset voltage drift although again, in most power supplies that will be an insignificant effect.

Who cares if the 0 to 30 volt power supply can only go down to 15 millivolts instead of 0 millivolts?

It's around 35 millivolts actually, not that it ever bothered me.  RV1, according to audioguru over there, was supposed to be able to take the output to true zero.  It doesn't.

Which just means that there is additional offset outside of the amplifier creating the other problem I just described.  I have never measured the offset on any of my PS-503s ...

On the one PS-503 that I have plugged in at the moment, its offset is -6.1 millivolts on one output and -7.6 millivolts on the other, meaning that the output ranges extend a few millivolts on the other side of zero, however the design deliberately arranges for the outputs to always go slightly negative given the worst case input offset voltages of the error amplifiers.  It would be easy to trim them to zero (without the operational amplifier offset pins!) but offhand I cannot think of a reason this would be important.

I checked for glitches at power-up and power-down both with the power switch and the standby switch and in no event could the output rise about the current voltage setting and there were no visible glitches as such.  There was a tiny millivolt level fast glitch at turn on which I suspect is common mode noise since the PS503 outputs are floated with respect to chassis ground.

If there was no load on the output and the output capacitor was not discharged, then the voltage sometimes dropped before rising again when the power switch was cycled which I suspect is caused by the action of those two protection transistors we discussed earlier.

I can tell you this does work, I have two of them in one case and use them.  I never turn them on hooked up to a circuit unless I'm absolutely sure the components are rated above what the spike produces.  At 30V, you'll get a 36-37V spike.  At 12V, you'll get a 19-20V spike.  At 3.3V, you'll get a 5-6V spike.  If you can live with this, build yourself one.  The nice part is that the components are reusable if you decide to go with something else.  The transformer, pots, and pass transistors doubly so.

That is outrageous.  I would consider the design broken if it produced a start-up or shut-down output spike like that.

I agree, it's unacceptable as it stands.  I'm attempting to learn enough to fix it though.  To be honest, I'm not even sure where it's coming from.  I have a hunch that it's got something to do with the transformer inrush but I always thought that the filter caps helped there.  I could be way off though, I'm a newbie.  What I don't get is that people have been talking about this design since 2003 and no one has noticed this enough to fix it.  They've blown up plenty of TL081's though, which is why it was redesigned with the 2141's from what I understand.

The only way the output can rise in this design is if the output of the operational amplifier rises for whatever reason or if there is leakage through the pass elements.

One thing I would try which is part of the PS-503 design is to add a low current preload from the output to the low voltage negative bias supply.  A 6.8k resistor would be about right.

I'd like to sim the circut but I suck at spice because I get absolutely lost.  That and I could never find a model for the 2141.

Simulating startup of an operational amplifier strikes me as pretty difficult to do.  Some operational amplifier models will definitely not handle it correctly if only because they do not always model common mode input range violations.

[liquibyte]

One thing I would try which is part of the PS-503 design is to add a low current preload from the output to the low voltage negative bias supply.  A 6.8k resistor would be about right.

You just lost me here.  I've tried several times to formulate the right question but you really lost me at output.  My mind wants to think of output as pin 6 on one of the op amps but I can also think of it as output of the power supply.  Honestly, my mind also wants to think "low voltage negative bias supply" means U3 and I think that corresponds to U55 but I'm not seeing the equivalent pre-load in the service manual.  I think if I just start hooking a 6.8K resistor to various outputs and different pins of U3 I'm going to blow something up or at least damage it past the point of working right and make tracking down the problem even worse.

[David Hess]

One thing I would try which is part of the PS-503 design is to add a low current preload from the output to the low voltage negative bias supply.  A 6.8k resistor would be about right.

You just lost me here.  I've tried several times to formulate the right question but you really lost me at output.  My mind wants to think of output as pin 6 on one of the op amps but I can also think of it as output of the power supply.  Honestly, my mind also wants to think "low voltage negative bias supply" means U3 and I think that corresponds to U55 but I'm not seeing the equivalent pre-load in the service manual.  I think if I just start hooking a 6.8K resistor to various outputs and different pins of U3 I'm going to blow something up or at least damage it past the point of working right and make tracking down the problem even worse.

I mean the PS-503 has a preload on the output of the power supply although you have to look hard to find it.  The preload resistors are R85 and R192 and attached directly to the emitters and collectors of the output transistors and go to the opposite bias supply.

On yours, the resistor would connect directly from the output of the power supply to the -5 volt bias supply which powers the operational amplifiers.

[liquibyte]

I mean the PS-503 has a preload on the output of the power supply although you have to look hard to find it.  The preload resistors are R85 and R192 and attached directly to the emitters and collectors of the output transistors and go to the opposite bias supply.

I thought those might be it but wasn't sure.

On yours, the resistor would connect directly from the output of the power supply to the -5 volt bias supply which powers the operational amplifiers.

Still not doing anything to mitigate the spikes.  I've also tried with a 1K to load it more because I'm measuring -1.5V and not -5V but still not making a dent in my readings.  I've been testing at 10V to make seeing things and the mental math easier, both between what my meter is seeing and what my scope is seeing.  I'm getting a consistent 7V spike at that 10V.  I'm starting to think that this circuit is just plain failure and not a whole lot can be done about it.  It's a shame really because other than the transient on power up, the rest of it works rather well.

[David Hess]

What does the output of U2 do at startup?

What does the non-inverting input of U2 do at startup?

I've been testing at 10V to make seeing things and the mental math easier, both between what my meter is seeing and what my scope is seeing.  I'm getting a consistent 7V spike at that 10V.  I'm starting to think that this circuit is just plain failure and not a whole lot can be done about it.  It's a shame really because other than the transient on power up, the rest of it works rather well.

I am confused about this.  With the output set to 10 volts you get a 17 volt spike or it spikes at 7 volts before rising to 10 volts?

[liquibyte]

What does the output of U2 do at startup?

What does the non-inverting input of U2 do at startup?

With the output set at 10 volts, pin 3 sits at 3.74 volts.  When I measure during power on at pin 3 the voltage rises to 6.3 volts before falling back to 3.74 volts.  The same thing happens at the inverting input.

Pin 6 outputs 10.65 volts.  On power on I get 17.65 volts.

I've been testing at 10V to make seeing things and the mental math easier, both between what my meter is seeing and what my scope is seeing.  I'm getting a consistent 7V spike at that 10V.  I'm starting to think that this circuit is just plain failure and not a whole lot can be done about it.  It's a shame really because other than the transient on power up, the rest of it works rather well.

I am confused about this.  With the output set to 10 volts you get a 17 volt spike or it spikes at 7 volts before rising to 10 volts?

With the output set to 10 volts, I get a spike up to 17 volts before falling back to 10 volts as it stabilizes.  Since I don't have a DSO, I can only eye it but it looks like this happens over about a half of a second.  It's hard to eye it but I have the scope set at 50mS and it looks like it spans between 5 and 7 cm before it settles.  Keep in mind that this scope is an old Conar 255 that I haven't calibrated but it does seem to agree with my meter more or less.