Drifting bench power supply and low impedance on DMM mV range

[Tuoppi]

Hi all,

I have recently started having electronics as a hobby and reading a couple of books and watching a few videos (Dave, my deepest, wholehearted thanks) and purchased a few pieces of gear to equip my home "lab". As one of those, I managed to get a Thandar TS3022S dual bench power supply in an auction from a company. One of units seems to work ok, but the other one exhibits a weird time drifting behavior. It can drift as much as half volt if left enough time on. This doesn't seem to be a reasonable feature on what is supposed to be a precision power supply. 

So, a couple of days ago I decided to try to something about that and see if I could solve that issue somehow. I started the unit and waited patiently and after a while I could see how the voltage in the indicator started raising (and this seems to match with the readouts from a DMM). Next, I tried unplug the unit and open the case to take a look inside and see if there was anything that would stand out as an obvious problem. There wasn't.

I have a UNI-T 61D DMM which can be plugged to a PC and record the measured values. So my next step was to try to document the phenomena because I had also seen at some point some fluctuations on the output of a few tens of millivolts in a few seconds period. I started recording (output of 12V with no load) and after a few hours the readout hadn't moved more than 10mV. Since I suspected some problem with overheating in the unit and now the cover was removed, I put the cover back and tried again. Nothing. So at this moment the power supply seems to be working fine. It shows a little offset on the voltage (about 70mV) and I even tried to measure the ripple and noise following recent Dave's video instructions (with the limitation that my scope's probes have a 6Mhz bandwidth limit in 1x) and the values are about 2mVrms and 5mVpp which seem to be a little bit out of spec (from specsheet: "typically < 1mV rms") but that could just be me incorrectly measuring. This values of offset and ripple and noise seem to be common for both units.

In any case, now it seems to be working fine for the time being, but it left me scratching my head. So any ideas on what might have been the reason and whether it will show up again are very much appreciated.

While I was trying to measure the output I observed something weird with my UT61D dmm. I tried to record the output in two ways. One of them was using the regular autorange in the V selection and just readout whatever it came out. But since this is a 6000 count multimeter, at 12V you only get 10mV resolution. So, I tried to select manual 600mV range in the mV selection and try to make a relative measurement zeroing the inital value. This seemed to work fine, but then I observed that the power supply was sourcing 6mA (in the V selection this is 0, at least in the readout of the unit which has a 1mA resolution).

I have checked the specifications of the UT61D and it says that for V range, the input impedance is 10MOhm and that for the mV range it should be >3000MOhms. So, I grabbed another dmm (a fluke 179) and measured resistivity on both modes on the UT61D. On the V range it is showing about 11 MOhms, but the funny thing is that in the mV range it shows about 1.54MOhms, way below specs, and actually matching relatively well the 6mA drawn from the power supply . But the measurement in the UT61D is accurate in the mV range, I tried to measure several values in that range and both the Fluke 179 and the UT61D are a match to the tenth of millivolt. So, the UT61D seem to be working fine and at the same time out of spec.

So, does anyone know have a hint on what is going on with both the power supply and the dmm?. Thanks a lot!

[KSP]

The power supply voltage will drift due to its age. The pots wear out and tend to be a bit unstable. Try exercising the V adjustment back and forth over its whole range, this should help clear any residue on the brushes. You see this problem a lot with old power supplies, most use rotary encoders instead these days.

As for the DMM, measuring the input impedance is not as simple as connecting one DMM to another, really you should use an LCR bridge. A DMM will measure resistance, impedance is a whole different ball game

[KSP]

While I was trying to measure the output I observed something weird with my UT61D dmm. I tried to record the output in two ways. One of them was using the regular autorange in the V selection and just readout whatever it came out. But since this is a 6000 count multimeter, at 12V you only get 10mV resolution. So, I tried to select manual 600mV range in the mV selection and try to make a relative measurement zeroing the inital value. This seemed to work fine, but then I observed that the power supply was sourcing 6mA (in the V selection this is 0, at least in the readout of the unit which has a 1mA resolution).

Also, I am not quite sure what you are doing here. You applied 12V to the DMM on the 600mV range? Nulling or not this will saturate the range.

As for the residual current I just simulated your setup on a TTi PL320 supply, and true enough there is about 5 mA residual current displaying on the supply, it is nothing to worry about. There is a setup procedure in the service manual for the TS3022S supplies (and most old TTi supplies) which will show you how to null it out if you really must, but I honestly wouldn't bother

[Andy Watson]

As one of those, I managed to get a Thandar TS3022S dual bench power supply in an auction from a company. One of units seems to work ok, but the other one exhibits a weird time drifting behavior. It can drift as much as half volt if left enough time on.

I had a similar problem with the TS3022. I tracked it down to a pair of back-to-back diodes (actually transistor C-B junctions) that were supposed to protect the op-amp from excessive feedback voltage. One of the diodes had developed a fault and was essentially behaving like a temperature dependant resistor. Do the fine voltage controls on each channel have the same feel? The faulty diode(s) causes the fine voltage control to be overly sensitive.

Edit: Also, the setting of the fine voltage control affected the amount of drift - because the fine control produces an offset voltage across the faulty diodes, whereas the main voltage control is driven to a null.

[KSP]

I had a similar problem with the TS3022. I tracked it down to a pair of back-to-back diodes (actually transistor C-B junctions) that were supposed to protect the op-amp from excessive feedback voltage. One of the diodes had developed a fault and was essentially behaving like a temperature dependant resistor. Do the fine voltage controls on each channel have the same feel? The faulty diode(s) cause fine voltage control to be overly sensitive.

Oo we have a lot of these supplies,  would you mind sharing the part number and location of the component in question? I have always attributed instability to worn pots (and for the most part changing/exercising the pots has been successful), but this would be great help if you still have the info available

[Andy Watson]

Oo we have a lot of these supplies,  would you mind sharing the part number and location of the component in question?

The parts are labelled as Q3 and Q4. The parts list says ZTX239/BC549. I could only obtain BC546, so that is what I fitted and it appears to function.

I have always attributed instability to worn pots (and for the most part changing/exercising the pots has been successful),...

I wouldn't rule out worn pots. One of the key indicators of the diode (transistor) failure is that the fine voltage control is overly sensitive. At 10V output, I found a good channel to have a range of just over 0.5V on the fine control.

[Tuoppi]

The power supply voltage will drift due to its age. The pots wear out and tend to be a bit unstable. Try exercising the V adjustment back and forth over its whole range, this should help clear any residue on the brushes. You see this problem a lot with old power supplies, most use rotary encoders instead these days.

As for the DMM, measuring the input impedance is not as simple as connecting one DMM to another, really you should use an LCR bridge. A DMM will measure resistance, impedance is a whole different ball game

Actually, at the same time that I removed the cover I was also exercising the knob quite a bit and the so far it seems that the drifting is not showing anymore. I wonder if that might have solved that issue. Thanks for the suggestion!.

Regarding the DMM, I suspected that the methodology I used might not be very sound. On the other hand, since this is constant DC, I assumed that the LC parts were irrelevant. In fact, if I do the measurement the other way round (measuring the fluke 179 with the UT61D) I get spot on 10MOhms in the V range and about 11MOhms in the mV range which seems to be in line with the specs.

Also, I am not quite sure what you are doing here. You applied 12V to the DMM on the 600mV range? Nulling or not this will saturate the range.

As for the residual current I just simulated your setup on a TTi PL320 supply, and true enough there is about 5 mA residual current displaying on the supply, it is nothing to worry about. There is a setup procedure in the service manual for the TS3022S supplies (and most old TTi supplies) which will show you how to null it out if you really must, but I honestly wouldn't bother

Well, since the maximum variation range was 500mV and the V measuring range only gives 10mV precision, I was trying to see better the difference by zeroing the absolute value and just measuring the variation. But you are quite right, it is not working. It seems to work though because the UT61D si not giving any sign of reaching the limit once you zero it and seems to be reading fine. But it is in fact not working.

I don't understand what you mean by zeroing the current. It seems that is the actual current which is drawn for the power supply when doing the measurement. Why would I want to zero that out?. In any case, unfortunately the unit came as it is without any additional documentation of any kind, so I don't have the service manual. I have tried to locate it but I cannot find it anywhere. If you know where I could find it that would be really nice to hear!

[Andy Watson]

Actually, at the same time that I removed the cover I was also exercising the knob quite a bit and the so far it seems that the drifting is not showing anymore. I wonder if that might have solved that issue.

If the transistor diodes are at fault the drifting will be less apparent as the fine voltage control is turned up. Try monitoring the output with the fine control turned fully anti-clockwise.

In any case, unfortunately the unit came as it is without any additional documentation of any kind, so I don't have the service manual. I have tried to locate it but I cannot find it anywhere. If you know where I could find it that would be really nice to hear!

It is available on the web, however, it is rather cunning called "S-series". If you search these forums you will find a more exact search term. Try here (thanks to Dimlow):
https://www.eevblog.com/forum/testgear/ts3022s-services-manual-here/msg199310/#msg199310

[Tuoppi]

I had a similar problem with the TS3022. I tracked it down to a pair of back-to-back diodes (actually transistor C-B junctions) that were supposed to protect the op-amp from excessive feedback voltage. One of the diodes had developed a fault and was essentially behaving like a temperature dependant resistor. Do the fine voltage controls on each channel have the same feel? The faulty diode(s) causes the fine voltage control to be overly sensitive.

Edit: Also, the setting of the fine voltage control affected the amount of drift - because the fine control produces an offset voltage across the faulty diodes, whereas the main voltage control is driven to a null.

I wouldn't rule out worn pots. One of the key indicators of the diode (transistor) failure is that the fine voltage control is overly sensitive. At 10V output, I found a good channel to have a range of just over 0.5V on the fine control.

I have tested the fine tune control at 10V. The unit which exhibited the drifting has a range of exactly 0.5V in the fine control. The other one has about 0.6V. Acording to what you said this should be ok, shouldn't it?. In any case, your comment regarding the diodes seems worth checking. As I was mentioning to KSP unfortunately I couldn't find the service manual. Although I guess I could probably locate those diodes based on your pics. Thanks  a lot for the hint!.

Also, the drifting pattern didn't seem to match a heating profile, and sometimes it would just oscillate a few tens of millimiters up and down. Also the drift would happen in a period of maybe 2-3 hours, and it seems to me that the power source should have reached a constant thermal state far earlier than that if there is no variation on the load.

[Andy Watson]

I have tested the fine tune control at 10V. The unit which exhibited the drifting has a range of exactly 0.5V in the fine control. The other one has about 0.6V. Acording to what you said this should be ok, shouldn't it?.

Could be - depends how bad (or not) the transistors are. 0.5/0.6V sounds good. On my unit it was very apparent within a few minutes of switching on, also, the drift would tend to be in the same direction (as the temperature increased).

[Tuoppi]

Could be - depends how bad (or not) the transistors are. 0.5/0.6V sounds good. On my unit it was very apparent within a few minutes of switching on, also, the drift would tend to be in the same direction (as the temperature increased).

It was the same for me and that is why I was suspecting something related to heating. Now, it seems to have dissapeared completely. I need to do further testing to see if it is really so and try with the fine tune knob fully turned anti-clock-wise as you suggest.

I followed your suggestion and managed to find the thread here in the eevblog you mentioned. The link they mentioned there was dead but after some googling for the 'S' series service manual I managed to locate it. That is great since then I will be able to hopefully adjust it and remove the 70mV offset it shows at the moment.

Excellent hints, many thanks!. I will update here the results once I do more thorough tests. For the time being I have had them about 15 minutes on and maybe that the unit has drifted one count in the dmm and voltage meter of the PSU but no more. In the past it changed far more quickly. The other unit hasn't changed a bit

[Tuoppi]

Hi,

I have tested the units a little bit more, the faulty one for a little bit more of 4 hours and the other one for about 3. It actually seems to be some drifting on the left unit at least. For this test I turned the fine control completely anti-clock wise as Andy suggested. The absolute value is not even close to what it used to be at the beginning, but definitely something can be seen.

I have also tested the C-B and E-B connectors of Q3 and Q4 with the diode mode of the dmm and all of them show 0.735V-0.740. Any measurements that I have done on those are identical for both units to the decimal point.

So, what do you think is the best bet?, worn pot or faulty transistors?. Any advice on which would be the best way to proceed for doing additional tests? In any case, thanks a lot for the very nice advices so far.

[KSP]

I don't understand what you mean by zeroing the current. It seems that is the actual current which is drawn for the power supply when doing the measurement. Why would I want to zero that out?. In any case, unfortunately the unit came as it is without any additional documentation of any kind, so I don't have the service manual. I have tried to locate it but I cannot find it anywhere. If you know where I could find it that would be really nice to hear!

Sorry I thought you meant just an offset on the meter, a lot of the old TTi supplies I have used tend to have a minor offset on the meters, but they can be adjusted. But tue enough, it may just be the current drawn during testing

[Tuoppi]

Hi all,

This last weekend I got some time to get back to this and see if I could locate the source of the drifting. I started by calibrating both units according to the service manual. The drifting was still there (this time decreasing voltage). So, I continued by measuring resistance across Q3 and Q4 and testing what would happen when blowing some hot air to the transistors. The result was that resistance would change significantly, more than 10%, and that it was oscillating wildly. So, no monotonically increasing or decreasing with heating or cooling. For comparison purposes I also tried with the "proper" power supply unit and variation was way smaller and consistent with heating and cooling.

So, it was clear that there was something strange going on. Then, out of the sudden, the unit just went to +42V output voltage and it wouldn't change no matter what I would do on the fine and coarse adjustment pots. My first guess is that there was something happening around the opamp controlling the voltage level, so I tested the voltage on the inputs of the opamp and as expected they were not the same and the opamp output was saturated to the positive rail.

I attach a diagram of the circuitry and voltages surrounding the opamp. The voltages listed are measured with respect to the positive output terminal. If I understand right how this system works, the output of the opamp should change so that the voltage at the negative output terminal of the power supply is such that the voltage of the non-inverting input of the opamp matches the value of the inverting input. Because of Q3/Q4, the voltage at the non-inverting input must always be clamped between -0.6972V and 0.6972V.  So, the measured voltage at the input of the inverting input of the opamp only makes sense if the input has shorted to the negative rail and is actually sinking current, right?. I am quite convinced of this, but I wanted to check what was your opinion before I start replacing the opamp. Also, do you think this might be the origin of the original drifting or could be a totally unrelated issue?.

My local supplier has in stock the TL072, so I could change part for part. But I was wondering if it would make sense to replace the opamp for a similar less noisy alternative and whether this would have an effect on reducing the noise on the unit output. For example, for about 0.2€ more, they also have the NE555AP which is also a dual opamp which seems to have very similar specs but higher CMRR and lower noise. They also have for 8-10 times the price some OPA2134, but that one seems to have very similar CMRR and noise specs to those of the NE555AP. Just for educational purposes I would like to get a 8 pin socket and try both TL072 and NE555AP and see what effect it has on the output. But if I add a socket, would that have some effect on the amount of noise?. In fact, I am not even sure that this opamp is the main source of noise on the output of the power supply or is perhaps coming from some other components. What do you think guys?.

[Andy Watson]

I would replace like with like. NE5532 have a higher bandwidth and you may find that you have stability problems by changing the loop characteristics. TL072 are plenty adequate. 

Did you check Q3 and Q4 by removing them from the circuit?

[Tuoppi]

I would replace like with like. NE5532 have a higher bandwidth and you may find that you have stability problems by changing the loop characteristics. TL072 are plenty adequate. 

Did you check Q3 and Q4 by removing them from the circuit?

No. I haven't. My skills wielding the welding iron are not exactly those of an artist so I try to keep my use of it to a minimum. Maybe with time I will get more confident on swapping parts in and out, but for the time being I am not. In any case, tomorrow I will try to pass by my local supplier and get the opamp and a few BC549's and if after replacing the TL072, and hopefully fixing the high voltage issue, the drifting persists I will remove Q3 and Q4.

[Tuoppi]

And update on this.

I replaced the opamp and that solved the issue about the voltage going to maximum. After this, also the readings of resistance across Q3/Q4 were stable. I checked the drift and it seemed that there was still 25-30 millivolts variation (downwards) in a one hour period or so. I thought this might be too much for just the unit warming up, so I went ahead and removed Q3 and Q4 (out of circuit they still seemed to test ok) and replaced them for new ones as Andy suggested. 

In the short time I could test the unit after the changes it seems that there might be no more drift anymore. I need to do a lengthier tests with some data logging to analyze it properly but it is looking good.

Thanks guys for the advices, I have learnt a lot troubleshooting this.