ADC: It's a current steering circuit, and U32 switch is used to control that current steering. The position of the analogue switches confused me for a while, the outputs of the switches are at virtual ground at the U28 op-amp inputs......so we were never gonna see any sort of measureable voltage at the right side of the analogue switches......only virtual 0Vdc. EEProm: As pointed out by others the 93C66 EEProm can be subject to data loss/corruption through bit flip issues, and did you notice the GPIB address was 8 at the start, then 9 at the end! I've also played with the 'faulty' EEProm by writing b00000000, b11111111, b10101010, b01010101 at every location then read it back and look for any changed or swapped bits but so far it's been fine. Need to do more testing by leaving it a few hours then reading back again. I'd love to find something to conclude the repair!
I was going to comment exactly the same explanation. Reading 0V on the non-inverting input of that op-amp actually means it’s working properly. If there is anything other than zero there, then that would indicate a problem in that area. Great work, Ian.
@@garyramsey4275what caught me out was not expecting the switches fitted after the series resistors and right at the op-amp inputs! I’d be tempted to reverse them if I was designing the circuit.
@@IanScottJohnstonyes, if those resistors were after the analog switch, it would be much easier to follow the signal. I wonder why they chose to do it this way?
@@garyramsey4275 R72 needs to be where it is, or the -10 VREF is shorted by the switch, but a resistor on the non-grounded output would not hurt for sure.
Nice repair, funny how we have tend to have a favourite bit of test kit. I used to get a bit angry if other engineers messed with my bench setup, letting a bit go off for cal was like losing an arm for a month. 🙂
Great work, Ian! My guess is the EEPROM was defective. The strange current and voltage readouts lead me to believe some calibration constants were changing. Copying the contents to a new chip transferred bad cal data to the new EEPROM. The re-calibration with the new chip fixed that. You could try a few reads of the old EEPROM to see if you get consistent data from it just to verify the fault. Thank you for another interesting repair video. Regards, David
It could be that the old calibration was too far away from the aging drift of the opamps or passives and it just needed a recal to compensate and re-centre the control range. Would be interesting to see the differences in the old and new EEPROM data to se if there are big changes ... that's assuming you know what each byte in the EEPROM represents! You would have thought a simple error code that says recal needed would have helped!
@@simonbaxter8001 I agree with you that drift in components could cause the error message that was being displayed - the BIT (built in test) was showing that correction value was out of range. The random errors in voltage and current ouput displays (calculated values based on stored calibration data) are an indication of errors reading the EEPROM based on the fact that the actual output was correct.
I was eagerly awaiting the answer to your homework question @17:30, but it didn't come! From my limited experience with this stuff, my guess is that U32 pin15 is always 0v, regardless of the switch setting because:- It's pulled to ground directly by pin1 in one state, but in the other state, I'm not sure. At first I thought it was pulled low by pin2 due to that being connected to U28 pin2, which is the virtual ground of its pin3 (assuming the feedback to pin 2 makes it so). But then U32 pin2 also goes to pin12, which is connected to +10Vref via pin14 in that state. So shouldn't pin15 be +10V? (or is there another 30K on the +10vref that we can't see? - that would give 0v at the junction of the two resistors) Edit: Just realized that the pinned comment was about the virtual ground, so I was on the right track. But then there's still the connection to +10v?
Both the +10v and -10v have series 30k resistors so nothing wrong at all with this part of the circuit. The other 30k is out of shot, but you do see it earlier. Why they put the analog switch on the op-amp side of the 30k resistors rather than the left side is curious though!
Wow!, great repair. Not sure what just happened, :), but great to see you get it squared away. Thanks for sharing the adventure, Rabbit Hole and all! - JRH
Brilliant! What a pain that was lol! I bet it was that flash ROM! I bet it gets updated fairly frequently with usage patterns - eg. user defined settings, as well as calibration data etc.
Great video as usual, I wonder if it would have just need a recalibration from the beginning, but if the eeprom did get corrupted then it was probably good o change it anyway, either way, an excelent learning / troubleshooting video
Hi Ian. @1:28 you said that the output of 7.955V would be "correct", but it was already unusually (for these HPAK power supplies) deviating on the DMM by +1%! For me, that already looked very suspicious.. and further measurements were as well far off. Such a 1% deviation is very improbably due to component drift. Very probably this EEPROM 93C66 from STM encountered a bit-flip. The old technology of these serial EEPROM memory cells were known to be critical, due to impurities in the isolating oxyde, and would later on develop this error, even if the content is not changed so often. As well, data retention time is limited to 10..20 years, although they - nowadays ? - claim 40 years in the datasheet, so it could as well be the 'normal' ageing / discharging mechanism, and would have needed a complete refresh/re-programming upfront. What is the Date Code of the old chip? The replacement of the EEPROM and re-calibration is in any case the best measure.
What I meant by “correct” was it was about as accurate now as it was when it was working before. It’s been in need of cal in a while. The analogue circuits use those coloured 1206 smt resistors same as in some other Agilent PSUs I have repaired and they drifted by 25%……..so this unit in need of call for a while but no point till the resistors are replaced. Good catch on the bit flip, certainly possible…….and notice also the GPIB address changed from 8 to 9 from the start of the video to the end!
@@IanScottJohnston Great catch, as well! This GPIB address change is really an indicator of a defective EEPROM. 28 years ago, at SQM E, I was responsible for fault analysis on these 24Cxx and 93Cxx EEPROMs in our Automotive applications .. such faults were very common at that time. Seems, that there was no proper error detection / error correction implemented in the firmware, at that time, as it's done nowadays inside, or for FLASH memories.
One would think if the EEPROM was going bad that it would throw cal checksum error codes from 740 to 750 just below that 718 in the listing...unless it just drifted to the point where it would use the cal value to set itself with the readback being far enough off from what it expected that it complained. I'm not really sure how that logic is set up. I have one of these that would go Unreg and stop purring some time after starting up. I ended up shotgunning everything in the preregulator circuit except the (now discontinued) phase control IC, and it has been working ever since. It looks like it got bonked once in its prior life, and the preregulator circuit parts are all located around one of the feet on the main heatsink that just kinda hangs there...probably a hairline crack somewhere.
I think the GPIB address jumped from Ch.8 to Ch.9 between the beginning and end of the video.....I don't remember changing it manually, so kinda indicates it could be a bad bit in the EEprom. Saying that, I tested the EEprom offline in my programmer and it appeared to read/write ok.
I suspect the EEPROM has passed its write limits. Maybe every time you power up, it write something to the EEPROM. There is a way to check by erasing the EEPROM and writing 1s and 0s and read them back to see if they get saved.
I've played with the 'faulty' EEProm by writing b00000000, b11111111, b10101010, b01010101 at every location then read it back and look for any changed or swapped bits but so far it's been fine. Need to do more testing by leaving it a few hours then reading back again. I'd love to find something to conclude the repair!
@@IanScottJohnston My old days, we used cold spray and soldering iron to crack down on the faults that appeared to be cause by temperature. Last resort would be to erase the old one and put it back and see if calibration alone would fix it or not. btw, I love your videos.
Hi, have simple repair question if you can help. I have a 12V 500ma input working Electronic Keyboard and when i push in the Power button, there a load pop from the speakers and maybe the main CHIP ect are getting hit hard too?. Its cheap chinese but i like it SO....... i was thinking of like a 240V inrush limiter sort of thing, Is there a way of delaying that 12V just a little when i push the on/off hard switch? I was thinking of a CAP in series that would charge up in a few seconds? or a inductor in series but that is for killing AC so probably wont do much for the DC?
@@IanScottJohnston Its cheap chinese BASE at least from the soldering inside is terrible but it works pretty good. SO.. Maybe they dont have that circuit or its faulty? Do they always have a delay on the speaker circuit when these things are turned on or could this one be faulty and its pooping or from transients? I may not be able to get a schematic assuming the board inside has any numbers, i will have a look, but i was thinking after making the comment, would a computer rated PSU Varistor in series slow the 12 volts from the plug just enough to cure this? They are usually rated at at least 2amps if PSU is 500Watts? and this keyboard is 12V 500ma. So a big green one should be plenty enough and i can test and monitor voltage and heat cranking the volume up and see how it goes it you think it will work or trace down the circuits and maybe a damaged board or poorly designed.
I have been using the PSU for 3-years since I repaired/refurbished (see my old YT video) it and it's been absolutely find till one day I went to switch it on. Consensus is the EEProm was faulty.
It's not, my bad I read the circuit wrong (I'll trim the video)......Yes, it's a current steering circuit, and U32 switch is used to control that current steering. The position of the analogue switches confused me for a while, the outputs of the switches are at virtual ground at U28 op-amp inputs......so were never gonna see any sort of measureable voltage. Not sure why they never put the analogue switches on the left side of R72 & R98 rather than directly at U28 op-amp input. Integrating ADC's always confuse me........
Interesting fault. I have one of these.... it's my LEAST favorite power supply. I just keep it in a closet. Turns out my GW Instek PSM-6003 is easier to use, has a nicer VFD and is actually more accurate. I always found Agilent/Keysight overpriced power supplies to be oddly under-powered. My much older HP units are IMHO better than this line of power supplies and trouble free.
ADC: It's a current steering circuit, and U32 switch is used to control that current steering. The position of the analogue switches confused me for a while, the outputs of the switches are at virtual ground at the U28 op-amp inputs......so we were never gonna see any sort of measureable voltage at the right side of the analogue switches......only virtual 0Vdc.
EEProm: As pointed out by others the 93C66 EEProm can be subject to data loss/corruption through bit flip issues, and did you notice the GPIB address was 8 at the start, then 9 at the end!
I've also played with the 'faulty' EEProm by writing b00000000, b11111111, b10101010, b01010101 at every location then read it back and look for any changed or swapped bits but so far it's been fine. Need to do more testing by leaving it a few hours then reading back again. I'd love to find something to conclude the repair!
I was going to comment exactly the same explanation. Reading 0V on the non-inverting input of that op-amp actually means it’s working properly. If there is anything other than zero there, then that would indicate a problem in that area. Great work, Ian.
@@garyramsey4275what caught me out was not expecting the switches fitted after the series resistors and right at the op-amp inputs! I’d be tempted to reverse them if I was designing the circuit.
@@IanScottJohnstonyes, if those resistors were after the analog switch, it would be much easier to follow the signal. I wonder why they chose to do it this way?
@@garyramsey4275 R72 needs to be where it is, or the -10 VREF is shorted by the switch, but a resistor on the non-grounded output would not hurt for sure.
Congratulations on getting this PSU repaired!
Nice repair, funny how we have tend to have a favourite bit of test kit.
I used to get a bit angry if other engineers messed with my bench setup, letting a bit go off for cal was like losing an arm for a month. 🙂
Woohoo it’s repair time again!
I know!.....it's been a few weeks!
Great work, Ian! My guess is the EEPROM was defective. The strange current and voltage readouts lead me to believe some calibration constants were changing. Copying the contents to a new chip transferred bad cal data to the new EEPROM. The re-calibration with the new chip fixed that. You could try a few reads of the old EEPROM to see if you get consistent data from it just to verify the fault. Thank you for another interesting repair video. Regards, David
Yes, I think so…….and I might put the old EEProm in my programmer and try repeat testing it, I think my software has some routines for this.
It could be that the old calibration was too far away from the aging drift of the opamps or passives and it just needed a recal to compensate and re-centre the control range. Would be interesting to see the differences in the old and new EEPROM data to se if there are big changes ... that's assuming you know what each byte in the EEPROM represents! You would have thought a simple error code that says recal needed would have helped!
@@simonbaxter8001 I agree with you that drift in components could cause the error message that was being displayed - the BIT (built in test) was showing that correction value was out of range. The random errors in voltage and current ouput displays (calculated values based on stored calibration data) are an indication of errors reading the EEPROM based on the fact that the actual output was correct.
I was eagerly awaiting the answer to your homework question @17:30, but it didn't come!
From my limited experience with this stuff, my guess is that U32 pin15 is always 0v, regardless of the switch setting because:-
It's pulled to ground directly by pin1 in one state, but in the other state, I'm not sure.
At first I thought it was pulled low by pin2 due to that being connected to U28 pin2, which is the virtual ground of its pin3 (assuming the feedback to pin 2 makes it so).
But then U32 pin2 also goes to pin12, which is connected to +10Vref via pin14 in that state. So shouldn't pin15 be +10V? (or is there another 30K on the +10vref that we can't see? - that would give 0v at the junction of the two resistors)
Edit: Just realized that the pinned comment was about the virtual ground, so I was on the right track. But then there's still the connection to +10v?
Both the +10v and -10v have series 30k resistors so nothing wrong at all with this part of the circuit. The other 30k is out of shot, but you do see it earlier.
Why they put the analog switch on the op-amp side of the 30k resistors rather than the left side is curious though!
Wow!, great repair. Not sure what just happened, :), but great to see you get it squared away. Thanks for sharing the adventure, Rabbit Hole and all! - JRH
Brilliant! What a pain that was lol! I bet it was that flash ROM! I bet it gets updated fairly frequently with usage patterns - eg. user defined settings, as well as calibration data etc.
Great video as usual, I wonder if it would have just need a recalibration from the beginning, but if the eeprom did get corrupted then it was probably good o change it anyway, either way, an excelent learning / troubleshooting video
Hi Ian. @1:28 you said that the output of 7.955V would be "correct", but it was already unusually (for these HPAK power supplies) deviating on the DMM by +1%!
For me, that already looked very suspicious.. and further measurements were as well far off.
Such a 1% deviation is very improbably due to component drift.
Very probably this EEPROM 93C66 from STM encountered a bit-flip.
The old technology of these serial EEPROM memory cells were known to be critical, due to impurities in the isolating oxyde, and would later on develop this error, even if the content is not changed so often.
As well, data retention time is limited to 10..20 years, although they - nowadays ? - claim 40 years in the datasheet, so it could as well be the 'normal' ageing / discharging mechanism, and would have needed a complete refresh/re-programming upfront.
What is the Date Code of the old chip?
The replacement of the EEPROM and re-calibration is in any case the best measure.
What I meant by “correct” was it was about as accurate now as it was when it was working before. It’s been in need of cal in a while.
The analogue circuits use those coloured 1206 smt resistors same as in some other Agilent PSUs I have repaired and they drifted by 25%……..so this unit in need of call for a while but no point till the resistors are replaced.
Good catch on the bit flip, certainly possible…….and notice also the GPIB address changed from 8 to 9 from the start of the video to the end!
@@IanScottJohnston Great catch, as well! This GPIB address change is really an indicator of a defective EEPROM.
28 years ago, at SQM E, I was responsible for fault analysis on these 24Cxx and 93Cxx EEPROMs in our Automotive applications .. such faults were very common at that time.
Seems, that there was no proper error detection / error correction implemented in the firmware, at that time, as it's done nowadays inside, or for FLASH memories.
That readback could use adjusting a bit.
Yes, it got a recal at the end of the video.
Neat. Well done.
Maybe just a stray cosmic ray zapped your EEPROM - all those lovely solar flares making the Aurora where you live!
Don't forget to change the address back to 8.
I never noticed the change……thanks!
One would think if the EEPROM was going bad that it would throw cal checksum error codes from 740 to 750 just below that 718 in the listing...unless it just drifted to the point where it would use the cal value to set itself with the readback being far enough off from what it expected that it complained. I'm not really sure how that logic is set up.
I have one of these that would go Unreg and stop purring some time after starting up. I ended up shotgunning everything in the preregulator circuit except the (now discontinued) phase control IC, and it has been working ever since. It looks like it got bonked once in its prior life, and the preregulator circuit parts are all located around one of the feet on the main heatsink that just kinda hangs there...probably a hairline crack somewhere.
I think the GPIB address jumped from Ch.8 to Ch.9 between the beginning and end of the video.....I don't remember changing it manually, so kinda indicates it could be a bad bit in the EEprom. Saying that, I tested the EEprom offline in my programmer and it appeared to read/write ok.
I suspect the EEPROM has passed its write limits.
Maybe every time you power up, it write something to the EEPROM.
There is a way to check by erasing the EEPROM and writing 1s and 0s and read them back to see if they get saved.
I've played with the 'faulty' EEProm by writing b00000000, b11111111, b10101010, b01010101 at every location then read it back and look for any changed or swapped bits but so far it's been fine. Need to do more testing by leaving it a few hours then reading back again. I'd love to find something to conclude the repair!
@@IanScottJohnston My old days, we used cold spray and soldering iron to crack down on the faults that appeared to be cause by temperature.
Last resort would be to erase the old one and put it back and see if calibration alone would fix it or not.
btw, I love your videos.
@@piconano I did use freezer spray and a heatgun on the Pcb when I first opened the unit up, but it made no difference in this case.
@@IanScottJohnston That's cuz you're good
The multimeter was reading the voltage at the start, and it was off by ~0.1V.
Did it get better after re-calibration?
After recal the PSU is more accurate, you can see that at the end of the video.
Hi, have simple repair question if you can help. I have a 12V 500ma input working Electronic Keyboard and when i push in the Power button, there a load pop from the speakers and maybe the main CHIP ect are getting hit hard too?. Its cheap chinese but i like it SO....... i was thinking of like a 240V inrush limiter sort of thing, Is there a way of delaying that 12V just a little when i push the on/off hard switch? I was thinking of a CAP in series that would charge up in a few seconds? or a inductor in series but that is for killing AC so probably wont do much for the DC?
Look up amplifier speaker protection/muting relay circuit.
@@IanScottJohnston Its cheap chinese BASE at least from the soldering inside is terrible but it works pretty good. SO.. Maybe they dont have that circuit or its faulty? Do they always have a delay on the speaker circuit when these things are turned on or could this one be faulty and its pooping or from transients? I may not be able to get a schematic assuming the board inside has any numbers, i will have a look, but i was thinking after making the comment, would a computer rated PSU Varistor in series slow the 12 volts from the plug just enough to cure this? They are usually rated at at least 2amps if PSU is 500Watts? and this keyboard is 12V 500ma. So a big green one should be plenty enough and i can test and monitor voltage and heat cranking the volume up and see how it goes it you think it will work or trace down the circuits and maybe a damaged board or poorly designed.
The problem is between the bench and the chair, simple!!!!
Nice !
The list of error codes is named "Calibration errors", maby that's a hint?
A great bit of lateral thinking, could someone have incorrectly calibrated it or been unable to on some of the ranges it needs.
I have been using the PSU for 3-years since I repaired/refurbished (see my old YT video) it and it's been absolutely find till one day I went to switch it on. Consensus is the EEProm was faulty.
Call me a brick but I'm still not sure what you saw on the schematic that told you that was a current reference.
It's not, my bad I read the circuit wrong (I'll trim the video)......Yes, it's a current steering circuit, and U32 switch is used to control that current steering. The position of the analogue switches confused me for a while, the outputs of the switches are at virtual ground at U28 op-amp inputs......so were never gonna see any sort of measureable voltage.
Not sure why they never put the analogue switches on the left side of R72 & R98 rather than directly at U28 op-amp input.
Integrating ADC's always confuse me........
Interesting fault. I have one of these.... it's my LEAST favorite power supply. I just keep it in a closet. Turns out my GW Instek PSM-6003 is easier to use, has a nicer VFD and is actually more accurate. I always found Agilent/Keysight overpriced power supplies to be oddly under-powered. My much older HP units are IMHO better than this line of power supplies and trouble free.