DIY SCPI programmable dual channel bench PSU 0-50V/3A (now EEZ H24005)

[Romain]

214 Ohm resistor mounted on the 12V rail across C77:
Standby switch off: 9.72V measured
Standby switch on: 12.77V measured

See attached reading during the reboot event.

[prasimix]

Ok, stand-by switch (front panel) did not affect powering of TR1. But you can measure different voltages and on top of that it seems that 12 V is in one moment shorted! Did you try to power-up PSU without fan connected? Maybe it is shorted.

[Romain]

Good call!
D11 was shorted. I replaced it with a 1N4004 (I didn't have an SMD part at hand).
The PSU now boots, the fan spins and it passes the fan test. 
It's just showing the "SD card failed" message. Why is that?

[prasimix]

It's just showing the "SD card failed" message. Why is that?

It's quite possible that SD card is not plugged properly any more. It's could happen during shipment. That SD card slot is on the rear side of TFT display, and it seems as cheap as display itself . Unfortunately you have to detach Arduino shield from front panel. Make sure to put everything back after disassembling. Consult building instructions and particularly Fig. 22, 23 and 24. As a precaution step, you can try to check if SD card is working on your PC with SD card reader.

Thanks for another video. If you are not going to use Ethernet, simply disable it to avoid waiting for connection timeout. You can activate it anytime from system settings page.

[phuebner]

Hi guys,

I have also put together a PSU based on the r5B12 revision documents (i.e. similar to the crowd supply version) and I am quite happy with it. First of all great job on developing and documenting the PSU so well !!!

One thing however which I encountered is that there is a considerable drift of up to +100mA from the set voltage after power on, which may take up to an hour of operation to stabilise. The drift is clearly temperature related as it stops drifting once the PCBs and ICs have heated up to a stable operation temperature. In my test I set a voltage of 3.5V and passed a current of 7mA, which caused a 100mA drift over the time of approx 45min to 1hour. I suspcet if I would pass higer currents the PCB would heat up faster and the drift would stop earlier. 

I tracked down the source of this drift to the OpAmps that were used for the Voltage feedback loop and current feedback loop (i.e. IC5 and IC6). Their output gain seems to be quite sensitive to temperature, and since they are mounted in close proximity to the bias supply circuit they heat up considerably. If I place a punctual cooling source onto the package of IC5 i see an immediate drop in voltage by 100-200mA, back down to the original voltage before the drift started. Cooling of other ICs on the power board causes no change in output voltage. This led me to belive these ICs are the main source of the problem.

So maybe we should considder to:
a) choose a zero-drift OpAmp model (e.g. http://www.linear.com/products/zero-drift_amplifiers). Idealy a drop-in replacement for the one that is used right now. But this may be hard to find
b) improve the thermal management on the PCB
c) implementing a feedback loop in firmware to counteract the drift

Has anyone else experienced this on his PSU? It would be great if someone could confirm this problem on his PSU, so I can be sure I did not do something wrong when assembling mine. Also I would like to get your input about how we should try to improve this. I? happy to take a shot at option c) but I would need some help with selecting a OpAmp replacement to try out option a). Option b) is probably not an option right now for those of us who have already a PSU.

I'm looking forward to your input.
Patrick

[prasimix]

Hi Patrick, that sounds pretty weird, but lets see what can be done. First, please let me know what mode did you test: CV or CC? If you are in CC mode, then at least part of the control loop has zero-drift op amp if you've used LTC2050HV (IC7). Drift of 100 mA on 7 mA is insane, but maybe I somehow missed to record it all this time.
Please let me know what is your output voltage, current and connected load that I can try to setup similar environment and see what is going on. Certain current drift exists when you are in CC mode due to overheating of Rsense which TCR is huge (75 ppm/oC) and you can see up to 0.6 % drift, but that is still only about 20-30 mA on the 5 A and that is reported as issue #40.

[phuebner]

Hi,

I just realised I wrote mA. Well that does not make sense at all  . I am in CV mode with set voltage of 3.5V and I have a drift of >100mV to 3.65V. To me it seems as if the drift is caused by IC5, which handles the voltage loop. If I put some ice on the IC the voltage drops right back down to 3.5V (so the set voltage).

My test setup:

• Recently powered up PSU (so all PCBs are initially at room temperature)
• U_Set: 3.5V, I_Set: 20mA
• Load resistor of 470Ohm resulting in 7mA current draw and CV mode
• Multimeter connected to output terminals to measure output voltage[\li]
  

I hope that makes more sense 

[prasimix]

I've made a first test now (PSU was turned off over night) and during initial one hour I can measure positive drift of not more then 3.7 mV with mentioned 3.5 V and 470R as load. I'll repeat it a couple of more time over a day.
That is far better then what you reported, but also maybe my construction is a little bit different! Please note thermal pad that is placed underneath Power board PCB (see Fig 14.). Nearby TL783 (IC3) is considerable source of heat (about 1.5 W) but that has a "bad" and "good" side: the good side is that you can count with more predictable PCB temperature over wider ambient temperature range. Therefore if we wait that temperature is stabilize and perform calibration then we can expect more predictable output voltage over the time (of course with connected load you have to use remote sensing if better precision is needed).

[phuebner]

I do not have the thermal pad yet that you've mentioned, as I found it hard to source. Maybe that is why my board heats up more than yours. But I wouldn't have thought that it has such a dramatic effect? 3.7mA sounds very good. Over what time frame did you record it? Mine keeps climbing very slowly but steadily over the course of ~1h.

[prasimix]

I've tested it for an hour.

5mm thick silicone thermal pad can be found online, usually as 100 x 100 mm sheet (I've used 25 x 70 mm only per board) e.g. at eBay or AliExpress.

[Helix70]

i am running this test too. I will post my results. I am running both channels, each at 3.5V with 20mA and a 470R resistor giving CV at 7-8mA.

[Helix70]

Ok, after 2 hours, here are my results:

Channel 1:
8mA
Start: 3.5018 VDC
End: 3.5036 VDC (Fluke 287)
1.8mV change

Channel 2:
8mA
Start: 3.500 VDC
End:  3.503 VDC   (Fluke 87V)
3mv Change

[phuebner]

Hm, curious. So maybe not having the 5mm thermal pad underneth the pre-reg/ bias circuit is having such a big effect? I have ordered one from AliExpress now. Can only take a month or so until it is delivered and i can test that ;-) Thanks @prasimix for the link.

In the meantime I will do some more testing on my boards to see if there is anything else that is causing this. I will report back. Thanks for trying this out on your devices

[prasimix]

Two more measurements: 3.5 V and full-scale (40 V):

[phuebner]

Ok I also did another set of measurements. I found some thin old thermal pads lying around at home and I stacked them up and placed them where the 5mm thermal pad would usually be. I also enabled the fan a bit earlier than usual. Using this configuration I was able to get a drift of only 5mV over 1 hour !!

Using the thermal pad and fan the PCB stays considderably cooler. Before I was measuring up to 45 degC on the PCB. With thermalpad and fan I reach a maximum temperature on the PCB of 26 degC. A dramatic improvement.

So to me this confirms that this is a thermal effect which can be reduced using the thermal pad. However, it may still be a potential candidate for further improvement in a future version, don't you think?

Anyway, I'm quite happy with this right now. As I said, the good thermal pads are ordered and on the way.

[prasimix]

However, it may still be a potential candidate for further improvement in a future version, don't you think?

Absolutely, everyone are invited with suggestions. Also we should remember that heating up of up to an hour is recommended practice before calibration that you can find in many high class commercial devices manuals (Keysight, etc.).

[prasimix]

BTW, I just noticed that my unit collect about 130 working days, or translated in hours its more the 3000



I wasn't been so active working in last few months but almost all of the time the unit has been powered on and subjected to various types of tortures . On of the most challenging so far was driving a high voltage converter that generate massive EMI that, for instance, my mobile phone was almost damaged on 1.5 meter distance, and TFT display controller painted in white a whole screen few times during nearby discharging sparks. Apart from that it also saved me a day a big time few times thanks to various protections that was implemented some of them is related to my foolishness, therefore I can say that foolproof testing is passed. So far, so good.

[Dwaine]

I ran a similar test while charging some batteries.  Had about 430+ma @ 3.0volts.  It stay pretty consistent.

[phuebner]

Thanks Dwaine for also testing this. It seems that this thick slap of thermal pad underneath the pre-regulator and bias supply is really important... I wouldn't have thought so before, but it became quite apparent now.

[Helix70]

Hi Guys,

Since upgrading 2 units to Firmware v1.1, I get "SD Card Failed" on boot. I have reseated the card, still get the error. Anyone else seeing this?

Josh

[prasimix]

Please note the Firmware v1.1 is not released yet! If you are using the latest build from master branch, you have to install also additional SdFat library used by new set of SCPI commands that works with SD Card (MMEMory subsystem).

[Helix70]

Ahhh. That would do it. Serves me right for using git. That answers the other question about why the web site says 1.02. 

[Helix70]

I tried to compile 1.1 again, i already had the SD-Fat library from your github installed, and initially it works, then after being in standby from the switch overnight, i get the error for the SD card again when i turn on. The SD Card is fine, i disassembled and checked it.

FWIW, I am using Atom and PlatformIO to compile. I have now selected the tag 1.02 and am trying that, (learning about GIT as I go!) initially it works. I will try an extended standby cycle tonight, but there may be an issue with 1.1's SD card handling as the repo stands right now. I know it isn't released yet, just letting you know what I found.

[prasimix]

Thanks a lot for info. Perhaps to be sure that SdFat lib is correct try to execute some of new added SCPI commands, for example MMEM:CAT? (use e.g. Arduino IDE Serial monitor, or if you have Ethernet, use telnet session initiated with telnet <psu_ip_address> 5025).

[Helix70]

*IDN?
Envox,EEZ H24005 (Due),CS17054,v1.1

MMEM:CAT?
"System Volume Information,FOLD,0"

MMEM:INFO?
12288,7974543360

*TST?
1

DIAG:INFO:TEST?
"1, EEPROM, installed, passed","0, SD card, installed, failed","1, Ethernet, installed, passed","1, RTC, installed, passed","1, DateTime, installed, passed","2, BP option, installed, skipped","1, Fan, installed, passed","1, AUX temp, installed, passed","1, CH1 temp, installed, passed","1, CH2 temp, installed, passed","1, CH1 IOEXP, installed, passed","1, CH1 DAC, installed, passed","1, CH1 ADC, installed, passed","1, CH2 IOEXP, installed, passed","1, CH2 DAC, installed, passed","1, CH2 ADC, installed, passed"


Still had SD Card Fail on startup, with a single high pitch beep.