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

[ChrisG]

Hi Prasimix, I've tested with 1.4V offset and 2.4Vpp. Following is a screen shot on the output signal when the output of H24005 is under load: 10 Ohm at 4A.

[prasimix]

Ok, so you can see that top is cut off, that seems ok to me since you reached top voltage and thanks to current limitation (4 A for 10R) OVP is not tripped on. Remaining question is "hole" on rising edge. Does it there even with lighter load, or no load, too?

[ChrisG]

Under no load everthing is perfect. 40Vpp, solid sinewave until 60Hz or so. At higher frequencies 100Hz I have the same result as you have with the bottom triangle. Though between 20 - 60Hz the falling edge of the sine-wave is a bit jerky but when fiddling with the BNC from ARB to H24005 it goes away. Guess the BNC with the wires to the H24005 is not of the best quality perhaps. Although that problem disappears with the e.g. 100Hz frequency.

[prasimix]

Ok, that "hole" also make sense to some extent with huge load, it's possible that pre-regulator cannot deliver enough power on time. Possibly with bigger Cout on its output situation could be optimal.

[morris6]

After reading all these pages I felt kind of challenged by the idea of building my own H24005. The author has done a lot of work to make this a great project. All the information is available on the EEZ website and on GitHub. Real encouragement can be found in the sections dedicated to DIY building. A complete Bill of Materials is published (GitHub) with references to part suppliers. Well, I decided to give it a go, ordered some parts (..), and a fitting enclosure.
 
Soldering all the parts kept me off the streets for quite a few evenings but was not too difficult.

The enclosure I used has a bit more height inside than the one specially designed for the group buy version. Therefore, I could use an eighty-millimeter fan for cooling, bigger is better isn’t it…? At least less noisy. More about it later.

Most hiccups were taken care of during assembly and after running the calibration procedure all seems well. However, two out of three startups give a “beep” -Fan failed- fault.
 
It looks like my choice for the bigger fan causes the startup test to fail: the fan accelerates slower obviously than the sixty-millimeter original.

The firmware has a parameter “FAN_NOMINAL_RPM”, default 4000 rpm. I ‘re’ defined this in the conf_user.h header file. My fan does about 2000 rpm. But that did not solve the startup test issue. Some searching reveals that the timing limits are hardcoded in the software: fan.cpp file. For testing this I doubled the hard-coded timing values and “tadah”, it passed the startup test all the time.

So far, so good. But the project deserves a better solution, I think.

@Prasimix:

Will it be possible to change the hard-coded timing limits to something that is dependent upon nominal fan rpm? And don’t you agree that this makes the firmware more flexible for builds with other cooling fans than the default one? Can this be integrated in the next firmware version?

Maybe a remark is needed then in the building instructions to make builders aware of the possibility to use adapted fan timings.

I did an attempt to write an adaptation in the firmware, the fan.cpp file, but I feel hardly qualified for this. I am a typical -following-the-examples- programmer. Nevertheless, the adaptation is working for my build.

The modified fan.cpp file is attached here, renamed to fan_cpp.txt for forum posting rules reason. I also included my conf_user.h header file for clarity.

[prasimix]

Hi @morris6, this looks amazing!

Fan issue shouldn't be a problem to fix at all. Just please be so kind and open a new issue in GitHub firmware, describe what you want and we can add it in v1.1 release.

Another possible issue with your fan is its position. It seems to me that power boards are not evenly positioned regarding fan center and that one channel will be better cooled then another. Anyway, even if that is a case firmware monitor temp sensors on both channels (together with "ambient" temperature) and should keep fan working depending of highest measured temperature.

[morris6]

Ok, thanks Prasimix. I will open a new issue for this.

In my build the fan exhausts warm air. Cooling air is entering through slits in the side panels of the casing. But you are right, cooling will not be equal. Just hoping it will proof to be sufficient.

[prasimix]

Hello all EEZ H24005 owners. We just finished adding PID for fan control, so anyone interested in testing it feel free to get the latest version from master branch and check #188. Many thanks in advance for your feedback.

[morris6]

@Prasimix

In this new version of the firmware the startup test failure fan issue with my bigger and slower fan is solved. Thanks for that.

@All

When using a different fan in your build than the 'BOM' one you can make the following adaptation in the conf_user.h file when compiling your firmware:

Code: [Select]

// Nominal fan RPM (for PWM=255)
#undef FAN_NOMINAL_RPM
#define FAN_NOMINAL_RPM   2000



[morris6]

Hi @saltech,

To answer your PM about the enclosure I used for my build, you are not the first asking me about it, so I will answer your questions here.

Enclosure is a Bopla Botego BO 62618. For sale at Conrad, ordering # 520455 (-only- 79 Euro  , but couldn't find them elsewhere). The material is ABS, I used a 3 mm aluminum plate as a sturdy chassis.

The heat sinks are from TME, ordering # RAD-A52317/70. They are 70 mm x 150 mm x 25 mm, and I shortened them by 9 mm to make them 141 mm for a better fit. The 70 mm width is just enough, 75 mm would be better for the mounting hole pattern. Quality wise there are better brands but these do the job.

Some pictures:

FrontInside picture shows how the heat sinks are mounted. Also you can see the openings in the side panel for cooling air entry. Left and right of the orange power relays the longer pin headers I choose to use; these give more distance between power boards and Arduino shield and add some mounting flexibility. 

The silicon pad between pcb and heat sink I got was a little bit thinner than the advertised 5 mm. To get some pressure on the pad I reduced the distance between pcb and heat sink. At the four 5 mm spacers a 0.4 mm recess is milled into the heat sink. It also needed a 1 mm pocket for the main power fet Q4, for better fitting. See heatsink picture.

Concerning the front panel, yes, always difficult. I now have some label printer strips on the front.

I enjoyed assembling this well designed DIY power supply, I hope you will too.

[prasimix]

I can see at least two nice details on your unit that I don't have: placing of SD Card that it can be easily removed without disassembling front panel and r5B13 (or r5B13a) power board that has three current shunt resistors in parallel instead of just one. I'd like to know how you manage to expose SD Card like that and did you measured current offset on hi-range (something that was reported in #661)? Does three resistors improve this issue to any extent?

[morris6]

Well, in my build it also is not easy to change the SD card, you would need to disassemble the whole front panel. That is why I put an SD card socket, a push push type, on a piece of board and placed this on a bracket beneath the Arduino shield. It is (wire wrap) wired direct to the signals on the LCD1 connector. Gnd and +3v3 from C140 and C105, shield wire to Earth connector.

Another detail, originally the reset switch hides partly behind the display. I placed the switch in the empty space of SP2 on the r5B13, that's the wire you see running to IC31. You can reach it trough a hole in the front panel or easily when the top of the case is removed.

To answer your question about current offset in hi-range: during calibration I observed no problem in this range. But the measuring time was not long because I was overheating my load resistor(s). I will try to test this issue again after assembling a better load, with cooling or so.

At what current does the change over from mid to high occur? The lowest range is not implemented yet in the software is it?

[prasimix]

At what current does the change over from mid to high occur? The lowest range is not implemented yet in the software is it?

Well, with r5B13 you also got three instead of two current ranges (i.e. 0-50 mA, 500 mA and 5 A). The lowest one is not supported in firmware since I don't have such Power board yet . We can try to add it blindly and send to you updated firmware to test it. In the mean time you can safely use middle and high range if Autorange is selected under Channel's advance settings ... Current ranges:



If I can recommend a huge, solid and cheap power resistor, take a look at this one from SR Passives:

[fuzzoli]

Does anyone in the US want/need the VTX-214-005-0512 power module for the Aux PS?  The only places that have it are Sierra IC ($100 min. order), and Newark.  However, the Newark stock is in the UK, with a $20 freight charge, so I was looking to see if anyone here in the States wants to share the wealth pain as $20 to ship a $14 part is a bit much. 

Thanks,

-Frank

[prasimix]

Phew, I understand your frustration and due to that I've tested another solution that probably can be placed in the same area on AUX PS board but based on #861. Also It's highly possibly that someone else (from China where VTX possibly also manufacture their adapters) has a pin-compatible replacement but with reasonable price. In a time when I chose that part it was about 5 GBP (unit price!) on Farnell and now it's still insanely high (~16 GBP).

[morris6]

About the issue in post #661:

I think it's a good idea to give the psu some time to stabilize during calibration of the high current range. After doing so the ADC indication is pretty much in line with the measuring DMM, within 1 - 2 mA near the 5 A. limit.

Measurements were done with a Keysight 34461A. This DMM has enough resolution to see (parts of) milli amps in the 10 A. range.

I observed, with the r5B13 boards that have the three parallel 0.03 ohm resistors for R65, R65A and R65B, a difference of about 15 mA between initial (cold) and stabilized at 5 Amps indication in CC mode. So about 0.3 % offset (?). The resistors used are Bourns CRA series 2512 size and spec'd +/- 50 ppm/C, Farnell 2321759. Screen picture 'current.png', at about 1' 30'' cooling fan started.
Compared to the r5B12 board only slightly better.

Hmm.. "Advanced settings > Current ranges > Currently selected" is not visible directly after switching the psu to on. It appeared only after some fumbling with the settings  ...  ?

[prasimix]

Thanks for that. It seems that without implementing a sort of active temperature compensation the last resort is to put some physically (and financially) huge current sense resistor like e.g. Isabellenhütte PBV-R010-F1-1.0. And yes, before that someone should try CSSH2728FT20L0 (TCR 15 ppm/°C) mentioned in #681.

[morris6]

The Isabellenhütte one is 30 ppm/K so has to be kept cool. The other suggestion looks better at 15 ppm/K. Improving the temperature induced offset comes at a cost, for me it's good enough for now.

However:

During trials I stumbled over the following: With the current range selected to 0.5 A I could not reach the full 500 mA current in the load resistor when increasing voltage. At about 460 mA for power board 1 and 490 mA for power board 2 the CC mode became active. Both channels were calibrated and external measurements were equal to the indication on the psu.
 
First I suspected the sense resistor for the 500 mA range being to high, R63: 0.082 ohm 1%, but that is not the case. The problem is in the pcb trace that connects R63 with the R65's. This trace carries the 'full' current in the 500 mA range and is part of the sense 'stack'. Although not long, about 9 mm before it divides and connects to the R65's, the trace is rather thin, 16 mill (from eagle), so 0.41 mm. But its resistance value must be below 0.008 ohm to arrive at the wanted total of 0.100 for current measuring in this range. With a resistance calculator (https://www.allaboutcircuits.com/tools/trace-resistance-calculator) I found 0.0108 ohms for this piece of trace. Combined with the tolerance of R63 this is too much. See picture Layout.png

The solution was a simple piece of wire from R63 to R65, picture R65R65C.JPG.

The layout below the three R65's is really nice I think however a bit critical. There is space for a wider trace, certainly when the fork is omitted.

Another reason to minimize the resistance of this trace is the resistance temperature coefficient of copper,  about 4000 ppm/K.
 
I had another issue with the layout during soldering the R65's: Several times the thin measure point traces did not even connect to R65's when I hand-soldered them.    Makes the problem above even worse.

So I would like to suggest to change the layout here. A wider trace from R63 to the inside of R65A, and maybe connect the measuring point traces direct to the islands of R65A. Like picture Suggest.png

For the low current range (50 mA) R179 I used a 0.82 ohm resistor, not tested since this range is not yet active in the firmware. But maybe soon...?

[prasimix]

The Isabellenhütte one is 30 ppm/K so has to be kept cool. The other suggestion looks better at 15 ppm/K. Improving the temperature induced offset comes at a cost, for me it's good enough for now.

A good thing about Isabellenhütte is that comes in huge package similar to TO-247 that can be easily mount to the additional heatsink. But still, it cost a small fortune.

However:

During trials I stumbled over the following: With the current range selected to 0.5 A I could not reach the full 500 mA current in the load resistor when increasing voltage. At about 460 mA for power board 1 and 490 mA for power board 2 the CC mode became active. Both channels were calibrated and external measurements were equal to the indication on the psu.
 
First I suspected the sense resistor for the 500 mA range being to high, R63: 0.082 ohm 1%, but that is not the case. The problem is in the pcb trace that connects R63 with the R65's. This trace carries the 'full' current in the 500 mA range and is part of the sense 'stack'. Although not long, about 9 mm before it divides and connects to the R65's, the trace is rather thin, 16 mill (from eagle), so 0.41 mm. But its resistance value must be below 0.008 ohm to arrive at the wanted total of 0.100 for current measuring in this range. With a resistance calculator (https://www.allaboutcircuits.com/tools/trace-resistance-calculator) I found 0.0108 ohms for this piece of trace. Combined with the tolerance of R63 this is too much. See picture Layout.png

The solution was a simple piece of wire from R63 to R65, picture R65R65C.JPG.

The layout below the three R65's is really nice I think however a bit critical. There is space for a wider trace, certainly when the fork is omitted.

Another reason to minimize the resistance of this trace is the resistance temperature coefficient of copper,  about 4000 ppm/K.
 
I had another issue with the layout during soldering the R65's: Several times the thin measure point traces did not even connect to R65's when I hand-soldered them.    Makes the problem above even worse.

So I would like to suggest to change the layout here. A wider trace from R63 to the inside of R65A, and maybe connect the measuring point traces direct to the islands of R65A. Like picture Suggest.png

No, something else is probably your problem here since I can easily go over 500 mA in low (in your case middle) range (current range set to Best/Default, and with Autoranging set to on).  Did you try to replace R179 with 0R for test before you put that extra wire?

For the low current range (50 mA) R179 I used a 0.82 ohm resistor, not tested since this range is not yet active in the firmware. But maybe soon...?

Hopefully yes, maybe not as a "official" release (i.e. the v1.2, since we are finalizing now the v1.1) but included in the one of the future version in the GitHub master branch.

[prasimix]

Reply to @RedSpanner question, since it can be useful for other builders, too:

The firmware (finally) loaded correctly, but first power up, I got error messages
on Channel 1:
Self-test is not passed
- CH1 IOEXP failed
- CH1 ADC failed

If one or more SPI peripherals is complaining during self-test then one or more of the following things could be the problem:

• 48 Vdc power is not present on that module (check if you have ~48 V on the module input)
• Bias voltages on the Power board is missing. The main suspect here is wrong soldering of LTC3260 thermal pad. 4. Digital isolators for Ch1 (i.e. IC19, IC20) are e.g. not soldered properly
• You cannot check that with visual inspection, but you can check if +/-5 V comes out of it. Additionally check output of LP2591 for powering digital section.
• An issue with connection between Arduino shield (X13) and Power module (x2)

You can get some additional info over the serial console when powering it up (or reseting). Additionally, you can also send DEBUG? command to get some info.
 

[morris6]

No, something else is probably your problem here since I can easily go over 500 mA in low (in your case middle) range (current range set to Best/Default, and with Autoranging set to on).  Did you try to replace R179 with 0R for test before you put that extra wire?

Sorry to bring it up once more, other builders of r5B13 boards may also run into this issue.

When working with the psu in the 500 mA range it seems not possible to reach the full 500 mA with the r5B13 power board as it is. Somewhere between 460 and 490 mA the CC mode takes over, although the set current limit is 500 mA. Seen this with all four power boards I have tested.

Comparing the r5B13 with the r5B12 power board layout the trace between R63 and R65 is much shorter, see picture r5B12.JPG. Also the shunt resistors are twice the value, halving the relative influence of trace resistance.

I soldered R63 and R65, -A and -B on a spare power board r5B13, picture measureC.JPG. While under a load of 500 mA, in series with an external 4 ohm load resistor, I measured these voltages:

left pad of R63, supply from load resistor   : 52 mV
other of pad of R63, trace to R65         : 11 mV
left side of R65, -A and -B            :  5 mV
right side of R65, -A and -B ,common       :  0 mV

I also measured these voltages on a 'life' power board in the psu while in 500 mA range and under almost 500 mA load. They are about the same.

So there is a 6 mV drop over the trace from R63 to R65, apparently a trace resistance of 0.012 ohm here! Total shunt for 500 mA sums up to 0.104 ohm in stead of 0.092. Compare this to a power board r5B12: R65 is 0.02 and R63 is 0.16 ohm. The trace between adds about 0.006 ohm, being less than half length. That is 3% of the total. The 0.012 ohm trace of the 5B13 board adds 13% to the total shunt resistance for the 500 mA range. So the voltage measured is 13% higher than intended... You can calibrate this away but the set value voltage from the DAC is limited, and so the current can not reach the intended 500 mA in CC mode.

Another thing: 13% of the 500mA shunt is the copper trace with 4000ppm/K Temperature Coefficient, significantly worsening the temperature stability of the 500 mA range.

@ Channel failure after assembly:

With one of my boards I had IC11 swapped, pin 1 to pad 4, that caused -5V rail failure.

[chenguoping]

Hi prasimix,
about the serial console, there has three questions, 1) what port should be used, native or programming? 2) does the arduino due should be modify, that means don't powered by usb cable. 3) which baud ratio should be set?

thanks alot

[prasimix]

Sorry to bring it up once more, other builders of r5B13 boards may also run into this issue.

When working with the psu in the 500 mA range it seems not possible to reach the full 500 mA with the r5B13 power board as it is. Somewhere between 460 and 490 mA the CC mode takes over, although the set current limit is 500 mA. Seen this with all four power boards I have tested.

Comparing the r5B13 with the r5B12 power board layout the trace between R63 and R65 is much shorter, see picture r5B12.JPG. Also the shunt resistors are twice the value, halving the relative influence of trace resistance.

I soldered R63 and R65, -A and -B on a spare power board r5B13, picture measureC.JPG. While under a load of 500 mA, in series with an external 4 ohm load resistor, I measured these voltages:

left pad of R63, supply from load resistor   : 52 mV
other of pad of R63, trace to R65         : 11 mV
left side of R65, -A and -B            :  5 mV
right side of R65, -A and -B ,common       :  0 mV

I also measured these voltages on a 'life' power board in the psu while in 500 mA range and under almost 500 mA load. They are about the same.

So there is a 6 mV drop over the trace from R63 to R65, apparently a trace resistance of 0.012 ohm here! Total shunt for 500 mA sums up to 0.104 ohm in stead of 0.092. Compare this to a power board r5B12: R65 is 0.02 and R63 is 0.16 ohm. The trace between adds about 0.006 ohm, being less than half length. That is 3% of the total. The 0.012 ohm trace of the 5B13 board adds 13% to the total shunt resistance for the 500 mA range. So the voltage measured is 13% higher than intended... You can calibrate this away but the set value voltage from the DAC is limited, and so the current can not reach the intended 500 mA in CC mode.

Another thing: 13% of the 500mA shunt is the copper trace with 4000ppm/K Temperature Coefficient, significantly worsening the temperature stability of the 500 mA range.

Thanks for more input. I also have 10 mOhm as R65 (with accompanying different R63, R60, R70). I still think that bigger error is announced with R179 then with thiner trace between middle and hi range resistors. Once again, did you try to replace R179 with 0R?

[prasimix]

Hi prasimix,
about the serial console, there has three questions, 1) what port should be used, native or programming? 2) does the arduino due should be modify, that means don't powered by usb cable. 3) which baud ratio should be set?

thanks alot

1) Starting from firmware v1.02 native port is selected as default, but with change in firmware configuration you can use programming port, too.
2) Arduino due should not be modified, since +5V that comes from USB is removed on the connector that comes from AUX PS board to Arduino Shield.
3) I'm regularly using 115200 bauds for all type of interaction (e.g. even when LISTs are executing on both channel). If should go up to 2Mbauds as offered on Arduino, I've tested it but not in all scenarios.

[morris6]

Thanks for more input. I also have 10 mOhm as R65 (with accompanying different R63, R60, R70). I still think that bigger error is announced with R179 then with thiner trace between middle and hi range resistors. Once again, did you try to replace R179 with 0R?

OK. I tried a power board with R179 shorted, and with only the original trace as connection between R63 and R65. After recalibration maximum current in 0.5A range is 494 mA. When I remove the short over R179 the results are the same, 494 mA.

In r5B13 board schematic R179 is at the bottom end of the divider R60-R62, compare with R70-R71. In 0.5A range current passes via Q13, R63 and R65,-A and -B. Only in 50 mA range current flows via R179 etc., with Q34 switched on. Well.. that's how I read it.

Using a single 10 mOhm resistor on the layout of r5B12 is less critical because the trace R63-R65 is about half the length compared with r5B13 layout where there are three parallel resistors of 30 mOhm. See my earlier post, trace resistance about 6 mOhm vs 12 mOhm (measured).

R179 can have some influence on the 'common mode rejection ratio' of opamp IC7A though.