EEZ Bench Box 3 (BB3) and beyond...

[Kleinstein]

The temperature of the isolator would change quite a bit less than the temperature of the shunt. So I don't think there is much hope for compensation.

One point could be that the TC of different shunt can be different. The 20 ppm/K number is just an upper limit. The actual TC is often lower, but should no be much higher. So it could be just good and bad luck.

400 ppm to 0.12% is only a factor 3. The fuse resistance would not directly comtribute, but the fuse would also get het. So for the complete board the remperature rise may be more than 20 K, though still likely similar for the 2 shunts.

The layout around the shunt may be the problem and effectively add a little copper to the effective shunt. The next higher current shunt should be more connected to the voltage sense side of the shunt. As far as one can see it from the picture it down not look so good.

[prasimix]

Ok, as another exercise I tried to disconnect low/mid range protection from Hi range current path, like shown below (in fact, only the diodes needed to be disconnected). No more drifting! AIN4 shows good results just like AIN2.

[exe]

One possible explanation is that the TCRs of Rsense and Isolators are of different sign and cancel each other out?

I'd be surprised, but this is something that we can measure . But you already did that, the resistor has much less tempco than 0.12% that we see, right? So, it's not a resistor then. You can also swap resistors to see if this makes a difference.

As of isolator tempco, you can gently heat it with a heatgun and see if it changes anything. Although, the thing is, heatgun is not an accurate test because it also heats surround components, the air temperature is not precisely controlled, etc. Also, fast heating may create additional mechanical stress which is not present when the board naturally heats up. However, still a good test, imho .

However, if there is leakage due to flux, this one manifests quite fast when the board is heated (imho). So, when you blow hot air on the board, does the measurement drift? If so, then we can heat parts of the board separately to locate the place more precisely. My best guess is still flux leakage, but may be it's a half-broken diode or something else in signal path.

[Kleinstein]

Ok, as another exercise I tried to disconnect low/mid range protection from Hi range current path, like shown below (in fact, only the diodes needed to be disconnected). No more drifting! AIN4 shows good results just like AIN2.

Good to see the drift gone, but normally the protection should not have that much effect. It should be very low leakage.

The MCP6001 come in 3 different pin-out versions.  The OP is not absolutely needed - chances are it could be good enough also without it. However the diodes are kind off needed to protect the mid and low current shunt.

[prasimix]

Ok, as another exercise I tried to disconnect low/mid range protection from Hi range current path, like shown below (in fact, only the diodes needed to be disconnected). No more drifting! AIN4 shows good results just like AIN2.

Good to see the drift gone, but normally the protection should not have that much effect. It should be very low leakage.

The MCP6001 come in 3 different pin-out versions.  The OP is not absolutely needed - chances are it could be good enough also without it. However the diodes are kind off needed to protect the mid and low current shunt.

Perhaps I could try to move bottom end of diodes to the upper leg of Hi current Rsense, and remove IC4, R33, R34 completely:

[prasimix]

In the meantime Martin added basic power and efficiency measurements. Here is the first test with DC-DC converter from eBay. So we have DC sourcing and measurement in the same chassis. We really now need el. load as the third module to have complete solution (at least for low power devices where e.g. 20-30 W of power has to be dissipated).

[Kleinstein]

Removing  / not polulating IC4 / R33 / R34 could be an option. The diodes should still not cause trouble. If they would cause trouble with the highest current they would cause even more problems with the lowest current range. The diodes should still be relatively high resistance compared to 33 Ohms.

So one should still find out why the diodes cause trouble. Is there a significan common mode voltage at the current measuring channel. That is quite some voltage at the input ground (D4,D5,R31) ?

[prasimix]

So one should still find out why the diodes cause trouble. Is there a significan common mode voltage at the current measuring channel. That is quite some voltage at the input ground (D4,D5,R31) ?

The JP6 is in place. Therefore the voltage is zero. Without it I can measure with DMM about 4 mVrms (AC), or 0.1 mV (DC coupled).

[wizard69]

Your electronic load might be better off as a separate product.   That to keep the heat out of the BB3 chassis and to allow for a higher power handling capability.   All you would really need is a faceless box connected to BB3 via USB or a serial port.   Keep it low cost by limiting power capability to 100 watts or less.   Plus you save the third port for more interesting or complex modules.

In the meantime Martin added basic power and efficiency measurements. Here is the first test with DC-DC converter from eBay. So we have DC sourcing and measurement in the same chassis. We really now need el. load as the third module to have complete solution (at least for low power devices where e.g. 20-30 W of power has to be dissipated).

[wizard69]

i was wondering when the lack of slots would come up.   I'm not sure how limited you are on addressable slots but keeping the same form factor for the expansion unit should be high on the required features list.   This mainly to allow for easy stacking.   I'd prefer off the shelf power supplies myself and would be OK with six slots if that is required to keep the box depth similar to the control unit.    I'd probably would lean towards Ethernet for communications.   USB could be better saved for user equipment, a printer or even an electronic load solution.   

One idea is to make an expansion chassis that could accommodate up to 7 non-power modules in the same form factor as the BB3 (i.e. those that do not need a Mean Well 48 Vdc converter). Communication with BB3 or PC could be via USB or Ethernet.
However, there are a lot of things to define: whether the AUX-PS and MCU modules will remain separate, or perhaps merge them into one? Should AUX-PS/MCU will be connected behind the new 7-slot backplane or should it be a module that goes into a dedicated (8th slot), will the FPGA be added from the beginning or will we just stay on STM32, etc.

[jbb]

Could it be thermal EMF? The resistor pad - PCB solder joints make two thermocouples. If both ends are at the same temperature they  cancel out. If one end is colder than the other you get a thermal gradient and a DC offset voltage. This will depend on layout, nearby components and orientation.

6mA indicates current on 10mOhm shunt is 60uV, which isn’t totally unreasonable.

Suggested test: run 5A * 15 min soak test, then disconnect the load and check the current measurement.

[Mike G]

Firstly, my BB3 arrived Thursday and went together very well, all holes lined up and every nut, bolt and screw present and correct. Only two problems encountered though.  When I took the MCU out of its bag something fell on to the bench. It was the USB socket, the solder was so dry it had fallen off, easy to refit but indicative of the dreadful quality of solder flow, a bit worrying for the future life exoentancy of these boards. On inspection the others look poorly flowed as well   The second problem was that when powered up with either of the DCP405 modules installed the unit was in a reset loop, unplugging the modules enabled the unit to boot up ok. This was solved with a firmware update from v1.0 to v1.6.1   All working now, very happy and just waiting for the DCM224 hopefully later this month according to Mouser.
Have calibrated it with a newly calibrated DMM6500 and the accuracy and stability of the BB3 is really impressive.  So far 10 out of 10, thank you to the developers, a great product.  Now to start programming

With regards to the electronic load you are proposing to design, are you in danger of reinventing the wheel?  There are plenty of loads available for very reasonable money already.  I have used several different models over the years but this one from Tenma   https://uk.farnell.com/tenma/72-13210/dc-electronic-load-prog-30a-120v/dp/2848407   is an amazing unit for around 250 euro. It is spot on accurate and comes as standard with USB and ethernet, from a business point of view could you not just utilise something like this and save a lot of development effort.

In fact spend time producing a 2  and then a 4 quadrant module that interfaces with the Tenma   I already use it with a power supply to make a crude 2 quadrant test bench to run intelligent charger tests.
Thanks again for a great project, Mike

[prasimix]

Firstly, my BB3 arrived Thursday and went together very well, all holes lined up and every nut, bolt and screw present and correct. Only two problems encountered though.  When I took the MCU out of its bag something fell on to the bench. It was the USB socket, the solder was so dry it had fallen off, easy to refit but indicative of the dreadful quality of solder flow, a bit worrying for the future life exoentancy of these boards. On inspection the others look poorly flowed as well   The second problem was that when powered up with either of the DCP405 modules installed the unit was in a reset loop, unplugging the modules enabled the unit to boot up ok. This was solved with a firmware update from v1.0 to v1.6.1   All working now, very happy and just waiting for the DCM224 hopefully later this month according to Mouser.
Have calibrated it with a newly calibrated DMM6500 and the accuracy and stability of the BB3 is really impressive.  So far 10 out of 10, thank you to the developers, a great product.  Now to start programming

With regards to the electronic load you are proposing to design, are you in danger of reinventing the wheel?  There are plenty of loads available for very reasonable money already.  I have used several different models over the years but this one from Tenma   https://uk.farnell.com/tenma/72-13210/dc-electronic-load-prog-30a-120v/dp/2848407   is an amazing unit for around 250 euro. It is spot on accurate and comes as standard with USB and ethernet, from a business point of view could you not just utilise something like this and save a lot of development effort.

In fact spend time producing a 2  and then a 4 quadrant module that interfaces with the Tenma   I already use it with a power supply to make a crude 2 quadrant test bench to run intelligent charger tests.
Thanks again for a great project, Mike

Hi Mike and thanks for the helpful feedback. I don't know if you followed what we went through in the first production batch of BB3 and, I would say, the criminal actions of the Bolek & Lolek bros, so it's no wonder that the USB socket fell out. I believe that no other problems will arise and that BB3 will serve you well.

It's strange that the DCP405 r2B11 (green PCB) was causing problems with 1.0, but it's good that it's working with the latest firmware now.

Regarding BB3 programming, I think the real fun is just coming because we are working on EEZ Flow, I will soon start writing a bit more about this in the https://www.eevblog.com/forum/testgear/]Test equipment[/url] section of the forum. In the meantime, feel free to install EEZ Studio and maybe play around with examples of MicroPython scripts that can be downloaded to BB3 with few clicks.

Regarding the mentioned electronic load: yes, it will look like a reinvention of the wheel, but the whole BB3 is something like that . In terms of price, if I manage to make a new module, it will hardly be as attractive as the mentioned TENMA. Such prices are impossible to achieve with production in the EU and with modest batches (100 or less pieces). However, what such a module could offer is convenient integration with other modules with GUI just as in the case of existing modules.
By the way, @wizard69's idea from post #558 isn't bad at all. It should definitely be taken into account.

Yes, a 2- and 4-quadrant DC power module would certainly make more sense, it has been talked about for years. I guess something will happen about that sooner or later.

[prasimix]

A few words about the problem with drifting when measuring current. It seems I needed to re-solder that section and clean it better. Now it looks quite fine at least for currents up to about 5 A, i.e. in accordance with the most accurate DMM currently available to me: Fluke 287 which is not a miracle of technology .

I also tested with higher currents, i.e. 9 A and found out a bit more about how much thermal stress this can be for a fuse (rated for 10 A). I naively left it covered and without the fan working. Here's what I got:



I tested it again, but this time without a cover and with the fan on. A temperature of about 60 oC at the ends of the fuse terminal seems acceptable to me.

[Mike G]

Good morning Prasimix,  thank you for such a complete reply to my post.  I was considering a major resoldering job but have decided to run with the green boards as they are and see how things progress.

Strangely the MCU, AUX-PSU and back plane are green PCB but both the DCP405 are blue PCB, what does the blue PCB material signify, are these froma different manufacturer or a different production run?

Yes I researched the forums thoroughly before ordering, I installed EEZ Studio a while ago and have been learning about this for a while, I admit I am struggling a bit with the GUI design and interfacing to a python script but the "Hello World" tutorial is a good introduction to work on from.  What with this and learning the Keithley Text Script Builder (runs Lua based script) plus the fact that I am an old school hardware engineer, who is not trained in programming at all, I certainly don't seem to have enough hours in a day

The comments on the electronic load were not meant to be taken too seriously but with the interfaces available ( I forgot to mention it has RS232 as well) and the fact it readily connects to EEZ Studio I intend to explore the possibility of doing something with it when I find time   I agree that a separate enclosure to accomodate more slots is looking the best way forward, if this is merely a slave module to BB3 then it could also be placed out of the main equipment rack so saving front row bench space.

I would like to think that the 2 and 4 quadrant modules become a reality as the load development progresses, the possibility of future modu was a major factor in my decision to go for a BB3. A module that can sweep 2 quadrants supply, by that I mean supply a voltage to sweep both negative and positive would be a good first step perhaps

It is good the way you include the forum in the loop when developing ideas, keep up the good work,
Mike

[prasimix]

Hi Mike, Please note that a number of BB3 kits from the first production batch (green PCBs) were completed with DCP405 modules from the second production batch because AssemTec Europe (aka Bolek & Lolek) never managed to deliver all the ordered DCP405 modules (thus further increasing our loss). The blue modules are version r3B3 and cannot work with the v1.0 firmware. It was not written in the building instructions which I have now corrected. I'm glad you were able to handle it yourself.

Blue PCBs come from another PCBA contractor, and as much as green PCBs are badly soldered, so much these blue ones are possibly the other extreme: from the first some components fell off themselves that I had to solder them again, from others the solder is so good that it needs to be equipped with quality heat gun and possibly a preheater plate so that the components can be detached.

MicroPython scripting certainly requires a lot of skill if you want to add a GUI. However, all the blame is on me as there is still a lack of extensive documentation for EEZ Studio and more described examples of MicroPython scripts (I believe that Chapter 16 of the User manual helps at least a little). Writing quality documentation is a full time job and we should have at least one dedicated person for that. I believe this will improve as EEZ Flow (conceived as an open source alternative to NI LabVIEW and Keysight VEE) progresses as we intend to add some wizards as well and finally fully document EEZ Studio.

[exe]

I also tested with higher currents, i.e. 9 A and found out a bit more about how much thermal stress this can be for a fuse (rated for 10 A). I naively left it covered and without the fan working. Here's what I got:

Ah, fuses... This seems to be a common problem. Check this thread, and may be my message here https://www.eevblog.com/forum/blog/eevblog-1377-the-amazing-unpredictability-of-fuses/msg3483916/#msg3483916 . TL;DR: fuses can get very hot and don't trip reliably unless heavily overloaded. Not sure what to suggest here.

[prasimix]

Basic power analysis in the AC domain has been added. This is something you can’t get easy with DMM and DSO.
The 47R in series with 20u was used as a load for the test to have an active and a reactive component. The calculated Irms can be compared to what DMM measures:



Vertical and horizontal view:



Here I have a dilemma for which I would need your opinions. If we are in the AC domain the NPLC should be 0. In that case if the current and voltage are measured it should actually display the RMS values. Is it acceptable to then add the option to select AC or DC mode or should it be done in a different ("smarter") way?

[Kleinstein]

The AC power analysis looks nice. One starts to see the limited screen size, but it is still OK.

The front-end has no AC coupling (with rather high resolution ADCs this is not a real problem) and thus no need to switch between AC and DC mode.  The more normal choice is a combined AC/DC mode with multiple possible results to show. Just for a single channel one could have: DC (average), positive and negative peak values, peak to peak value, RMS (DC coupled) , RMS (DC subtracted) and possibly even more. The averaging time is a prameter that can apply to both DC and AC, though with AC it really makes sense to only choose full periods. With just numbers on the screen very fast does not make much sense.
With enough space there is relatively little need for just a simple DC and AC reading. I would more consider the combined DC + RMS the natural choice and the user decides to ignore the number he may not need.
For the RMS there is the RMS reading with full bandwidth and the std. dev of the slower DC readings which is kind of a slower fitlered RMS value.

With 2 channels combined there are additional readings like the ones shown for power and also correlation (kind of lock in function - practical use may need an expternal input amplifier, but this is not a big problem).
So one has a large number of possible results to display and choose. So it is kind of pick some 6 values to show. For convenience there may be preset combinations like the power analysis as shown, or single channel with something like DC, RMS(AC)  and maybe Std. dev for DC

[prasimix]

Many thanks @kleinstein. In addition to the vertical and horizontal views, the results of the power analysis will be possible to display in a maximized view, similar to what we have in post #555 where voltages, currents, powers and efficiencies are displayed.

It looks like there will be a lot of work to be done so the functionality would be gradually added through multiple firmware versions. Some of the things we will try to calculate in real time, some more demanding such as harmonics analysis can be done on a recorded sample.
Having combination preset seems like a good approach: in fact it's the answer to my dilemma, but I see that for a start it would be good to add next to DC and RMS and Std.dev for DC.

I will need to review what the power meters of others have to offer, if anyone has a particularly interesting/attractive model to recommend, please let me know.

[prasimix]

... and the power measurement for AC/DC converter (my abandoned CF-DIC project) directly connected to 230 Vac:



Same thing for off-the-shelf Mean Well LRS-150F-48 used in BB3:

[prasimix]

In case anyone is concerned about the safety of the connection to the 230 Vac mains, here is one close-up image. Everything is properly insulated:

[Kleinstein]

The connectors are rated for 250 V, but at some areas (e.g. around Q1) the clearance does not look like it is up the standards for mains safety.  It depends a bit which inputs are actually used. It can work, but there is some risk and there could be quite some damage from an arc over.

[prasimix]

The connectors are rated for 250 V, but at some areas (e.g. around Q1) the clearance does not look like it is up the standards for mains safety.  It depends a bit which inputs are actually used. It can work, but there is some risk and there could be quite some damage from an arc over.

You're probably talking about AFE1, this is AFE3. Here is how it looks like:



Probably I should cut PCB over L1, R1, R3, R5 and between X1 (pin 1 and 2) and F1 to get extra clearance.

[Kleinstein]

The 2 isolated inputs (should be the 4 lower pins on the board and the 4 upper pins on the connector when build into the BB3 should be OK for mains. It is the other 2 input where the spacing is too close for mains, especially around Q1 and R13. 

The board picture also shows the weak point at the link from R35 to R32.  Ideally R35 should have separate contacts for current and voltage and R32 connected to the voltage reading contacts.  With more solder for this link the drift in the current may be reduced.

Another point I just notices is that the lowest current range (33 Ohms shunt R30) may have problems with 48 mA, as the diodes would start conducting. So the maximum usefull current would likely be lower, more like 30 mA.
For a different reason the mid range is likely also not really useful to 1.2 A as was noted in the schematics - likely more around 0.5 A. With more current the FETs may get quite warm and help heat up the resistors. One may have to consider a smaller fuse (e.g. 630 mA).