EEZ Bench Box 3 (BB3)

[Kleinstein]

For the standard pitch, there are very cheap cables with low quality. So one may have to be a little more careful in the choice. However I so no reason why a smaller (less common) pitch would make things much better. The soldered connections may be more reliable then IDC.

There seem to be some unreliable IDC cables, but I am afraid this is no limited to the standard pitch.
With a low pin count one may use classical crimped cables.  Usually there is enough space between the boards, to no need to use the angled connectors. For an analog signal it may be just 2 or 4 lines that are needed.

[prasimix]

An apology for what was said in the previous post: it's not that the cable has a bad connection, but the connection between the plug and the socket is problematic because the number of pins is too small. This is especially pronounced for the 6-, 8-pin 0.1" version. The 10-pin is slightly better but still questionable.

[prasimix]

Follow up on post #438 where I expressed dissatisfaction with the bias voltage display when the input is floating which is why I would try to use another ADC. @Kleinstein in #440 suggested using ADS131xxx which is a very interesting series and offer 24-bit (in theory) and simultaneous sampling for affordable price. However, I now have a problem with the bipolar signal. They like most others do not have bipolar input. I have two options: to add an op-amp (per channel) for level shifting or to look for another candidate.

[Kleinstein]

The ADS131A04 actually has the option to run with a +-2.5 V or similar analog supply and this way operate also with negative voltages at the input. There even is a charge pump integrated to create the negative supply.  In this respect it is one of the few SD ADC chips that can work bipolar even with a single ended signal.
The ADS131E... also can use a negative supply and have buffers included.
The description is a bit confusing however. 

Some difficulty comes from the limited (e.g. +-2.5 V) range, so a 10 V voltage would need an extra divider at the input. On the upside the noise is relatively good, so that one could directly connect something like a thermocouple.

[prasimix]

New MCU module prototype powered by ULX3S displays the image concurrently on two screens: a TFT touchscreen and an external monitor. This could find its purpose in presentations and in classrooms.

[prasimix]

Thanks to @Kleinstein assistance I'm progressing well with MIO168 AFE modifications. New ADC will be definitely ADS131E04IPAGR that is pin compatible with 6- and 8-ch variant.
Existing Traco DC-DC converter will be also replaced with new one build around SN6505B, that looks like more flexible solution that could be also more silent. Since mentioned ADC is 24-bit, I believe that is at least 19-bit resolution achievable and in that case it will be good enough for simple automated voltage and current calibration, before a fully featured DMM module become available.

The question remains which voltage reference to use? ADC comes with 2.4 V internal reference with ±0.2% accuracy and 20 ppm/°C temp. drift. I can always set this as the default, but I would like to provide a place on the PCB for external reference.

It is not entirely clear what reference voltage I can use as I will use ±2.5 V for powering the analog part (to accept the bipolar input signal). In DS there are recommendations for AVdd of 3 and 5 V. If ±2.5 V can be interpreted as 5 V then 4 V is specified as the nominal reference voltage. Your thoughts are welcome.

[Kleinstein]

With an analog supply of +-2.5 V and having the negative input more or less tied to ground, it does not make much sense to have a reference voltage of more than 2.5 V.  The typical value is also 2.5 V - so this would be obvious choice. If there is a true differential drive, one could consider a higher reference voltage, like 4 V.

The ADC itself is specified with 3 ppm/K max., so the reference does not have to be a much better than the ADC. However the ADCs upper limit may be conservative to allow for easy testing.
An external reference would only make sense if considerably better than the internal one.
I would consider the max6225 or ADR441. Not sure if an extra reference driver is needed.

[prasimix]

Great suggestions, many thanks. MAX6225 looks somewhat better, and price is accordingly higher. The good thing is they comes with standard pinout so one can experiment (REF50xx has that pinout, too). However, I believe that buffering is needed, or at least the Fig. 29 is described as "typical external reference drive circuit". This isn’t the first time TI has pushed a buffer for its ADCs that is typically built around the OPA350.

[prasimix]

... and we have the first MicroPython application created by a community member thanks to @Jan28 (on Discord): Reforming electrolytic capacitor ...



The process is logged and can be viewed locally or on a PC thanks to EEZ Studio:

[jeremy]

Hi prasimix,

I have just finished assembling my BB3 kit. When assembling, I had the following issues:

- The hole in the front panel for the 5-pin push-in connector is slightly offset to the right, so it is quite difficult to push the connector in (it rubs on the left side of the connector). It would probably be a good idea to add some more tolerance to this. All of the other connectors are accessible (incl SD card).

- I am missing two of the black countersunk screws

- One of the meanwell power supplies (the one that came with the DCM220) had missing screws in the screw terminal blocks. Upon closer inspection, the screws were actually lodged inside the power supply chassis itself (inside the "honeycomb" case)! It's possible that if I didn't look for this, plugging in the power supply could have ended very badly... This was probably my biggest issue with the assembly.

- The MCU board was shipped with the nut and washer from the encoder loose inside the sealed antistatic bag; they were just sitting on top of the PCB. This is a problem because there is a battery inserted in the board, so it could have shorted something on the PCB. Also, in general having metal fragments from the washer on the PCB is a bit of a concern if the shipping was rough.

- It was quite difficult to insert module #1 in the backplane, as the upper tab on the front panel sheet metal interfered with the front panel of the DCP405. It would be helpful if you could add some more clearance here. Modules 2 and 3 were fine.

Overall, I think the assembly went well and the instructions were easy to follow. I look forward to tinkering with this device more.

Thanks!

[prasimix]

Thanks for reporting this so far unique collection of issues.

I wonder what that offset hole for a 5-pin push-in connector looks like. I would expect the holes on all the front panels to be the same, but they don’t seem to be.

The problem with mounting the module in the first slot I believe was caused by a improperly mounted display frame. You should probably move it a little to the left so as not to disturb the front panel of the module.

The issue with Mean Well seems to me the most serious. Since Mean Wells came in the factory packaging it is easily possible that their quality control did not notice that the screws had fallen inside (I was not in a position to open all the packages).

Anyway, I'm glad to hear that you managed to fix everything.

[prasimix]

I think that I have now all in place for redesigned AFE for MIO168 module that is based on 4-ch simultaneous sampling 24-bit ADS131E04. First I have to thank @Kleinstein once again for his selfless assistance and patience as we went through a dozen iterations. The ADS131E04 is a pin compatible with the 6- and 8-channel versions which can be handy for future upgrades.

The ADC is powered with +/-2.5 V to be able to accept a bipolar signal. A place for buffered external reference in SOIC8 packaging is also allocated. The picture also shows the driver for the latching relays shown below. In the bottom right corner are analog bus extenders that I called ADIB: these will be two 10-pin connectors on which two analog inputs with sense and guard and Gnd are exposed. On the MIO168 they will be connected to the inputs of the ADC which have active clamps on them (IC28).



I called the next AFE "Hi-voltage" because it can measure up to ±150 or ±240 V, depending on the divider used (Caddlock or discrete). For better accuracy, the voltage can be measured in three ranges. This AFE also has the option to measure current in case a 4-24 mA sensor is connected to the input. The current shunt has a limiter in front made of a combination of two depletion FETs, a resistor and a PTC, and at the AFE output there is an active clamp that cuts everything over ±2.42 V.



The next AFE I called “Hi-current” because it allows measuring of current up to 10 A in three ranges. It can also be used in voltage mode when it is possible to measure voltages up to ±15 V in two ranges.
The selection of the operating mode (voltage/current) is selected with a signal relay (ISEL_1), and hi current mode with a power relay (ISEL10A). Since the entire MIO168 is floated and uses a small DC-DC converter for isolation, it is important to take care of consumption. For this reason latching relays were used instead of regular ones. Relays with two coils are used, and the MCU will take care of set and reset actions. Relays with two coils are used, and the MCU will take care of set and reset actions.



To summarize, we'll have two pairs of AFEs that can measure voltage and current simultaneously. This means we can also measure power. In combination it can be used to measure the efficiency of, e.g. a DC-DC converter (input and output are monitored simultaneously). We basically have a rudimentary DC power analyzer as well as a entry-level DMM (the real one should follow, if we can reuse a 6.5-digit DMM module that @Kleinstein is working on). I will be more then happy if I manage to get 18-bit precision.

[Andrew McNamara]

Thanks for reporting this so far unique collection of issues.

It sounds to me like the package experienced a lot of vibration during transit.

[prasimix]

Yesterday I sent JLCPCB to make a new set of PCBs for MIO168 and AFE. Eventually I also moved the newly selected ADC to AFE piggyback and added two “ID” lines so that the MCU could recognize the different AFEs we might have in the future.



One detail remained open: I did not yet find a satisfactory board-to-board pair of connectors that would allow a distance of 22.5 mm between the AFE and MIO boards. Currently the closest I have found is a combination that gives about 18.8 mm (see picture) i.e. it is shorter by about 3.7 mm (approx. 150 mils). Since the receptacle is more or less the same height, it remains for me to find a larger pin header. The required number of pins is 36 (2 x 18). Any suggestions are welcome.

[exe]

Any suggestions are welcome.

Does it need to be exactly 22.5? If not, stacking of multiple headers could work . In the worst case, a short ribbon cable. Or pogo-pins or other spring-loaded contacts

[prasimix]

22.5 mm would be optimal, in that case 8-pin 3.81 mm connector on AFE piggyback (X1) will be in line with neighbor 16-pin two rows 3.81 connector (X3) that is 22.5 mm high. The AFE connector will hanging over same type of 8-pin connector on the MIO base board (X4). I already have enough distance between AFE and MIO but the front panel would probably look nicer if we could achieve those 22.5 mm.

[prasimix]

Got a recommendation from Samtec support to try the following combination: SSW-114-01-L-D + DW-14-10-L-D-550

[prasimix]

Since MIO now has a separate AFE I think we could have more different types in the future. The first one I presented earlier has the ability to measure high voltage and current. Here's what else comes to mind:

1. "Entry level" (i.e. cheap version) - similar to the initial version where we had voltage measurements up to +/- 10 V and current (4-24 mA).
2. "Isolated" - a Hi-voltage/Hi-current version that has two inputs isolated. In this way, the voltage and current on the primary and secondary sides could be monitored simultaneously. Isolated modulators such as AMC3306M25 could be used here
3. "Autoranging" - A version that would have fast current autoranging (two ranges would be enough to start) for accurate monitoring of IoT and similar devices where there are large and fast changes in power consumption.

I am especially interested in the third version, since we will have an ADC with simultaneous sampling, both current ranges can be monitored simultaneously, however the question of very fast (preferably of the order of a few microseconds) switching from low to high current shunt remains. I still have no real idea how to do it. Maybe a power MOSFET with a powerful gate driver could help here.

[prasimix]

MUX14D is a new multiplexer module that could be combined with MIO168 and SMX46. It has two banks of 7:1 multiplexers with dual lines and it can be combined in single 14:1 multiplexer. Multiplexer outputs are exposed on front panel connectors and ADIB connectors for internal wiring with other modules.

Module details are available on GitHub.

[prasimix]

Next level of modularity, MIO168 module with AFE4 piggyback are ready to be tested and support in firmware. Front panel is used from previous MIO168 prototype, next one will come without hole between two upper 8-pin terminal connectors:

[prasimix]

I'd like to share with you first results that I got with MIO+AFE presented in previous posts. It seems when the voltage divider is used (e.g. /100) that significantly degrade the measurement, or perhaps better to say it significantly amplifies the common mode noise and 50 Hz (mains frequency here) stands out so much. For example, then 2 V is measured with /1 range (+/-2.4 V) FFT looks like this:



On the same input (AIN1) with /100 range (+/-240 V) FFT for the same signal looks like this:



It's massive 40 dB difference! Connecting PE on AIN1- makes situation ever worse, adding additional 5 dB:



I put a Caddock voltage divider (1776-C68) on the AIN2 input, and the situation looks a bit better. For /1 range it looks like this:



... and with /100 range is about 10 dB better then with voltage divider made of discrete 1206 Susumu resistors (RG3216P serie) in AIN1 input:



The pleasure is that the complete process is made within the EEZ environment: the signal is sampled with 500 SPS for 15 seconds, with one click it is transferred from Data Logger to the EEZ Studio where with a few more clicks I have FFT.

The question is what can be done here. Should I deal further with the PCB layout, should I put an additional filter on the ADC input?

[Kleinstein]

Getting 40 dB higher 50 Hz hum with the divider active means the same voltage level is present at the ADC. This points to something like inductive coupling. Capacitive to the ADC input would be even weaker with the x1 range as the input is lower impedance.
Seeing mainly 50 Hz and not much 100 Hz suggest that it does not come from the supply voltage. At least the 50 Hz don't seem to come from the 2 V source.
It could still be coupling to the reference side - one would see the difference, when looking at the signal with a shorted input. This case would not reactor to the reference side.

With no extra shield the level of hum does not look so bad. The -100 dB without the divider actually looks quite good.
AFAIK the ground of the CPU board is not directly connected to PE and the case. So there may be some residual AC, e.g from Y caps at the supplies or similar. The level of hum may vary with other modules and possibly how the PC is connected.

One could try with an extra probe with an AC signal to see which area of the board is sensitive and may want extra shielding.

It is a little odd that the different dividers make so much different - I would normally not expect that much difference. The expected downside from the SMD resistor version would be more higher relative TC and maybe more effect of self heating - though usually not so bad at only 240 V max. The frequency response may also be different between the dividers - though usually not a big deal at 50 Hz.

[prasimix]

Thanks Kleinstein for your input and good idea tom perform measurement with input shorted. I got the following results:

1. AIN1 shorted /1 range (discrete divider):



2. AIN1 shorted /100 range:



3. AIN2 shorted /1 range (Caddock divider):



4. AIN2 shorted /100 range:



MIO+AFE was installed in Slot #3 (far right) when DCP405 power module was in Slot #1. Removing DCP405 does not improve results.

You're right about MCU module (CPU board): it is not connected directly to PE. MIO from other side is completely "floating" but its Gnd is connected to PE with 4n7/2kV. Removing PE wire on MIO does not make any difference.

[Kleinstein]

The hum does not look like it is coming from the reference.
I don't think the difference between the channels is from divider itself, more like the 2 inputs having different coupling to the outside and maybe to the unused channels 3 and 4. One could check the channel separation to get an idea the coupling: so apply some AC signal to channel 3 or 4 and check the response on 1 and 2. Even if not the source of the hum the channel separation should be checked anyway.
To measure even weak coupling, one can compute the correlation, like in a lock-in amplifier. So use 1 channel as reference and than look in the other channel for the same frequency. Ideally this could be done real time one board - though it is quite some effort, and the board still misses a low noise amplifier input to make it a lock-in amplifier.

The x1 range has a relatively large resistor at the input for protection - so it may still be capacitive coupling. 100 K for protection is about the same as the /100 divider. So coupling to behind the protection and to behind the divider would look similar.

One could try a makeshift shield (connected to the MIO ground), and for a test connecting the MIO ground to the case.

There is not much filtering that could be applied to suppress the hum - at least not analog. For slow measurements one can have digital filtering by averaging over multiples of 20 ms.

As a final solution one may need an extra shield. A revised layout (e.g. additional ground areas, maybe some thinner lines) may also help, though not sure how-much it would help.
I do not consider the hum so much to warrant a new PCB revision - just keep in mind if a revision is needed anyway.

As the measurements work, I assume the DCDC converter is also working OK. Is the injected common mode signal reasonably low (measured voltage over some 1 K from the isolated ground to the ground of the CPU board)?

[prasimix]

ADC channels 3 and 4 are not unused: they also have input stage, and for testing they are set in current mode, i.e. shunt is active over their inputs. Test with 1K connection between MIO and MCU GND is performed with x10 probe, and long Gnd clip.
If probe Gnd is connected to MCU side it looks like this (100 mV/div):



... and when probe tip and ground are swapped, i.e. Gnd is connected to MIO side it looks like this  (500 mV/div):