Analog Input Module signal issue

PrsaLab

Hi Josh,
We recently assembled a couple of new AI modules v2. With both of them we observe a strange "staircase" signal when recording (see image below):
https://drive.google.com/file/d/1jg2c9bDnFUkXpy8UFUBEmKE_Rf3g4YW8/view?usp=sharing

For comparison we record the same signal with an older AI module v2 that we assembled previously, which looks normal:
https://drive.google.com/file/d/1_EcZlbBZ1DTuOv1OOpVKNz_Xnj6FyrUw/view?usp=sharing

Both modules are running the same firmware and Bpod software version. It therefore seems to be a hardware issue. Do you have any ideas what it might be?

Thanks!

Josh

Hello PrsaLab,

Looking at the screenshot, there are 8 voltage levels within each 5V box, with the range set to -10V : 10V. With 32 voltage levels spanning the full range, it looks like you somehow have a 5-bit ADC! There is no configuration register in the AD7606C that reduces the bit depth from 16-bit to 5, so I'm pretty much stumped.

Which vendor did you source the components and circuit board from?
Can you please share a close-up photo of the board?

Thanks,
Josh

PrsaLab

Hi Josh,
Thank you for the quick reply!

Here are the closeups of the circuit board:
https://drive.google.com/file/d/1pVYKVj7Hu9Ev50wIvBgOsz1kv2lK5Gas/view?usp=drive_link
https://drive.google.com/file/d/191UT6gRfBx-sb5iY72EVj1FngA1qwlmw/view?usp=drive_link
https://drive.google.com/file/d/1b83Pb2mxncaZPHUmh6ZUZyO3NkfizCOD/view?usp=drive_link

We sourced our components from DigiKey and Mouser and the PCB from SeeedStudio.

Here are two more comparisons of our older (good) AI module and the new (bad) module with the same 8 V, 2 Hz, sine wave input:
https://drive.google.com/file/d/13F8ivRJjCFjIXHAT6cKJTAQ3J9lLE6ja/view?usp=drive_link
https://drive.google.com/file/d/1Lz8nVnCaBodY-udqDenkiGJx5cPf61ZX/view?usp=drive_link

Josh

Hello PrsaLab,

Looking at the photos I'd double check the solder joints on pins 2 and 24 of the ADC chip. Cold solder joints are harder to detect. In general, I'd recommend a reflow soldering method for boards containing chips with fine pitch. For prototypes, we use a convection toaster in-house. JLCPCB is our new favored supplier for PCBs. You can add a metal solder paste stencil to your PCB order for a few extra bucks. We've had good luck with ChipQuick SMD291SNL10 solder paste. There are many videos showing how to hack your toaster to automate reflow, but if you're willing to babysit the 5 minute process with a flashlight and turn off + open the oven 20 seconds after the final component reflows, you'll have better control over the outcome. I'd kapton-tape a thermocouple to the metal tray to ensure that you don't overheat the board. Definitely reflow in an area with good ventilation (e.g. under a heat-tolerant fume hood).

That being said, it's suspicious that both of your builds are failing in exactly the same way, which leads me to suspect something other than soldering. Can you please send photos of both boards, including the bottom side and closeups of the chips? If you add light to both sides of the board to reduce shadows, the details will be more clear.

Thanks,
Josh

PrsaLab

Hi Josh,
Thanks for all the suggestions and tips for reflow soldering! We indeed have a reflow oven in the lab but for some reason our technician preferred to use manual soldering. We'll definitely use reflow in the future!

Here are the pictures of both boards:
Board1:
https://drive.google.com/file/d/1UAxX7BIQvYk4LrsPnRGOR1GYy2pXF-u9/view?usp=drive_link
https://drive.google.com/file/d/1yNu59L8LOBW2VBMUmNY0Et3ZF1ALrpJL/view?usp=drive_link
https://drive.google.com/file/d/14ehkzeolN_QwuTUWSYf_FOMlo6eM_rGV/view?usp=drive_link
https://drive.google.com/file/d/1ojx3GakbDq9R3JsDxp0e2UtfZAUJZsuT/view?usp=drive_link

Board2:
https://drive.google.com/file/d/1SM7i03yc5C3S0o-_D0A_ovUn_CvxJUWS/view?usp=drive_link
https://drive.google.com/file/d/1Aw2ZeowDLzlUx_BcK6qyvlJWzkEaIwIy/view?usp=drive_link
https://drive.google.com/file/d/1hwn9dOUhICdSf7YyIVcFeqE2M86b3nbV/view?usp=drive_link
https://drive.google.com/file/d/15-gdBWSZcDbZmN2DPuGMNFzrWWSmSYfP/view?usp=drive_link

Thanks again for looking into this! We'll double check all the solder joints on the ADC chips in the meantime.

PrsaLab

... I just finished testing both boards and all the pins of the ADC chips seem to be well soldered to the PCB pads.

Josh

Hi PrsaLab,

I don't see a mechanical issue that is common to the two boards.
I'm still suspicious of the soldering on the ADC chip. Board 1 has bent pins (e.g. pins 35, 46) and pins with globs of charred flux+solder that may indicate a poor connection (e.g. pins 64, 16). Bent pins can occur when solder bridges are removed during rework, and care has to be taken to avoid overheating and mechanical stress on the pins.

Again, since both boards have the same bug and no common mechanical issue, I'm leaning towards other explanations.

Were the PCBs for the first build (1x working) and second build (2x non-working) from the same batch of PCBs from Seeed?

Is the firmware a fresh pull from the repository? If you already have the current version of Bpod_Gen2, you can also load using LoadBpodFirmware.

Thanks,
Josh

PrsaLab

Hi Josh,

We have 2x boards working from the 1st build and 2x boards with the "staircase" issue from the 2nd build. All PCBs are from the same batch from Seeed.

The firmware is the latest version uploaded using LoadBpodFirmware.

I'll do some more testing of the ADC pins to see if I can identify soldering problems.