AD7124 pyadi-iio rx() function is constant

Hi Bowei , the link you have provided is correct. The  move to change rpi-ad7124-8-all-diff-cs0-int25-overlay.dtbo to rpi-ad7124-8-all-diff.dtbo was to make the cs and interrupt pins reconfigurable at boot time via device tree overlay parameters. This way, we could use either of the 2 SPI ports of PMD-RPI-INTZ Evaluation Board | Analog Devices without the need to recompile the device tree.

To obtain the same configuration as rpi-ad7124-8-all-diff-cs0-int25-overlay.dtbo, we could use this in the /boot/config.txt instead:

dtoverlay=rpi-ad7124-8-all-diff,cs_pin=0,irq_gpio=25


As for the rx() being constant, there is an existing driver issue present in branch rpi-5.15.y (2022_r2)  that is being looked upon. This might be the cause of rx() readings being constant.

(+) Inconsistent channel position in a libiio buffer - Q&A - Linux Software Drivers - EngineerZone (analog.com)

Hi Kimpaller,

Thanks for the reply! It might be the same problem. In any case, could you reproduce my problem using

Hi Kimpaller,

Thanks for the reply! It might be the same problem. In any case, could you reproduce my problem using

Sorry Bowei unfortunately, I don't have an eval-ad7124-8sdz I could use right now. The issue on the libiio buffer was communicated to the developers  and looks like it affects any device using the ad7124 driver (cn0554, eval-ad7124-8-pmdz, cn0508). I believe there is a fix in rpi-6.1.y branch  (analogdevicesinc/linux at rpi-6.1.y (github.com) which will be part of 2023_R2 (not yet released). I'm still trying this myself and would love to update you on the result

I see. Do I need to build the rpi-6.1.y branch linux driver according to https://wiki.analog.com/resources/tools-software/linux-build/generic/raspberrypi or is there a prebuilt version available for download?

Yes, there are pre-built files from the latest build of rpi-6.1.y.

Download them from:
https://swdownloads.analog.com/cse/linux_rpi/rpi-6.1.y/rpi_latest_boot.tar.gz (should be  extracted in /boot)
https://swdownloads.analog.com/cse/linux_rpi/rpi-6.1.y/rpi_modules.tar.gz  (should be extracted in  /lib/modules)


Unfortunately, there is no rootfs yet for 2023_R2 but we could try the above files on an existing 2022_R2 image, though it is not guaranteed to work yet.

Thank you so much for the links! I am not very familiar with the Linux file structure. Right now on my Raspberry Pi, which has 2022_R2 Kuiper Linux installed, shall I delete the contents of /boot and /lib/modules directories, and put the files in your links into the respective directories? The rpi_modules link has 3 folders, which one should I use?

No need to delete, I just usually overwrite (/boot and /lib/modules) by copying all files as it is.
Its fine to copy all the 3 folders under rpi_modules to /lib/modules but if you are constrained with storage, you could just copy the folder that corresponds to your RPI variant,
ex. RPI4 -> 6.1.54-v7l+, RPI3 -> 6.1.54-v7+

Hi Kimpaller, I extracted the files as you said, but now when I use iio_info command, it can not find the ad7124 device. I also don't see an error loading the device tree overlay, so I guess the new kernel is not loading the device tree. Could you get the kernel to work properly for rpi-6.1.y?

Do you know which branch/pull request the fix corresponds to in github? I can't seem to find it.

Hi Bowei 

Sorry for the late response as I was also waiting for the fix myself.

There was a commit to rpi-5.15.y branch that was supposed to fix issue when multiple channel is enabled. I have built this and loaded the boot files/modules to an existing 2022_R2 image and tried my eval-ad7124-8 to test it


Please refer to this script I used with the experiment. I have enabled 3 channels and so far the ouput shows as expected.

Sample ad7124 script (github.com)

Thank you so much! I noticed in your code

ad_channel0 = 0 #voltage0-voltage1 plug to CH1 on my m2k

ad_channel1 = 4 #voltage2-voltage3 plug to CH2 on my m2k

ad_channel2 = 5 #voltage4-voltage5 left open

which is the same as what I found using 2021_R2:

I want to ask: why is the device tree configuration number different from the channel number in Python?

You're right, this ordering is a bug on pyadi-iio

I've investigated a bit and the culprit is on the sorting done here:

https://github.com/analogdevicesinc/pyadi-iio/blob/b579bbc055818537a0a28c291cee4cf0b323a048/adi/ad7124.py#L40

however, this was corrected on the latest version of pyadi-iio

https://github.com/analogdevicesinc/pyadi-iio/blob/2900d329d47a885ae443253dc90e94ce0b4d4d2c/adi/ad7124.py#L39-L43

another problem though in using the latest version is that a non-existing attribute "name" is referenced on a "channel" object, we could fix this by changing "chan.name" into "chan.id" on this line:

https://github.com/analogdevicesinc/pyadi-iio/blob/2900d329d47a885ae443253dc90e94ce0b4d4d2c/adi/ad7124.py#L39-L43