PDM on 5430 at 3.072MHz random glitches in the stream

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Fri, 18 Oct 2024 11:41:48 GMT

Hi,

Yes, I understand that the documentation is actually unclear regarding this.

[quote user="tamisoft"] previously you said your microphones from steval-mic005 worked for you, or was that the same mono sampling problem?[/quote]

Yes, I tried again with them and it does look like there was the mono sampling issue.

We guarantee higher speeds in the later SoCs like the 54, but for the 5340, it's the 16kHz.

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Wed, 16 Oct 2024 10:45:29 GMT

Hey,

the documentation doesn't explicitly states it *only* supports 16k and the driver uses the product spec's calculations to determine the requested vs. real sampling rates, it imposes no limits on them. Even faster speeds work (like 50kHz when 48kHz requested by with the 32M oscillator) so it is really not a speed issue. - and previously you said your microphones from steval-mic005 worked for you, or was that the same mono sampling problem?

If this is something the processor really won't support I think either the driver and/or the documentation should be updated (doc: only supports 16Khz, driver: reject any rates other than 16k, and then than we can close this thread.

Levente

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Wed, 18 Sep 2024 07:08:11 GMT

Hi Levi,

Unfortunately these thoughts were very early in the case and after further discussion internally, they mention that the 5340 does not support the 48kHz, and since the mentioned 16kHz works as expected, there does not seem to be much that we can do in this case. And as mentioned earlier, there are the 54 series which does support 48kHz now.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Tue, 17 Sep 2024 10:07:17 GMT

[quote userid="107729" url="~/f/nordic-q-a/91991/pdm-on-5430-at-3-072mhz-random-glitches-in-the-stream/409095"]

Priyanka said:

3) Perhaps they could try using the different taps in the phase mux to shift the PDM_CLK at IO versus the PDM_CLK at the filter: https://projecttools.nordicsemi.no/infocenter/index.jsp?topic=%2Fyoda%2Fpdm.html&cp=3_3_0_0_6_25_7_29&anchor=register.PHASE

[/quote]

Referring to this, sorry page confusion :D

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Tue, 17 Sep 2024 10:06:42 GMT

Hey, it is great you have managed to reproduce the problem, can you try the workaround from this, because we can't open the internal page. Maybe you get a magic setting that makes it work.

Thanks

Levente

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Tue, 17 Sep 2024 09:35:55 GMT

Hi Levi,

I tried playing around with it and finally was able to figure it out. It does look like only one of the mic is being detected and I can record mono and not stereo. Nevertheless, it does look like the 48kHz sample causes glitches whereas the 16kHz is smooth, just like you mentioned.

And as earlier mentioned, we guarantee the working for the 16kHz, and that does seems to work well.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Thu, 12 Sep 2024 07:33:55 GMT

Hi, that's odd, but on the other hand the two look identical, like if it was mono. Can you try to touch the two mics independently to verify that it is indeed stereo with two independent streams? Because the two mics noise floors are visibly different normally.

Levi

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Thu, 12 Sep 2024 06:01:29 GMT

Hi,

I am not sure why, but this is what I get:

Both Left and Right with no glitches.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Tue, 10 Sep 2024 10:06:20 GMT

Hi, correct, Brown GND, Red VDD.

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Tue, 10 Sep 2024 08:00:37 GMT

Hi,

Could you confirm what the red and brown wires are? (Is it GND and VDD):

It's a bit difficult to make out since it's not labelled on the board.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Wed, 04 Sep 2024 09:59:15 GMT

Hey Priyanka, I think it was using ncs2.4.0. I can't comment on the supported sampling rates, because we are just using whatever the sdk let us. And at least from the product spec the f_actual is provided as a calculation, and that gets integer numbers. It has something to do with that one exact frequency provided by ACLK, because as you saw above, turning it to 32M clock input it won't have the problem, but the actual sampling rate will be 50kHz.

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Wed, 04 Sep 2024 08:27:39 GMT

Hi Levi,

I discussed this internally with few experts and they mentioned that the nRF5340 supports only a 16 kHz output sample rate. So then setting the device on 48KHz, it will produce some issues.

We do have products that support 48KHz, 16-bit samples, such as nRF54L15 and others. Maybe these can be your choices.

Also, I am trying to reproduce the glitches here using your hardware, but when trying to build your sample from you Github link, I face some issues (probably because of the updates we have had in our SDK over the years.) Could you tell me which SDK version you had tried it approximately?

(Even if you can't remember it's fine, I am anyway working on resolving this at the moment. Just wanted to know.)

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Mon, 02 Sep 2024 09:20:45 GMT

Hey Priyanka, 🙂

I have jumped on the opportunity to address this one more time when they guys visited us. So the way we have worked around currently is not to use 48k sampling on the pdm, but rather 16k. This cost us in SNR 3-5dB.

The problem is with those microphones I sent you (we have those on our hardware versions since the v1.0) is that when PDM is used from the NRF5340 with 48k 16bit stereo (like the usb audio example does), one of the channels has a continuous random clicking noise (bad samples). The github repo above was the most minimalistic example I created, works with the nrf5340 dk and shows the symptoms on the usb audio recording. Same mics were unaffected in the same mode when external converters were used to PDM 2 I2S, we even used on 2 separate builds - 2 different converters, none showed this behavior.

You tested it with the steval-mic005 microphones, and those worked for you.

This is pretty much where I didn't have time unfortunately to continue, since we had to move forward.

I hope you'll be able to reproduce the problem because then I am sure you'll have a fix, as above you had some links to internal information that could have been tried, but since we have no access to that only you'll be able to test those, and maybe let us know a quick fix 🙂

If you need anything else, just let me know.

Levi

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Mon, 02 Sep 2024 09:05:37 GMT

Hi Levi 🙂

I have received your hardware and am looking into this. It would be great if you could mention here a short version of what the current issue is so that I can confirm it here.

Do let me know and in the meantime I will catch up with the previous works that we have discussed here in the ticket as it's been a long time 😃

Hoping to hear from you.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Fri, 03 Mar 2023 14:58:55 GMT

Interesting, hmm. In the meanwhile we have ordered the same microphones, so will test with those as well. Also I have created a 16kHz version of the code, to use that unfortunately you have to patch zephyr as well, since they have hard coded the 48kHz sampling rate for the usb audio endpoint, but I have included the patch in the zephyr-patch directory. offcode-oy/nordic_pdm_test at 16kHz (github.com)

What is interesting, that works flawlessly with the same setup, where the 48k glitches.

And as it was mentioned way above, we have configured the PDM3140 (PDM to PCM converter) to output the same 3.072Mhz as the 48k example code does, that recorded though the i2s was correct (hinting that , there is a difference how NRF converts the PDM and how the TI chip does it).

So somehow you should recreate the same problem on your end so you can try the suggested hidden register changes, if any of those help. We can send you the microphone board you can plug to your DK if that helps. 🙂

Some scope pics:

16kHz high pin drive

48kHz high drive

48kHz normal drive (original code)

As you can see we have changed also the pin drive strength to reduce the rise/fall times, since our microphones have a specified maximum 12ns (your ST microphones didn't specify that) and even the high drive is slightly out at 14-15ns. I don't have a measurement of that from the PDM3140 yet, and I think it shouldn't really matter, but who knows. This is still an unverified action point for me, because if I cannot reach the desired rise/fall times, I'll have to add more hardware to the board to test.

Levi

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Thu, 02 Mar 2023 13:32:05 GMT

Hi,

Tried with v1.9.1 too. And still no glitches observed.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Thu, 02 Mar 2023 12:49:07 GMT

Okay. I will try with v1.9.1 too. I used this module: https://www.st.com/en/evaluation-tools/steval-mic005v1.html

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Thu, 02 Mar 2023 12:47:28 GMT

Oh wow, now can you tell us which microphones you were using, so we can source some and try? Our SDK is 1.9.1 btw, I forgot to tell this in the previous mail. L.

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Thu, 02 Mar 2023 12:43:34 GMT

Hi,

I was able to do this, but unfortunately could not see any glitches in any channel.

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Wed, 01 Mar 2023 15:49:42 GMT

Hey, you can hear it on the USB audio output. It is hard to tell directly from the PDM signals if they look ok or not. The error seems to be added when the PDM gets decoded to PCM. Here is an audio sample I have just recorded https://mega.nz/folder/Bz4FybqJ#8JLwTfUgzYt-k3G8BM68ag First it has music from the speakers then I stop the music and you can hear the problem on the right channel samples. (if I can believe Audacity which is Left and which is Right :D ) Looks like this:

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Wed, 01 Mar 2023 15:19:19 GMT

Hi,

I have tried to reproduce this. I have connected two PDM microphones (standalone units) to the 5340DK as you had mentioned:

The CLK and DATA of both PDMs are connected to P1.04 and P1.05 of the DK.

I am trying with nRF Connect SDK v1.9.0.

I have inspected the clock signal with a Salaea logic analyser:

And something seems to be happening, but it could be that sampling pauses at certain intervals. How should the glitches look like?

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Mon, 27 Feb 2023 12:39:42 GMT

Sure, Infineon IM69D130 and Knowles SPK0641HT4H-1 .

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Mon, 27 Feb 2023 12:13:12 GMT

Hi,

Thank you very much. Could you provide the details of which PDM microphone you use?

-Priyanka

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Tamisoft — Fri, 24 Feb 2023 15:10:09 GMT

Hey. I have created a quick minimalistic demo that runs on the 5340DK. The code is here offcode-oy/nordic_pdm_test: Reproducing the PDM 48k stereo glitch (github.com)

The 2 PDM microphones are connected to:

P1.04: PDM CLK

P1.05: PDM DATA

The USB audio device reproduced the problem for me, simply record the audio and one channel should have the glitch.

Levi

RE: PDM on 5430 at 3.072MHz random glitches in the stream

Priyanka — Fri, 24 Feb 2023 09:43:15 GMT

Hi,

Is it possible to share an example project showing this, which we can reproduce here on the nRF5340DK? Also, how is the PDM microphone connected to the DK? Are you using some kind of small board which can be connected to the DK?

-Priyanka