ADAU 1467Z- Incorrect sensitivity measurement with ICS 41351
Hi forum,
Lately , I purchase some EVAL-ADAU 1467Z for measuring ICS 41351 mem microphone from TDK.
And i encouters 2 problems with this setup :
+Offset 1 dB FS compared to data sheet , the value in datasheet is 26 dBFS ± 1 , I got 27.5~28 dBFS for 10pcs measurement.
+Quite high noise floor from the board.
This is how I connect the mem to EVAL board:
+Mem CLK to BCLK0 at 768khz
+Mem DATA to LRCLK_IN0(MP10)
+3V3 from board as VDD for Mem.
The DSP sample rate is 48khz so I guess that makes the decimation value at 16X.
I wonder what I have done wrong to get incorrect sensitivity and quite high noise floor from the board.
My company has another analyzer with built-in PDM analyzer and I got correct sensitivity with this setup (around 26.5 dBFS for 10pcs)
That is an odd results. The hardware setup sounds fine. Can you post the project that you are using for configuration and test (you may need to create a zip file)? How are you accessing the mic output signal to check the level and noise?
Ken
Hi Ken,
Thank you for support.
I attach my project as zip file as suggested. PDM Audio_TKA preprogrammed.zip
The picture below is how ICS 41351 is measured here:
The dual analyzer system consists a BK4227 mouth simulator , BK4938 as reference signal and a mem microphone as dut signal.
Adau 1467z 's output 1 is analog signal of mem microphone after decoding.
Here 's what I have tried:
1) Measure with different types of PDM mem microphones (Merry MDM300 , Memsensing...) and still see 1dB offset compared to datasheet ( not just 1khz , full response curve is offset down by 1dBFS).
2) Switch between left and right channel (mono PDM format)
3) Try with CLK range from 768khz up to 3MHz
Nothing works .
Based on observation , it seems the board has quite high noise from output 1 even without microphone plugged in , it is around 210mVpp , i am thinking about how to suppress this noise with configurations but I haven't found anything yet.
When I downloaded the program to board , i have made sure to put DSP core on hibernating mode as suggestion in ADAU 1467z data sheet .
![20200710_080146[1].jpg](/media/forum/sigmadsp/167778/20200710_080146_1_.jpg)
You seem to be basing your levels on the analog output. Is that correct? If so, how are you calibrating full scale? The gain of the DAC output circuit is not calibrated exactly to have 0 dBFS equal to 1 Vrms. You would have to generate a clipped signal and find the corresponding analog voltage.
Alternatively, given your lab setup, I would expect that you have an audio analyzer that can accept I2S input. I recommend using a serial output to check the digital levels. I suspect (hope) that you'll find that the decoded gain as measured in an I2S stream matches the levels that you get when decode the PDM stream with the analyzer.
Thank you , Ken.
It 's true that we use 1Vrms as 0 dBFS as the value is printed on the board.
We don't have analyzer supporting I2S input here. However , the clipped signal sounds like a good idea , I will try that.
The offset problem is clear to me now.
About the high noise floor , do you have any recommendation ?
-With the board , the mic noise value is about -60 dBV (A-weighted)
-Without the board , mic noise value is aboud -85 dBV(A-weighted)
Please log in to reply.