Hi community,
I have 2 MM8108 working in the mesh mode. While making a long observation, I saw that, the noise in one MM8108 (the right hand window) jumped and stayed stable from about -100dBm to -80dBm
It went back to -100dBm after a reboot
Please note that, there was no settings changes during the observation.
Is it something relating to AGC, LNA or so?
V.
Hi @bigboy061293,
That’s a fairly strange one… Was it a once off or are you able to reproduce this? Can I also get you to specify a few things on your setup?
Cheers,
Michael
Thanks for your reply.
This just happened randomly, but I will let it run freely with iperf test to see if I can reproduce. This is my setup:
In case it happend again, do you need some information such as dumpcore, cli…? If you provide some test scenario, I will try to set it up and run test cases.
Thanks,
Vu.
Thanks for the update.
If you do manage to reproduce, the outputs from morse_cli stats, iwinfo, and logread would be helpful.
Cheers,
Michael
Dear @michael.mccandless ,
Sorry for this testing procedure takes a bit longer than I expected. Unfortunately, I could not reproduce 100% what I witnessed. However, my findings below might (hopefully) provide some insights for your further investigation.
During last time, I had a chance to read back some points in the Wifi standard that might be relate to the noise measurement. I realized that “noise” could never be updated if the module does not receive anything from its associating network. This could be the part of the OFDM RX chain - I guess. So I set up a peer-to-peer transmission for logging the noise and other information
Let me describe my setup and how the data is capture:
iperf3 server, A connect to B with UDP. After that, they change the rolesmorse_cli stats was capture for both cases: Dataset 1 (DS1) and Dataset 2 (DS2)This is my general findings for both devices regarding the NB Interference Noise Floor FFT which has a strong correlation to the Noise of Device A
In DS1, the NB Noise Floor FFT swept broadly from 19 to 26 dB (range = 7 dB), becoming the dominant noise predictor because it directly reflected varying levels of external interference energy. LNA bypass was also a major driver because the weaker link (−55.8 dBm) meant the AGC was doing heavy lifting, and bypassing the LNA degraded the noise figure significantly.
In DS2, the NB Noise Floor FFT is almost completely frozen at 21 dB (range = only 3 dB).
I have attached the Datasets for further investigation.
I would like to know how do you think about the results
Regards,
Vu.
device A - DS1.csv (717.9 KB)
device B - DS2.csv (1.4 MB)
Hello @bigboy061293 ,
Thank you for trying to reproduce this. Spurious system level issues are a chore to reproduce.
Your testing setup sounds like some of ours, but with different hardware, so the source of noise could come from a variety of factors, eg one of the clocks could be causing a harmonic in the spectrum you’re running your tests in.
bypassing the LNA degraded the noise figure significantly.
Removing a gain stage increases noise?
Dear @michael.mccandless
In the common sense, adding the LNA/gain causes more noise. However it is not what I observed in this case. I don’t conclude so but this is what the correlation showed.
This graph shows factors that have the strong correlation with noise in Device A (DS1)
