0
\$\begingroup\$

I bought STM32F4DISCOVERY board and I developed a software that records .wav file using MP45DT02 MEMS microphone on this board. .wav files that I record are really good (almost no noise).

I designed a board based on the schematics of this evaluation board. When I used the same software, I have high noise at the background. I also can't record my voice unless I shout or scream. You can listen the record by this link.

I have 3.3 volts on the board and the capacitors are appropriate in terms of value and closeness to the microphone.

I have two boards and I have the same problem on both boards so it seems to me that the problem is systematic.

Could it be a problem of soldering or PCB cleaning after production of the board? Could it be a problem of noisy power supply? What could be the reason?

Edit: Below is the spectrum analysis of the two records. (Green: Eval board | Blue: My board)

...

Edit 2: Below is the channel meter of the two records. (Top: Eval board | Bottom: My board)

enter image description here

\$\endgroup\$
  • 2
    \$\begingroup\$ I'd suggest you check the microphone and its related connections. It sounds like it isn't getting power - the MEMS unit should have a built in amplifier, and without power you would have to shout to generate a signal large enough to hear on playback. The recording has (to my ears) NO background noises from where ever you are, but it does have lots of white noise. There's also a 100Hz signal buried in it (probably from your local AC power system,) so you aren't just generating random crap with your sampling. \$\endgroup\$ – JRE Jan 22 '15 at 11:39
  • 3
    \$\begingroup\$ Just because the schematics says it's got power doesn't mean it has power. You need evidence to rule it out (and yes I know this is sometimes not easy to confirm!)... dirty contaminated contacts/pads can cause plain bad soldering. For designs based on previous working examples I always look at connections being the problem before anything else... \$\endgroup\$ – Spoon Jan 22 '15 at 12:59
  • 1
    \$\begingroup\$ I used my ears and Audacity. My ears tell me there's no background sounds and that it sounds like white noise, and Audacity shows me the 100Hz signal, and that it isn't really white noise, but that the intensity falls off at about 3dB per octave - so more like pink noise than white, which means my ears need recalibration. @Spoon is right. What the schematic says is one thing. What really got built and connected is another. Check the power to the mic with a voltmeter, and while you are at it check the mic output with an oscilloscope. \$\endgroup\$ – JRE Jan 22 '15 at 13:04
  • 2
    \$\begingroup\$ If nothing else, the traces should run someplace where you can see them. Scrape the insulation off and measure there. \$\endgroup\$ – JRE Jan 22 '15 at 14:09
  • 2
    \$\begingroup\$ Just had a horrible thought. Could the MP45DT02 be mounted wrong? If it is rotated 180 degrees, it would be soldered OK, but nothing would be connected right. \$\endgroup\$ – JRE Jan 22 '15 at 15:26
1
\$\begingroup\$

It seems that microphones were damaged during PCB cleaning process. The producer puts spray on the PCB after production and small particules may enter inside the sound inlet during the process. When we changed the microphones, they started working. This explains the problem.

\$\endgroup\$
0
\$\begingroup\$

Audio circuits can be quite tricky.

A noisy power supply is definitely a possibility; SMPS's can be pretty dirty if you don't get component selection and PCB layout spot on.

Other possibilities include:

  • Poor power supply decoupling.
  • Circuit layout allowing noisy signals (e.g. a digital clock) coupling into the analog audio.
  • Ground bounce caused by a high current device, e.g. your storage media - are you using an SD card?

What it isn't is dirty PCB's.

\$\endgroup\$
  • \$\begingroup\$ - LM1117MP-3.3 is used to generate 3.3V from USB 5V. - I don't think that I have audio/analog circuits because I only use digital MEMS microphone and no analog components between them. - I use microSD card for recording. \$\endgroup\$ – sanchop22 Jan 22 '15 at 10:26
  • 1
    \$\begingroup\$ @blabla Capture dead silence (cover the port on the MEMS microphone). Use an audio editor (e.g. Sound Forge) to look at the audio signal that you're capturing. Compare the period of the noise (which you will be able to see/calculate from the audio editor) with the rate at which you're writing to the SD card, or any other high current periodic event. \$\endgroup\$ – markt Jan 22 '15 at 10:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.