Sound acquisition using PCB 378B02 condenser microphone and NI 9250; need help

[pts92]

Hi,

I am working on a project whose goal is to acquire sound ranging between 5 Hz and 5,000 Hz.

The problem is high levels of noise in my sound acquisition setup, leading to poor SNR especially in the low frequency range (below 200 Hz).

The setup I am using is:
1. 50 mV/Pa condenser microphone, PCB 378B02: http://www.pcb.com/products.aspx?m=378B02
2. National Instruments 9250 C-series module: http://sine.ni.com/nips/cds/view/p/lang/en/nid/214634
3. National Instruments CompactDAQ 9181 one-slot chassis, Ethernet output
4. Windows 10 laptop running custom-built LabVIEW application

Please find attached several power spectrum density (PSD) measurements and a WAV recording that present the issue.

Screenshot 1: The NI 9250 is connected to the PCB 378B02 condenser microphone (via 1.5 m cable)
Screenshot 2: The NI 9250 is connected to a 1.5 meter BNC cable, unterminated
Screenshot 3: The NI 9250 BNC input is terminated with a BNC 50-ohm terminator resistor
WAV file: A sound recording taken with the current setup. Note the "hiss" in the recording. There is also low-frequency hiss or rumble, which is best heard using over-the-ear headphones or loudspeakers (not laptop speakers).

Notes on measurements/recording:
- the measurements and recording were taken in a quiet room (not an anechoic chamber)
- IEPE current of 2 mA was used throughout
- Sampling rate of 25.6 kS/s was used throughout. Increasing  the sampling rate up to 102.4 kS/s did not improve the SNR.
- The NI 9250 was chosen due to its low input noise density: 90 nV/sqrt(Hz) at 10 Hz and 40 nV/sqrt(Hz) above 1 kHz. (Source: NI 9250 datasheet, page 12: http://www.ni.com/pdf/manuals/378009a_02.pdf)

Questions:
1. How to remove the low-frequency "noise spike" seen in screenshot 1? What is causing the noise spike?
2. Do I need an amplifier in the signal path, between the microphone and the NI 9250? What amplifier would you recommend?
3. Can you please offer any examples of sound acquisition setups with good SNR between 5 Hz an 5 kHz? What microphone-amplifier-data acquisition setup would you recommend? What would be some jellybean devices that people use for this kind of application?

I realise my questions may be quite basic. Please excuse my inexperience with sound acquisition/audio electronics.

Any tips or resources are much appreciated. Thank you!

[BrianHG]

Pleas provide logarithmic frequency plots.  Linear plots are not too useful here.  That noise spike is below 200hz and I want to see if it is just background hiss, or, 50hz/60hz injection from an AC source.

Can you also provide 3 recordings with and without mic, NO music in the background please, just as silent as you can get it.
10 seconds of each should do.  You may use flac to compress the files as it is a true lossless codec & identical to a .wav, just the files should be half sized or less.

[pts92]

Hi Brian,

Thanks for your reply. I am going to provide logarithmic frequency plots and recordings without music in the background tomorrow, 21st of March.

Did you spot anything wrong with the current setup (PCB 378B02 + NI 9250)?

[BrianHG]

I am not familiar with the hardware in question, your plot #3 shows me the noise floor of your sampler, so it is good.  I want to diagnose what your microphone is picking up & I may have a question after that.

Keep the 10 second audio samples of everything...

[DaJMasta]

You say it was taken in a quiet room, was there any air circulation on?  How is the shock mounting of the mic?

I don't really think it would account for such a large spike, but low frequency vibrations could at least be contributing to that, and they will pass through structures easier than higher frequencies and will generally be less noticeable to humans.

Are you using the 426E01 preamp listed at the mic element's page as well?  If not, maybe it's not as good of a low end performer. As mentioned maybe there's mains hum or something similar being coupled in as well.

Probably better than trying just an audible analysis, run that same FFT in just the lowest frequency range (DC to 200Hz) and see if the extra resolution gives you some peaks that will help track down the source especially if it's something like mains hum interference.

[pts92]

Hi Brian,

Please find attached:

1. Logarithmic plots of:
a) NI 9250 connected to PCB 378B02 microphone
b) NI 9250 unterminated input
c) NI 9250 terminated input

2. Ten-second WAV recordings of microphone in quiet room. (I took note that it would be best to use FLAC encoding. I am aware it is lossless, however I have no easy way of converting from WAV to FLAC in LabVIEW. I am going to find a conversion tool if needed.)

Thanks for any help and suggestions!

[pts92]

Hi, DaJMasta,

1. There was no air circulation in the room. The mic was placed on a bed (mattress). There was no heavy machinery (diggers, etc.) working near the room. Vibration was taken into account: in the system deployment location the microphone is going to be supported through rubber dampers.

2. Indeed, I am aware that low frequency travels further and is harder to detect if present when recording is done, but thanks for pointing it out.

3. Indeed, I am using the 426E01 pre-amplifier. The pre-amplifier is directly attached (screwed-in) to the 377B02 microphone transducer.

4. I can sample at 102.4 kS/s (max of NI 9250) and then use a software digital low pass filter. I am going to measure the Power Spectrum Density of the filtered signal (0 to 200 Hz).

Thanks!

[BrianHG]

You forgot to send me the 10 second wav file for plot 'C'.  I can tell you at this point you do not have any contamination from electrical supply or poor grounding.


Here is a hint to your possible problem.  Plot the spectrum of a 'Pink Noise' source waveform, both in linear and logarithmic frequency plot.  You might see something similar.

Your audio tools should be able to provide you with reference White Noise, Pink Noise and Brown Noise.

[BrianHG]

1. There was no air circulation in the room. The mic was placed on a bed (mattress). There was no heavy machinery (diggers, etc.) working near the room. Vibration was taken into account: in the system deployment location the microphone is going to be supported through rubber dampers.

After amplifying your sample & playing it on a really flat sound system down to 5hz, I'm getting some VLF which sounds like either highway traffic, or, basic wind against the windows.  We are talking things below 25hz, with the signal getting massive below 4hz.  Do you have a highway within 2 miles of your place?

The audio is mostly real, though, there is some general hiss mixed in.  Try doing your recording with the mic stuffed between a bunch of pillows.

Also, filtering the sound with a 20hz high-pass filter gets rid of most of your noise, but, I suspect you want the VLF due to the use of instrumentation ADC.

Also, sample 2b has a sound like you are typing on a laptop keyboard.
I've attached my FFT of sample 2a, I scanned the region where you weren't shifting your position in your chair...

[DaJMasta]

Those plots do show a clear picture, from just listening to the samples with the gain substantially boosted, the sort of noise (at least in the audio band) that is present does seem like recorded sound.  Some degree of white-noise like truly random sound, but then predominantly very low whirring and other noises - what I usually think of as normal ambient noise rather than just true random noise or hum of any sort.


I was testing out a DIY sound isolation chamber yesterday and observed very similar results using a commercial low noise recording mic (Rode NT1).  In a quiet room, with the air circulation off, the door closed, on a bed, in a box made of acoustic foam sandwiching a layer of neoprene, I see a very similar overall response.  It's also worth noting that in evaluating the performance of the chamber, it's clear that below about 200Hz, there is not much attenuation, though this is to be expect since the foam was "only" 4" wedges on one side and 2" pyramids on the other - not nearly thick enough to do a great job of attenuating low frequencies.  You will get a little damping and some noticeable decreases, but the low end noise you're observing is the reason why concert halls are frequently made with a floating slabs for the hall itself and why anechoic chambers have such thick absorption material, insulation, and isolation from outside noise sources.  If you have an entirely interior room, you may be able to reduce the wind and traffic noise that can get in somewhat, but the low frequency stuff will still predominate.  Sealing gaps above and below the door for the room being used or even covering the vents could help keep noise out.

From inside the chamber, I could still see a number of periodic signals - the one at 120Hz may have been from the mains since the cable runs went by a fair bit of mains wiring and were not especially well shielded, but most of the rest is external noise, you can see in the spectrogram cars and trucks driving in the neighborhood (and the Doppler effect of them passing) and the patch of higher noise in the middle is actually a plane high overhead, only barely audible from the neighboring room where I had the FFT running.