Using 40 low frequency Mics.

Question

Hi,

I am working on a device that takes input from 40 microphones and send data to PC via Arduino.

I want to use low frequency microphones (5Hz onwards) with around 200-250Hz sampling for each mic.

For this purpose, my preliminary design is like this...

40 Mics --> 40 PreAmp circuits --> Mux Shield II --> Arduino DUE --> PC

Please suggest me suitable commercial available preAmp circuits and any suggestion on the design is most welcome.

Thanks.

michaelkellett · Answer

To measure at up to 100Hz you need to sample each microphone at 500Hz or faster. For each microphone you will need a a buffer amplifier as shown on the Invensys data sheet and an anti-aliasing low pass filter which must allow the 100Hz signal thought but reduce the gain by, for example, 60 dB at 400 Hz. This requires 30dB per octave which is a fifth order filter. You could then sample at 800Hz and have no aliasing problems. To sample each of your 40 inputs at this rate is quite feasible but will need a sampling rate of 800 x 40 = 32000 samples per second at the processor. You might find it easier to sample faster and have a more relaxed spec for the anti-alias filter. I don't know of a commercial pre-amp that would do this - it will need to be custom made - I could do it for you (for money) or else I'm happy to advise you about how to set about it publicly on E14. I wouldn't use an Arduino for this - you'll need a custom board anyway so i would go for a processor with plenty of on chip RAM and number crunching. A board with 40 channels would be quite big - might be simpler to split it over a few boards. MK

dougw · Answer

Your answer suggests you are willing to spend over $1000 per channel to get the best possible. If that is not true, the description needs to be more complete in order to get a helpful answer.

rachaelp · Answer

DAB is right, the USB on the Arduino is serial over USB, i.e. it just shows up as a serial port on the PC and you'll probably be maxed out at 115200 baud (I may be wrong here but you should look into it to make sure before you build everything up). So assuming this is the maximum you can get and you are doing 12-bit sampling so you'll need to spread this over two serial words then with 7 bits, no parity and 1 stop bit you are at 18 bits per sample (two serial words) which is 6400 samples per second maximum which leads to a maximum sample rate of 160 samples per second per microphone which is well short of the 200-250Hz sample rate you were looking for.

Best Regards,

Rachael

noblemaster · Answer

I did like this... connected 1 Mic and read 40 times in a 5milisecond-loop for a duration of about 5seconds. I applied 2-3 thumb-press signals during recording. Printed 40 results on the same line and sent to PC with this huge baud after every 40 recorded samples. Initial thought was that if all goes well then all 40 recordings will be same.

Analysed 40 time series recordings and calculated correlation coefficients that were 0.99-0.999... So i feel that I've got pretty much success.

What do you say?

rachaelp · Answer

Most commercial mic pre circuits I believe have DC blocking caps so they are suitable for use with phantom powered mics. As such they have an inherent roll off at low frequencies so you'll be struggling to get a decent level through most sensibly priced ones at 5Hz. There are some exceedingly expensive options such as http://www.studiocare.com/summit-ecs-410-everest-channel-strip.html which are fairly flat down to 5Hz but for your 40 mics it'll set you back nearly £30k..... Given that your design uses an Arduino, I am thinking this is not the price budget you're aiming at.... I think you might be better off designing your own 40-input mic preamp (or a single one 40 times or something in between) which work for your specific requirements. Speaking of which, if you could expand on those with more details (i.e. a little more than "As much as possible") and maybe give some background for your specific application then people might be able to help you more easily. Best Regards, Rachael

WarrenW · Answer

Hi Noble. When calculating the low frequency cut off 3dB point you can do some simple maths of 1/(2 x Pi x Freq x Capacitor) to give the reactance and given the 10Kohm input on the 9816 module your reactance would need to be the same for a simple -3dB point (1/2 the amplitude). It's a simple voltage divider at that point. The 1uF gives about 33Kohm at 5hz. A 4.7uF would be approx. 6Kohm so should be able to amplify you 5 Hz with 'some' reduction in amplitude from the original waveform. DAB suggested superimposing a 1Khz 'carrier' and then filtering this which is a logical solution if you are going to MUX the inputs as the MUX rate will alter you LF waveforms. Now we take the MIC into account - whoa - it has a 3dB point of 6Hz! That will mean your levels will already be skewed. I hope you are not looking to analyse on amplitude and hope you are analysing freq only (i.e. presence of frequency not how loud). After about 10Hz things should flatten off. Hope this helps. Warren.

DAB · Answer

I am not sure the USB can handle the amount of data you can collect from 40 microphones. Are you going to do any preprocessing before you send the data, or just send up the raw data? DAB

rachaelp · Answer

I was thinking more along the lines of sending a known data set at rate which is utilizing a reasonable percentage of the link bandwidth, e.g. a counting value or a PRBS sequence incrementing at a set rate and send this over an extended period of time with something on the other end monitoring what it receives continuously and counting errors. I'd set this up and run it for hours or days to be sure something wasn't going to go wrong occasionally.

But at the end of the day it depends on what your application is doing and whether you can tolerate the occasional error once in a while and deal with / recover from it as to whether its worth the effort to validate the data link any further, it might not be so if you are happy with the validation you've done so far then that's great as then it looks like you should have enough bandwidth for what you are trying to do.

Best Regards,

Rachael

DAB · Answer

Hi Noble, Can you supply some additional information, like some description of your application. At the frequencies you are wanting to measure, you will find it very hard to get a good result and keep your expenses low. I would suggest using an intermediary frequency, say 1K Hz and then measure your low frequency component using software. Otherwise you will have a lot of issues with signal resolution and extraction. I have been looking at a design to record the low frequencies of thunder, so I know what it is like to assess low frequency components. DAB

noblemaster · Answer

I am interested in measuring vibrations in 5Hz to 100Hz band and then analysis in MATLAB.. Mic model ICS40300. Signal is good with DC offset at 0.8V and max output at around 1.2V. I am using Arduino DUE, it has 12 bit ADC. I have few options in mind... 1. https://www.sparkfun.com/products/9816 What if I change its input cap at both stages from 1uF to 4.7uF? 2. https://www.sparkfun.com/products/9868 What if I replace its microphone with my desired model and change its input cap to 27uF? Which option is better from above two? or something else is required? Thanks.

michaelkellett · Answer

If you are using a microphone then you pretty much HAVE to have an anti alias filter - if you don't then any sound at frequencies above half the sampling rate is converted into signals within the same band as the wanted signals. Unless your application has some feature that ensures unwanted sounds never reach the microphone you will need the filter. Can you tell us a bit more about the application.

The Sparkfun board could be modified as you suggest.

How far from the multiplexer will the microphones be ?

MK

noblemaster · Answer

I've tested Arduino Due, working good at 624000 Baud over USB. I hope it will be quite good for me. What do you say?

Using 40 low frequency Mics.

Top Replies