1
\$\begingroup\$

I am building a 16 channel multiplexer which will deliver the input from 16 microphones to a single ADC. I am clocking the ADC at 1.6 MHz and therefore sampling each audio channel at 100 kHz.

The ADC resolution is 12 bits, I cannot change the ADC because I am modifying a design, not starting from scratch. The ADC range is 2V. This is the microphone I am using:

http://www.farnell.com/datasheets/97498.pdf

I am realising this using 5 LMH6574 4:1 multiplexer chips, 4 used for the 16 mic lines then the fifth for the 4 outputs to produce a single output - The select lines for the first four are driven from a 2 bit binary up counter - a second up counter runs the 5th chip at CLK/4.

My question is this: Is it better to use an audio pre-amplifier for each channel - before being put through the multiplexer - or use a single pre-amplifier after the multiplexer stage?

I assume that if a single pre-amplifier is used it must have a bandwidth of at least 1.6 MHz, is this correct? - Will a high frequency pre-amp cause a loss in quality at the audio frequencies (compared with a hifi pre-amp chip)?

I also assume that the settling time of this amplifier is the all important factor, are there any rules/common practices defining what percentage I should be within (2%,1% 0.01% etc)?

\$\endgroup\$
  • \$\begingroup\$ What is the maximum un-amplified instantaneous signal level from your microphone AND what is the DC input range of your ADC? Also, what is a "high frequency preamp" and what is a "hifi preamp"? \$\endgroup\$ – Andy aka Apr 22 '15 at 10:43
  • \$\begingroup\$ By "High frequency pre-amp" I just meant "higher that audio frequency" By hi-fi pre=amp" I meant "audio frequency" \$\endgroup\$ – Tim Mottram Apr 22 '15 at 10:58
  • \$\begingroup\$ The data sheet requires fonts that I don't have - this still leaves me pondering about the peak signal level of the microphone and your need for an amplifier. \$\endgroup\$ – Andy aka Apr 22 '15 at 11:41
  • \$\begingroup\$ Also, what is the ADC you are proposing to use? \$\endgroup\$ – Andy aka Apr 22 '15 at 11:49
  • 1
    \$\begingroup\$ For what you are asking, the 1.6MHz ADC clock rate is not the important number. The 100KHz sampling rate is. That give a Nyquist frequency of 50KHz. Therefore signal components above 50KHz are detriments and would distort the ADC samples by aliasing. You mentioned metal striking metal, I don't know the signal profile that would produce after the mic, but it could have >50KHz components. So not only you don't want a "high frequency" amp, you may want to add a low pass filter at 20KHz that drops off fast enough to meet your noise requirement at 50KHz. \$\endgroup\$ – rioraxe Apr 23 '15 at 4:29
2
\$\begingroup\$

The range of microphones you are using appear to have a sensitivity of about -42 dB/Pa. This means they produce a nominal 8mV RMS for an RMS input sinewave pressure of 1 pascal. This pressure (1 pascal) is equivalent to 94 dB SPL (sound pressure level).

If you are measuring ambient sounds or even music, the peak/RMS level ratio (aka crest factor) can be as high as 20dB and this pushes the peak-to peak voltage output from the microphone up to 160mV from 8mV.

If you are measuring sound levels at up to 20dB higher than 94dB SPL, you might see an output from your microphone that is 1.6Vp-p.

You have a 2V range on your ADC so I'm questioning whether you need pre-amplifiers at all.

\$\endgroup\$
  • 1
    \$\begingroup\$ 94dB SPL is REALLY LOUD. \$\endgroup\$ – pjc50 Apr 22 '15 at 12:21
  • 1
    \$\begingroup\$ @pjc50 that's a rather subjective statement to make given that SPL deteriorates with distance so markedly. We do not know what the OP is measuring and clearly (from comments he has made) he's not that familiar with microphone specifications. 100dB could be 1m from a speaker in a disco and 90dB could be the sound levels on a bus. The op could be measuring a sound source at only an inch away. I'm trying to demonstrate that a pre-amplifier may not be needed at all. \$\endgroup\$ – Andy aka Apr 22 '15 at 12:26
  • \$\begingroup\$ Thanks Andy, that's some really good information on microphone specs! As regards what I am measuring, I am striking a piece of metal with a small metal head, however the microphone will be very close indeed, only a few mm away. However, I have no clue what the dB SPL will be at that point, I suppose I should test it and base my decision on that. I think for now, seeing as cost isn't really much of an issue, Ill just use 16 audio frequency pre amps which have a gain setting, I saw a few that went from 1 to 60 dB, then I know I'm safe. \$\endgroup\$ – Tim Mottram Apr 22 '15 at 12:49
  • \$\begingroup\$ @TimMottram I'd also be concious of the microphone you have chosen - most electret microphone suppliers are very non-committal on what the peak pressure the microphone can work with before clipping. On the other hand MEMs microphones (also quite cheap) seem to have a more trustworthy spec. Check-out ADI for MEMs microphones. \$\endgroup\$ – Andy aka Apr 22 '15 at 13:14
  • \$\begingroup\$ Thank you for the detailed answer, but I'm curious if you could edit to describe how you got from -42db/Pa to 8mVRMS, and how the crest factor of a waveform ties into the p-p output from nominal? Understanding microphone specs is one of those areas of electronics that I feel is overwhelmingly overlooked and misunderstood, and I (and others I'm sure) would appreciate the opportunity at learning it from someone such as yourself. \$\endgroup\$ – akohlsmith Jun 28 '15 at 23:07
1
\$\begingroup\$

I imagine the solution will involve leg work and compromises, possibly trial and error as well. Cost, complexity, audio quality, and so on. Also, PCB design and shielding will play a big part at these frequencies.

You have not said what kind of quality you expect from the microphones, nor the application, nor what range of amplitude you expect the microphone signals to be, nor the specs of your ADC (number of bits etc).

I suggest using an ADC that has a built-in 16+ channel multiplexer, it will make your life a lot easier. Or alternatively, separate ADCs. Digital is the new analogue; it solves a lot of problems.

If going for a separate analogue mux and ADC, I would consider a variable-gain amp configured as a compressor for each, to get the most resolution out of the ADC for any given signal. You can then implement an expander at the other end if necessary.

But no, I don't think you could use any hi-fi amp for a 1.6Mhz signal! As you say, it needs good bandwidth, which the hi-fi pre-amps would not have. Although you could use 16 of them as pre-amps. But at that point I'd ask myself if there was a better way.

\$\endgroup\$
  • \$\begingroup\$ Ok, thank you, unfortunately I can't change the ADC as I am modifying an existing design. The signal itself is not 1.6 MHz rather 10 ~ 20 kHz signals which I am switching. I have found amplifiers which have the required bandwidth, but I was wondering if it was better to have a separate amplifier for each stage. If I have a separate amp for each stage the individual amplifiers will always be following the voltage of each respective microphone, using one amplifier means constant switching between them, hence the question about settling time. \$\endgroup\$ – Tim Mottram Apr 22 '15 at 11:15
  • 1
    \$\begingroup\$ It sounds like you're answering your own question? You are correct, the settling time is a factor. If you use a single pre-amp, you will need to get the timing right with the ADC so the settling time of both the ADC and amp are optimal. The lower the settling time of the amplifier, the less critical the timing, assuming the ADC and mux settling times do not become a limiting factor as well. Try doing a simple worst-case calculation for samples, e.g. high, low, high, low, using the specs of the ADC, mux and amp to give an idea of the amount of cross talk. \$\endgroup\$ – CL22 Apr 22 '15 at 11:21
1
\$\begingroup\$

You acquire two lots of extra noise. If these are electret microphones with no internal pre-amp, it almost certainly will result in terrible quality unless you have a pre-amp channel per microphone. You're really much better off trying to find an integrated solution such as pre-amp+ADC in one package (eg https://www.fairchildsemi.com/datasheets/FA/FAN3850A.pdf , although that's not available in sensible packages) and then doing the multiplexing in the digital domain. Sticking to the existing ADC regardless will almost certainly give you a larger, more expensive, and lower quality solution.

\$\endgroup\$
  • \$\begingroup\$ Just to confirm, do you mean using 16 of these devices and multiplexing in there outputs (digital) ? \$\endgroup\$ – Tim Mottram Apr 22 '15 at 11:25
  • \$\begingroup\$ Not necessarily this device, but something similar (preamp with integrated ADC) and then yes multiplexing the digital outputs. \$\endgroup\$ – pjc50 Apr 22 '15 at 12:22
  • \$\begingroup\$ OK, thank you, looking at that device I can see its output is PDM, which I don't think I know how to multiplex, its breaking my brain! But I will look in to the devices you mentioned. \$\endgroup\$ – Tim Mottram Apr 22 '15 at 12:44

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.