-> Knowles SPU0410LR5H-QB -> ? -> FPGA

[Mr_Cheerful]

Hi there

May I kindly ask for your opinion on what the ADC (maybe also: pre-amp) of your choice would be to provide an FPGA with a (relatively) noise-free input from a Knowles SiSonic SPU0410LR5H-QB

output signal range: 20-85kHz
ADC sampling rate: >=256kSPS -1MSPS
ADC Output: PCM and or I2S

I would like to acquire, buffer, modify and output digitalised sound (as a stream or on SD-card) with an intel FPGA.
the bus between the FPGA and the ADC may have up to 16 signal paths.

I am also interested in an ADC I can easily use with my (amd64, Linux) workstation or even raspberry pi (clone) to ensure I get that part of the project right.

I will share all documents on Github and I will also share the source code for the FPGA (license tbd, probably MIT).

Currently I‘m considering the AD7768 but I‘m not quite sure.

It isn’t my first FPGA rodeo, I am a retired automation engineer who has always had time for tinkering.

It isn’t my first dabbling in analog circuits either, but this time I‘m having a harder time.

Thank you for taking notice!

 

[AnalogSteph]

1. That particular model MEMS mic does not seem to be of the "ultrasonic boosted" variety that they also seem to offer, according to the SiSonic Design Guide. Intentional?
2. You may be finding more options when opening yourself up to typical audio sample rates, notably 384 and 768 kHz. (Maybe even 192 / 216 kHz if filter performance is good.)
3. How many mics are you planning on using? Is this going to be a "bunch of MEMS mics on a sphere" kind of job? You're talking of up to 16 signal paths and an 8-channel ADC, that's why I'm asking.
4. There are a number of ADCs/codecs with built-in mic preamps and mic bias supply but I doubt very many of them have digital filters that will do well with ultrasonic signals, and even support for 192 kHz is not a given. You may have to do this the hard way and use an external preamp circuit for an ADC of your choosing. You'll need something like 30-36 dB worth of total gain, I'd go for two stages.
5. AD7768 would get the job done when provided with a preamp stage, I can't help shake the feeling that it would be a bit overdressed though. There's no way in hell you'll need a -120 dB THD, and even if noise level above 20 kHz is less than below, a total dynamic range of 87 dB(A) with a major following wind (mic limit levels are traditionally spec'd at 1% or even 0.5% THD, not 10%) does not suggest the need for a truly top-flight converter. Mind you, there is the issue of out-of-band noise, which is not necessarily well-documented.
6. Digital filter performance is going to be a major concern when dealing with ultrasonic signals. Look at the filter performance of some older audio-centric ADCs at 192 kHz (e.g. CS5361/81) and you'll see what I mean. That doesn't mean all older ADCs have this issue, e.g. PCM4220/4222 don't. Still, I would check out more recent designs first, e.g. from AKM and ESS.

 

[Mr_Cheerful]

1. That particular model MEMS mic does not seem

to be of the "ultrasonic boosted" variety that they also seem to offer, according to the SiSonic Design Guide. Intentional?
2. You may be finding more options when opening yourself up to typical audio sample rates, notably 384 and 768 kHz. (Maybe even 192 / 216 kHz if filter performance is good.)
3. How many mics are you planning on using? Is this going to be a "bunch of MEMS mics on a sphere" kind of job? You're talking of up to 16 signal paths and an 8-channel ADC, that's why I'm asking.
4. There are a number of ADCs/codecs with built-in mic preamps and mic bias supply but I doubt very many of them have digital filters that will do well with ultrasonic signals, and even support for 192 kHz is not a given. You may have to do this the hard way and use an external preamp circuit for an ADC of your choosing. You'll need something like 30-36 dB worth of total gain, I'd go for two stages.
5. AD7768 would get the job done when provided with a preamp stage, I can't help shake the feeling that it would be a bit overdressed though. There's no way in hell you'll need a -120 dB THD, and even if noise level above 20 kHz is less than below, a total dynamic range of 87 dB(A) with a major following wind (mic limit levels are traditionally spec'd at 1% or even 0.5% THD, not 10%) does not suggest the need for a truly top-flight converter. Mind you, there is the issue of out-of-band noise, which is not necessarily well-documented.
6. Digital filter performance is going to be a major concern when dealing with ultrasonic signals. Look at the filter performance of some older audio-centric ADCs at 192 kHz (e.g. CS5361/81) and you'll see what I mean. That doesn't mean all older ADCs have this issue, e.g. PCM4220/4222 don't. Still, I would check out more recent designs first, e.g. from AKM and ESS.

Thank you for your prompt and well structured response

1: The model is the one I have on a break-out board
2: Ideally a half-sphere of microphones on a tbd carrier. Maybe I need to reconsider my options as they do sell a model that has a digital output. I‘ve seen a few projects using this specific type for all sorts of applications.

Requirements:

Must have only microphones used between 20 and 100kHz.
Sample rate must be over 300kHz.
Enhanced noise filtering shall be implemented in ADC and/or FPGA.
An FPGA is used for collating records of the individual detections and the context to them.
Must be stand-alone in outside conditions in Central Europe.
Must save recordings and generated metadata (i.e. time; class; species; direction; duration.
May be mesh-networked, hard encrypted.
May have at least one node that manages other nodes.
Nodes must be light and rugged.
Shall cost up to 200 USD/light node.

My S.O. can‘t hear them. I can just barely hear one of two of the species that live here.
I can‘t cough up the cash for a professional device but already have a few parts in a bin and capable suppliers and time too.