Michrophone amp with D/A for continuous speech recognition

Started by Richard Owlett June 27, 2017
My theory dates from 3 years towards a BSEE in the 60's.
Real world experience focused on instrumentation ( 1 mV large signal, 
10Hz very wide band ;)
[Ever try to convince a fresh chemistry post doc that demanding a system 
gain-bandwidth product allowing measuring less than 1 electron per 
second wasn't reasonable?]

My voice recognition background is limited to reading end user oriented 
literature a decade ago. At that time cost/benefit was unacceptable.

Once again I'm interested. I have decided to split the system into two 
black boxes.

Box 1 - Microphone in, clean "hifi" digitized audio out
Box 2 - The rest of the system ;/

This post is about Box 1.

Part of the reasoning behind my specification for Box 1 is to some blue 
sky experimentation with filters etc.

The overall transfer function will yield a minimum of 40 thousand 16 bit 
(15 bits + sign) samples per second. Internally I assume significantly 
higher numbers. I explicitly am leaving the latency specification WIDE 
open. This black box may include a dedicated laptop computer for 
required processing power.

Comments on availability of consumer grade and price components I should 
investigate? The goal being to decide what price/performance trade-offs 
I wish to make.

Isn't retirement for doing what you couldn't?

Richard Owlett wrote on 6/27/2017 5:54 PM:
> My theory dates from 3 years towards a BSEE in the 60's. > Real world experience focused on instrumentation ( 1 mV large signal, 10Hz > very wide band ;) > [Ever try to convince a fresh chemistry post doc that demanding a system > gain-bandwidth product allowing measuring less than 1 electron per second > wasn't reasonable?] > > My voice recognition background is limited to reading end user oriented > literature a decade ago. At that time cost/benefit was unacceptable. > > Once again I'm interested. I have decided to split the system into two black > boxes. > > Box 1 - Microphone in, clean "hifi" digitized audio out > Box 2 - The rest of the system ;/ > > This post is about Box 1. > > Part of the reasoning behind my specification for Box 1 is to some blue sky > experimentation with filters etc. > > The overall transfer function will yield a minimum of 40 thousand 16 bit (15 > bits + sign) samples per second. Internally I assume significantly higher > numbers. I explicitly am leaving the latency specification WIDE open. This > black box may include a dedicated laptop computer for required processing > power. > > Comments on availability of consumer grade and price components I should > investigate? The goal being to decide what price/performance trade-offs I > wish to make. > > Isn't retirement for doing what you couldn't?
I don't recall the details, but I believe they make a chip that contains a microphone and produces the digital signal you are asking for. It might not be quite as wide band as 20 kHz, but rather intended for voice in phones. But you might sniff around a bit for chip level devices. The incentive for such inventions is the smallness of the package, important in phones. It has potential for high quality audio as well. -- Rick C
On 06/27/2017 06:30 PM, rickman wrote:
> Richard Owlett wrote on 6/27/2017 5:54 PM: >> My theory dates from 3 years towards a BSEE in the 60's. >> Real world experience focused on instrumentation ( 1 mV large signal, >> 10Hz >> very wide band ;) >> [Ever try to convince a fresh chemistry post doc that demanding a system >> gain-bandwidth product allowing measuring less than 1 electron per second >> wasn't reasonable?] >> >> My voice recognition background is limited to reading end user oriented >> literature a decade ago. At that time cost/benefit was unacceptable. >> >> Once again I'm interested. I have decided to split the system into two >> black >> boxes. >> >> Box 1 - Microphone in, clean "hifi" digitized audio out >> Box 2 - The rest of the system ;/ >> >> This post is about Box 1. >> >> Part of the reasoning behind my specification for Box 1 is to some >> blue sky >> experimentation with filters etc. >> >> The overall transfer function will yield a minimum of 40 thousand 16 >> bit (15 >> bits + sign) samples per second. Internally I assume significantly higher >> numbers. I explicitly am leaving the latency specification WIDE open. >> This >> black box may include a dedicated laptop computer for required processing >> power. >> >> Comments on availability of consumer grade and price components I should >> investigate? The goal being to decide what price/performance trade-offs I >> wish to make. >> >> Isn't retirement for doing what you couldn't? > > I don't recall the details, but I believe they make a chip that contains > a microphone and produces the digital signal you are asking for. It > might not be quite as wide band as 20 kHz, but rather intended for voice > in phones. But you might sniff around a bit for chip level devices. The > incentive for such inventions is the smallness of the package, important > in phones. It has potential for high quality audio as well. >
Sounds interesting. As a practical matter I'll need a board level product as I haven't done any component level soldering since the 8085 and 7400 series logic were current technology. Any suggested key words for searches? I assume digikey.com would be one place to search for board level product. They feature an appropriate microphone on their homepage. Haven't checked yet for a board. Anywhere else catering to the experimenter market? Thank you.
> ... Comments on availability of consumer grade and price components I should
investigate... If you can use USB, there are many affordable quality USB mic-preamp/audio-interface devices available today. As it happens, I just received a Behringer UMC404HD four-channel, 44100-192000 sample rate, 16- or 24-bit sample for my microphone development work, a very versatile device; it cost US$90 shipped. https://www.sweetwater.com/store/detail/UMC404HD A two-channel 192kHz/24-bit device is also available. https://www.sweetwater.com/store/detail/UMC204HD Another device that might suit is the Art USBDualPre, a two-channel device that only does 44100/48000 16-bit, for ~$75. https://www.sweetwater.com/store/detail/USBDualPrePS If you want spend more, look at Sound Devices. You might consider the USBPre2, for ~$900. https://www.sounddevices.com/products/portable-audio-tools/usbpre2
On 06/27/2017 07:16 PM, Richard Owlett wrote:
> On 06/27/2017 06:30 PM, rickman wrote: >> Richard Owlett wrote on 6/27/2017 5:54 PM: >>> My theory dates from 3 years towards a BSEE in the 60's. >>> Real world experience focused on instrumentation ( 1 mV large signal, >>> 10Hz >>> very wide band ;) >>> [Ever try to convince a fresh chemistry post doc that demanding a system >>> gain-bandwidth product allowing measuring less than 1 electron per >>> second >>> wasn't reasonable?] >>> >>> My voice recognition background is limited to reading end user oriented >>> literature a decade ago. At that time cost/benefit was unacceptable. >>> >>> Once again I'm interested. I have decided to split the system into two >>> black >>> boxes. >>> >>> Box 1 - Microphone in, clean "hifi" digitized audio out >>> Box 2 - The rest of the system ;/ >>> >>> This post is about Box 1. >>> >>> Part of the reasoning behind my specification for Box 1 is to some >>> blue sky >>> experimentation with filters etc. >>> >>> The overall transfer function will yield a minimum of 40 thousand 16 >>> bit (15 >>> bits + sign) samples per second. Internally I assume significantly >>> higher >>> numbers. I explicitly am leaving the latency specification WIDE open. >>> This >>> black box may include a dedicated laptop computer for required >>> processing >>> power. >>> >>> Comments on availability of consumer grade and price components I should >>> investigate? The goal being to decide what price/performance >>> trade-offs I >>> wish to make. >>> >>> Isn't retirement for doing what you couldn't? >> >> I don't recall the details, but I believe they make a chip that contains >> a microphone and produces the digital signal you are asking for. It >> might not be quite as wide band as 20 kHz, but rather intended for voice >> in phones. But you might sniff around a bit for chip level devices. The >> incentive for such inventions is the smallness of the package, important >> in phones. It has potential for high quality audio as well. >> > > Sounds interesting. As a practical matter I'll need a board level > product as I haven't done any component level soldering since the 8085 > and 7400 series logic were current technology. > > Any suggested key words for searches? > I assume digikey.com would be one place to search for board level > product. They feature an appropriate microphone on their homepage. > Haven't checked yet for a board. Anywhere else catering to the > experimenter market? > > Thank you. > >
My search just reminded me of sparkfun.com . Will have to browse their site more carefully tomorrow when I'm awake.
Richard Owlett wrote on 6/27/2017 8:16 PM:
> On 06/27/2017 06:30 PM, rickman wrote: >> Richard Owlett wrote on 6/27/2017 5:54 PM: >>> My theory dates from 3 years towards a BSEE in the 60's. >>> Real world experience focused on instrumentation ( 1 mV large signal, >>> 10Hz >>> very wide band ;) >>> [Ever try to convince a fresh chemistry post doc that demanding a system >>> gain-bandwidth product allowing measuring less than 1 electron per second >>> wasn't reasonable?] >>> >>> My voice recognition background is limited to reading end user oriented >>> literature a decade ago. At that time cost/benefit was unacceptable. >>> >>> Once again I'm interested. I have decided to split the system into two >>> black >>> boxes. >>> >>> Box 1 - Microphone in, clean "hifi" digitized audio out >>> Box 2 - The rest of the system ;/ >>> >>> This post is about Box 1. >>> >>> Part of the reasoning behind my specification for Box 1 is to some >>> blue sky >>> experimentation with filters etc. >>> >>> The overall transfer function will yield a minimum of 40 thousand 16 >>> bit (15 >>> bits + sign) samples per second. Internally I assume significantly higher >>> numbers. I explicitly am leaving the latency specification WIDE open. >>> This >>> black box may include a dedicated laptop computer for required processing >>> power. >>> >>> Comments on availability of consumer grade and price components I should >>> investigate? The goal being to decide what price/performance trade-offs I >>> wish to make. >>> >>> Isn't retirement for doing what you couldn't? >> >> I don't recall the details, but I believe they make a chip that contains >> a microphone and produces the digital signal you are asking for. It >> might not be quite as wide band as 20 kHz, but rather intended for voice >> in phones. But you might sniff around a bit for chip level devices. The >> incentive for such inventions is the smallness of the package, important >> in phones. It has potential for high quality audio as well. >> > > Sounds interesting. As a practical matter I'll need a board level product as > I haven't done any component level soldering since the 8085 and 7400 series > logic were current technology. > > Any suggested key words for searches? > I assume digikey.com would be one place to search for board level product. > They feature an appropriate microphone on their homepage. Haven't checked > yet for a board. Anywhere else catering to the experimenter market?
You can start with Digikey, but I think I would start with TI, they are big into CODECs and I think the mic I saw some time back was theirs. Either way I would look for the chip and then once I found a chip find the eval boards they offer. Starting with the board may be a little harder depending on what name they use for it. "digital microphone" would be a good search term to start with. -- Rick C
On 06/27/2017 08:39 PM, Tom Becker wrote:
>> ... Comments on availability of consumer grade and price components I should
investigate...
> > If you can use USB, there are many affordable quality USB
mic-preamp/audio-interface devices available today. USB was one of the extras I was wanting ;}
> > As it happens, I just received a Behringer UMC404HD four-channel, 44100-192000
sample rate, 16- or 24-bit sample for my microphone development work, a very versatile device; it cost US$90 shipped. https://www.sweetwater.com/store/detail/UMC404HD A two-channel 192kHz/24-bit device is also available. https://www.sweetwater.com/store/detail/UMC204HD
> > Another device that might suit is the Art USBDualPre, a two-channel device that
only does 44100/48000 16-bit, for ~$75. https://www.sweetwater.com/store/detail/USBDualPrePS Both of those are in the price range I was thinking of. I'll check specifications in the morning when I am functionally awake.
> > If you want spend more, look at Sound Devices. You might consider the USBPre2,
for ~$900. https://www.sounddevices.com/products/portable-audio-tools/usbpre2
>
That is too much. I'd like to bring the whole thing in for no more than half that. Thank you.
On 6/28/2017 3:35, Richard Owlett wrote:
> On 06/27/2017 08:39 PM, Tom Becker wrote: >>> ... Comments on availability of consumer grade and price components I >>> should investigate... >> >> If you can use USB, there are many affordable quality USB >> mic-preamp/audio-interface devices available today. > > USB was one of the extras I was wanting ;} > >> >> As it happens, I just received a Behringer UMC404HD four-channel, >> 44100-192000 sample rate, 16- or 24-bit sample for my microphone >> development work, a very versatile device; it cost US$90 shipped. >> https://www.sweetwater.com/store/detail/UMC404HD A two-channel >> 192kHz/24-bit device is also available. >> https://www.sweetwater.com/store/detail/UMC204HD >> >> Another device that might suit is the Art USBDualPre, a two-channel >> device that only does 44100/48000 16-bit, for ~$75. >> https://www.sweetwater.com/store/detail/USBDualPrePS > > Both of those are in the price range I was thinking of. > I'll check specifications in the morning when I am functionally awake. > >> >> If you want spend more, look at Sound Devices. You might consider the >> USBPre2, for ~$900. >> https://www.sounddevices.com/products/portable-audio-tools/usbpre2 >> > > That is too much. I'd like to bring the whole thing in for no more than > half that. > > Thank you.
About 10 years ago whwn I was working with a small company on acoustic sensors., we used Knowles microphones http://www.knowles.com/eng/Products/Microphones The ones we used were pretty flat from 100 Hz to 10 KHz. They were coming out with some digital microphones, but I didn't have a chance to play with them. -- Best wishes, --Phil pomartel At Comcast(ignore_this) dot net
On 06/28/2017 07:07 AM, Phil Martel wrote:
> On 6/28/2017 3:35, Richard Owlett wrote: >> On 06/27/2017 08:39 PM, Tom Becker wrote: >>>> ... Comments on availability of consumer grade and price components I >>>> should investigate... >>> >>> If you can use USB, there are many affordable quality USB >>> mic-preamp/audio-interface devices available today. >> >> USB was one of the extras I was wanting ;} >> >>> >>> As it happens, I just received a Behringer UMC404HD four-channel, >>> 44100-192000 sample rate, 16- or 24-bit sample for my microphone >>> development work, a very versatile device; it cost US$90 shipped. >>> https://www.sweetwater.com/store/detail/UMC404HD A two-channel >>> 192kHz/24-bit device is also available. >>> https://www.sweetwater.com/store/detail/UMC204HD >>> >>> Another device that might suit is the Art USBDualPre, a two-channel >>> device that only does 44100/48000 16-bit, for ~$75. >>> https://www.sweetwater.com/store/detail/USBDualPrePS >> >> Both of those are in the price range I was thinking of. >> I'll check specifications in the morning when I am functionally awake. >> >>> >>> If you want spend more, look at Sound Devices. You might consider the >>> USBPre2, for ~$900. >>> https://www.sounddevices.com/products/portable-audio-tools/usbpre2 >>> >> >> That is too much. I'd like to bring the whole thing in for no more than >> half that. >> >> Thank you. > > About 10 years ago whwn I was working with a small company on acoustic > sensors., we used Knowles microphones > http://www.knowles.com/eng/Products/Microphones > The ones we used were pretty flat from 100 Hz to 10 KHz. > They were coming out with some digital microphones, but I didn't have a > chance to play with them. >
Their site's search function turns up no hits for "usb". Now that I know mikes with USB connectivity are readily available, that is now one of my requirements.
On 06/27/2017 08:39 PM, Tom Becker wrote:
>> ... Comments on availability of consumer grade and price components I should
investigate...
> > If you can use USB, there are many affordable quality USB
mic-preamp/audio-interface devices available today.
> > As it happens, I just received a Behringer UMC404HD four-channel, 44100-192000
sample rate, 16- or 24-bit sample for my microphone development work, a very versatile device; it cost US$90 shipped. https://www.sweetwater.com/store/detail/UMC404HD A two-channel 192kHz/24-bit device is also available. https://www.sweetwater.com/store/detail/UMC204HD
> > Another device that might suit is the Art USBDualPre, a two-channel device that
only does 44100/48000 16-bit, for ~$75. https://www.sweetwater.com/store/detail/USBDualPrePS
> > If you want spend more, look at Sound Devices. You might consider the USBPre2,
for ~$900. https://www.sounddevices.com/products/portable-audio-tools/usbpre2
>
The sweetwater.com is going to provide me with a least a week's worth of reading material that should bring me into the current century. It also demonstrates that the universe of possible trade-offs is much different than I was thinking. Right now I think my goal would be something with the form factor of the Art USBDualPre but the electrical specification of the UMC204HD. I think I'll have to see what the current state of Linux voice recognition system is and what input they expect/make effective use of. I assume I'm just as far behind the times there also.