Sound Recording & Playback (not MP3 or Voice)

Go To Last Post
12 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A new industrial app proposal requires recording and playback of "high quality" sound. The functionality might be very similar to answering machine prompt(s), or the animal sounds in a children's learning toy.

But the requirement is a higher-quality recording than voice or toy.

An example of the industrial app is to record the sound of a running device and use the sound when new/good with the current sound to see if it is old/bad. (That isn't exactly the app, but like that.)

The total time is like 30 seconds, in 6 or so "sound bites". The customer thinks 96ksps is needed, at 16-bits or so.

We are starting with a clean sheet of paper, but would like to use an AVR. :) The first cocktail-napkin sketch has an audio codec, the AVR, and a parallel Atmel flash with 1 or 2 megabytes. Streaming 16-bit samples @96kHz (~5us/byte) can maybe, just maybe, be done with a tightly-tuned loop. The storage time would be very, very close to the specs on the flash.

I think that 48kHz sampling with the AC'97 codecs might be enough, but sound ain't my normal app area. A search of the Projects section doesn't seem to uncover anything. Google will be my friend.

What should I be looking at--something other than an AVR? Does the audio codec proposal sound right? (They seem to be cheap, fast, and readily available.) Do I need a faster buffer like SRAM for the "sound byte", and then process and store to the permanent memory for the later playbacks?

Lee

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Whats the DVD audio specs in bits and sample rate? 20 bit and 96KHz? CDs are 16 bit and 44.1 KHz. A few self proclaimed golden ears say they can hear the diff, but I'd need to see it (the double blind experiment setup) and hear it to believe it. If they'd give in to 16 bit and 48KHz that would be easier to stuff in the flash rom. The specs on the cirrus codecs are good. This is mono, right? I guess you can throw away half the codec output?

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

CDs are 16 bits and 44.1kHz. WAV files (RIFF format) support mono or stereo, 8 through 16 bits at 11, 22 or 44kHz, with the vital parameters conveniently passed in the file header. I've had surprisingly good results just feeding WAV files to an 8-bit ADC. You wouldn't think it would be particularly high quality at 8 bits, but for "industrial" applications (gaming machines, in my case) the governing factors are the speaker quality and the environment. The sampling rate makes a difference and I ended up using 22kHz, but without hi-fi speakers in good enclosures, and with public-area levels of background noise, you just plain can't hear the low half of the dynamic range anyway and I doubt anyone could tell the difference between (say) 12 and 16 bits. I'd try something simple and cheap first before going to a lot of trouble with codecs and such.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I would get samples of the old and new machine with a 44K sample rate 16 bit using a windoze machine. Then run them through FFT for comparison. I suspect that the client's idea of the required quality is excessive.

Ralph Hilton

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:

This is mono, right? I guess you can throw away half the codec output?

Mono is the RFQ, but stereo is mentioned in that it may be advantageous for the app. (Kind of a new/proposed application area, so the results of the "research" won't be known till the "development" is done. :) As all mentioned, some A/B comparisons with stereo/mono, sample rate, & bit depth will determine the final direction.

Lee

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I agree with peret.... you need more dynamic range for 'live'.. after its recorded, you can make it fit in 8 bits (really +-7... only 42 dB... not very good unless the track is really 'packed')

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

As an alternative I would suggest looking at some cheep, easy to use DSP like Blackfin from AD. There are many application notes with these chips and examples to get you going quickly. SRAM may not be necessary, but it depends on your application.

I agree with Bob, above 44.1Khz the difference in quality becomes very small (I think the defference between 44.1 and 96 is still large enough for a engineer to hear, the golden ears guys are on a completely different plane , I am a little less sceptical on thier abilities having worked with a few.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I agree with Bob, above 44.1Khz the difference in quality becomes very small (I think the defference between 44.1 and 96 is still large enough for a engineer to hear, the golden ears guys are on a completely different plane , I am a little less sceptical on thier abilities having worked with a few.
=====================
Hi Michael. Can you recall what source material and listening conditions were present when you heard the golden ears go 'ewwww... that sounds crappy.... mustbe 16bit 44khz!'? I hate to be such a skeptic, but I'd love to see the experimental setup, and I'd love to see the actual experiment and see the published results. Can you think of test conditions that would reveal CD deficiencies?

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi Bob,
I'm in the office , but I will put something together a bit later. This (as you know) is a much debated subject, so to do credit to your question (which is a good one) I actually deleted me reply I was going to write, as although it talks about quantization differences between 16 bit and 24 bit and so on, in the end doesn't take away the subjective point of view. I've read and seen many articles, which can be sceintifically reproduced. I'll just need a bit of time to dig around for them.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I hate to mention this but..... there is a dSPIC....which is a PIC whith lots of DSP features and NO paged memory!! I just got an offer for a development kit at 75% discount, perhaps they are not selling as well as they anticipated.

John Samperi

Ampertronics Pty. Ltd.

https://www.ampertronics.com.au

* Electronic Design * Custom Products * Contract Assembly

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Lee, don't forget that you'll need $500 Shure microphones and $1000 speakers, to take advantage of all that audio quality ! ;-)

/Jesper
http://www.yampp.com
The quick black AVR jumped over the lazy PIC.
What boots up, must come down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi Bob,
back again. Before I post some more technicahal references, better to have first hand experience. This is from discussions I had today wth a pair of "golden ears" acomplished musician, producer and worked for pro audio companies such as digidesign.

The diference in sample rate 44.1 to 96 is a phenomena that in a studio environment, multi track recordings and so on, (with top of the line equipment) is easily heard and does like you suggest give the Uggh feeling (to him, less talented recording engineers I know would say if it is a test they are aware of they can pick out the difference but otherwise not, so it really is subtle). Because, with so many elements in a recording chain, so many things can and do go wrong. The sample rate or incorret clocking is not uncommon and the phenomena presents itself like this.

For 44.1 to 96Khz from a listening point of view , the stereo imagary begins to be lost in particular with regard to the azimuth of the higher frequency components. This is the key he uses for identifying this problem.

"Golden ears' should not be thought of as someone who can just hear the difference, but someone who can explain and direct engineers to reproduce and evaluate what they are hearing .

But that aside, even if there is no audiable diference, doesn't mean that there is no difference. This might sound absurd. But think about it. It all depends on the application. For a live recording, reproducing exactly as heard is the criteria. Equipment is designed with distortion of 0.001%, would a person hear a difference if it was 0.01% ? of course not. The point is, it is part of a chain, each element is aimed at reproducing exactly what is heard.

So it is with sampling, mabee you can't hear the difference between 44.1 and 96. However , process the sound by DSP or something and while the 96KHz signal remains clear the 44.1 starts to have audible distortion.

For more detailed reading on how sample rate please check out the following ...

David Griesinger's home page http://world.std.com/~griesngr/

Discussion on the effects of sine/square wave sampling http://www.smr-home-theatre.org/surround2002/technology/page_07.shtml

and some way out stuff http://jn.physiology.org/cgi/content/full/83/6/3548