Alright, ready to join the dark side.

[JimRayden]

I need a general overview of digital sound processing and synthing.

I don't know where to start. A very convenient thing would be a nice big rewritable DSP chip so I could start experimenting with small programs at first.

This DSP stuff needs a beginner project.

-----------
Jimbo

[R.G.]

This DSP stuff needs a beginner project.

I thought that's what you were doing... 

Semi-seriously, a single chip that puts out a bass note in response to a pressed switch that can do bass accompaniment sounds like a GREAT beginner project. It's easy enough to do, needs no fancy hardware or software, and is immediately useful with essentially only the chip involved.

A very convenient thing would be a nice big rewritable DSP chip so I could start experimenting with small programs at first.

Yeah, yeah, we'd all love to have that. They're too expensive and involved just at the moment, so we're all nibbling at the edges with non-pure-DSP chips for the moment.

I need a general overview of digital sound processing and synthing.

That's not that hard. Read up the stuff posted here. There are some "intros to DSP" you can find through google.

A very basic overview is:
- to do digital signal processing,  you have to sample the data only at certain times
- if you sample often enough, you can reconstruct the results as though it were continuous data
- Nyquist's theorm defines "often enough" to be more than 2X the highest frequency you allow in.
- in practice, it's usually 3x or more the highest input frequency
- Anything over the Nyquist frequency gets aliased - it gets folded back into the sampled data range around the sampling frequency, so it sounds like SSB quacking ugliness
- bit resolution matters; if you sample data, you get a signal to noise ratio of S/N = 1.07 +6*b db, where b is the number of bits. Sample with two bits, you get 13db signal to noise possible. Sample with 12 and you get 73 possible and it starts to sound like audio.  The noise is the leftover quantization sounding like hash.

There are many other things involved, but those are the biggies that always come up. If you're sampling more than 3x the highest frequency involved, you can in general get back all the data you put in, except for noise. In general for hifi audio, you need "CD quality" - 16 bits, 44kHz. That's why "CD quality" was set that way.

Generation of sound is in some ways simpler than sampling, processing, and reproduction. For one thing, there's no "original sound" to judge it against.

[JimRayden]

What I was asking, was if DSPs or any other uCs were rewritable. This seems to be such a basic knowledge that I cannot seem to find any info on it. I don't feel like buying a new chip every time I write a test program. I also wish to know if DSPs usually have internal DAC or do I need an extra chip for that.

Hmm, here's the cheapest DSP chip at the local store: http://www.elfa.se/pdf/73/733/07333024.pdf

Tell me what you are able read from that datasheet. I'll take time to read it this weekend.

-----------
Jimbo

[Peter Snowberg]

Be sure to look at this project!  (this is bargain DSP using a microcontroller)

http://www.diystompboxes.com/smfforum/index.php?topic=43177.0

DSPs are really just another type of microprocessor which have special math capabilities. They run on software, just like the computer you are using now. The main addition a DSP brings to the table is the MAC or Multiply-ACcumulate instruction. This is a high speed multiply followed by an add instruction and it gives the capability to model the response of filters using the same math you would use to select the component values in the analog world.

There is no getting around the fact that DSP systems are complex. They are high performance computers. These days many DSPs have lots of basic elements like RAM and specialized peripherals built into a single chip. This makes system development much simpler and reduces system cost along with size and external complexity. By external complexity, I mean the number of devices that need to be integrated into the system and the interconnections between them.

I would suggest looking at the chips from http://www.wavefrontsemi.com which used to be the chip design department of Alesis. They sell a couple of very simple DSPs that are much less complex than a typical microprocessor but provide advanced processing capabilities.  I have a project that I will publish here in the future using these DSPs, but as R.G. said, it is better to crawl first and then walk.

Read everything at Wavefront about the 1K DSP and DRE. That will be a much better suited introduction to audio DSP hardware because that is what the Wavefront chips are made to do unlike the TI chip you posted a like to which may be used for a very wide variety of applications. That TI chips doesn't actually look very good at all for audio.

Also, be sure to look at the Axoris Miss Parker http://www.axoris.be/

Most of the time the converters are external to the DSP. There are some chips that contain DACs and some that contain both ADC and DAC but the converters are inexpensive and it pays to locate them in a different part of the board to reduce interference from the digital circuits.

[JimRayden]

Great links, great! The AL3101 looks like a sweety, though I'd prefer one in DIP package.

Too bad I'll soon be going out of town for two months and won't have the time to dig into it before mid August. I'll make sure to read this section alot before I leave and when I return.

Talk to ya later,
----------
Jimbo