[linux-audio-dev] Open firewire audio interface: A back-of-an-envelope prototype plan

[Simon Jenkins]

  Marcus Andersson wrote:

> Hi,
>
> what you need to figure out is how hardware can be developed by people 
> living in different countries, the same way email/sourceforge/CVS 
> pretty much solved the distribution problem for software. Here is just 
> a couple of ideas.
>
> All interested developers buy a prototype card each. You manufacture 
> it and then mail the cards to the developers.
>
> You set up a PC with remote login and give each developer an account. 
> Then hook up the prototype to the PC with firewire, JTAG, software 
> controllable power supply (GPIO?), webcam and other things that makes 
> it possible to test the card remotely.

I think it would have to be a prototype card each.

The development bottleneck would be making sure the first, physical 
prototype was basically correct. Someone would have to sit down at a 
bench with it and make sure the power supply was OK, get the dsPIC 
going, get it programming the FPGA,  get some communication with the 
1394 link going and get some basic end-to-end dsPIC to host over 1394 
transfer going. Anyone wishing to collaborate at this stage would 
probably want to arrange to be in the same room while it was happening 
although crude telepresence (webcam/irc/voice) might be worthwhile if 
things got really stuck.

Before the first prototype all collaboration would be electronic. Its 
just as easy to share a schematic across oceans as it is a body of code. 
After the first prototype a small batch could be made up and posted to 
interested developers and then its back to electronic collaboration again.

The developer prototypes would probably have to be self-funded although 
it might also be a good time to see if a bit of money could be sourced 
from somewhere. Another per-developer expense here would be a 
debugger/programmer for the dsPIC, probably the MPLAB ICD2, currently @ 
125UKP. The IDE is free but there's a question mark over the s/w 
toolchain. (Compiler/assembler/linker are gnu with source available, but 
the toolchain as a whole costs money. AFAIK its something to do with the 
C library being commercial).

Oh, and there's A/Ds and D/As to consider. Probably ignore them at first 
and just stick a scope on the audio interface header, but sooner or 
later will have to get some sound out of the thing. <thinks for a 
while/>. Its probably worth putting a  couple of mid-price converters on 
the proto board.

> It will not be as easy as pure software though. With software, you can 
> just pop in and actually contribute something with a very small 
> investment. In this case all developers have to be dedicated enough to 
> actually buy the hardware.

Yes, this would cost money to develop. Hardware does.

But remember, the early prototypes are useful little dev boards in their 
own right. I wouldn't mind paying to own one and I guess there are 
others who wouldn't mind either, including maybe people who aren't 
involved in linux and/or aren't involved in audio. Anyway, the "lets 
reverse engineer a driver for product X" approach also requires 
developers to actually buy the hardware. In both cases, I'd expect the 
people involved to be people who also had other reasons to want to own 
the hardware.

> Even if the hardware design is open doesn't prevent a company to make 
> money selling it. Why not design everything the open source way, then 
> let one or more companies make the investment required to prototype 
> and test the card. These companies can then make money by selling the 
> hardware, either completely assembled or in parts. To get someone to 
> do an initial investment is probably required to get the price down. 
> It is often necessary to buy components in the thousands to get a 
> reasonable price.

Yes, and there's lots of additional costs involved in a production 
model, eg packaging, production engineering, CE marking etc. It would 
require the involvement of a business, or the developers would have to 
put on business hats.

> Maybe you should first investigate if there already is a firewire 
> prototype card on the market, which can be hooked up to audio converters.

I'd be surprised to find anything combining dsPIC, FPGA and firewire. 
Buy a dev board for one and you'd still have to lay out the others. 
Might as well just lay out all three. This thing *is* a dev board which 
can be hooked up to audio converters.

Cheers

Simon