[sdiy] Using analog for physical modeling

[Eric Brombaugh]

On 09/21/2011 02:03 PM, Matthew Smith wrote:
> Quoth Eric Brombaugh at 22/09/11 00:08...
> ...
>> I've done a few boards for this sort of thing using
>  > low-end Xilinx Spartan3A parts.
>
> How low-end? Wondering how much I can squeeze out of the 3AN part I have
> here, which has 50,000 gates. Sadly, it's the smallest of the Spartan
> 3AN parts, but the largest that Element14 carries in a solderable
> package - all the larger ones are wretched BGA. (Digikey/Mouser are out
> of the running for me due to their arm-and-a-leg shipping prices for
> sub-$200 orders.)

I've been using the 200kgate version in a VQ100 pkg. Available from 
Digi-Key for ~$13 in single qty. That's about 1/2 what Element14 charges 
for the BGA version of the 200kgate parts.

The 50k part can probably do a fair amount of stuff though. I've found 
that my simpler designs are only taking about 25% of the resources in 
the 200k parts.

> I can get a better-sized Spartan 3E and store the bitstream in a
> Platform Flash device, but it's extra cost/board real-estate. Object of
> the exercise here is to see how much I can cram into something really
> cheap.

You don't need to use the expensive Platform Flash configuration 
memories - Spartan 3E parts do support SPI configuration which is a lot 
cheaper. Just needs a $2 8-pin flash chip, a decoupling cap and a few 
traces on the board.

>  > ...but for longer delays you'd probably want an external SRAM.
>
> Think I'll incorporate a load of SRAM (which is cheap) into my DVCA, to
> add flexibility.

You'll want to if you're using the 50K FPGAs - they've got only 3k x 
18bits of SRAM. Enough for about 64ms of delay at audio rates.

>> ...good-quality audio codecs...
>
> Something I haven't had anything to do with, up to this point. What's
> the advantage of an audio codec - is it that you get two (I guess
> simultaneously sampled) channels in and out all in the same package?

A good audio codec has high resolution (16 - 24 bits) for good SNR, high 
sample rates (even most cheap ones will do 96kHz) and all the 
anti-aliasing filtering built in. It will usually also have some handy 
routing and mixing functions. And they're dirt-cheap (think in the $4 
range). Only downside is that you have to figure out how to implement an 
I2S interface in your FPGA, but that's easy (I've got some sample code 
on my website).

> I was about to do the schematics for the 3A-based board, and was going
> to use discreted ADCs/DACs, but could certainly be swayed towards a
> more optimal solution.

Whatever floats yer boat. I prefer codecs for simplicity, cost & 
performance.

Eric