[sdiy] DSP synth applications (was DIY Polysynths)

[Eric Brombaugh]

thx1138 wrote:
> On 7/29/09 9:45 AM, "Antti Huovilainen" <ajhuovil at cc.hut.fi> wrote:
> 
>> On Wed, 29 Jul 2009, thx1138 wrote:
>>
>>> SoundBite board was designed to be as low-cost possible due to University
>>> support program requests.
>> Fair enough. I can understand leaving out the external ram interface, but
>> surely SPI/I2C header wouldn't have cost that much?
> 
> Our audio team was hoping to have a series of these platforms and then the
> economic downturn forced us to put priority on to other projects with a
> greater return on investment.
> 
> Having said all of this, I tend to agree with many of the comments people
> make, but at the end of the day I was lucky to get this platform out and not
> cancelled.

Terry,

First, thanks to you and your colleagues at Freescale for making the 
SoundBite boards available. I picked one up early last year and got as 
far as writing a few simple test apps. I worked a bit with David DiCarlo 
on reviewing some of the documentation and debugging some IDE conflicts 
with ARM/OCD tools installed under WinXP.

The DSP56371 used on the Soundbite looks like a potentially useful part 
for a number of synth applications I'm familiar with. That said, the 
lack of standard interface peripherals is a significant hurdle. Anyone 
wishing to do more than rudimentary digital bit-banging for 
user-interfaces is probably going to have to add a second, more 
general-purpose MCU to handle things like MIDI, CVs and user interfaces. 
Not that this is an onerous problem, but it does create barriers that 
will discourage a significant fraction of the potential user base.

Most of my audio DSP experience has been at the lower end of the 
performance spectrum - things with 20-60 MIPS and 16-bit wordsizes. I've 
become pretty familiar with the compromises that these restrictions 
impose and I'd love to see some inexpensive parts that are easy to use 
in a DIY context and provide more horsepower and resolution. My wish 
list looks like this:

* >= 24-bit words
* >= 60MIPS
* >= true DSP with MACs and multi-bus operand prefetches/addressing
* fast vectored interrupts from on-chip sources
* single-cycle (no wait) on-chip memory accesses (program & data)
* >= 16kword on-chip data SRAM
* >= 32kwords program flash (or SRAM auto-loaded )
* At least one I2S-capable interface
* SPI, both master and slave capable
* UART
* >= 8 GPIO
* >= 3 32-bit timers
* >= 4 channels 12-bit ADC
* <= 144-pin TQFP (48, 64, 80 preferred)
* < $10/ea in small qty from a common distributor
* free / low-cost development tools

It's a bit of a list, but several of the DSP56k parts already meet many 
of the bullets. I believe that the biggest hole is with the SPI master, 
UART and ADC (and the ADC would be less important with several SPI 
master channels) - adding these capabilities would make the 56k family 
quite attractive for my future projects.

Eric