SV: Re: [sdiy] audio microprocessors and C++

Veronica Merryfield veronica.merryfield at
Fri Mar 7 01:16:36 CET 2008

> If one looks closely at mobile phones each one has quite capable
> synthesis engine, closed for the phone user offcourse but still
> there are a vast amount of computing power in a regular phone.

Indeed. Most have an ARM9 or better in them together with a  
significant DSP processor. Some have just a DSP and some have a second  
ARM core. Take a look at TI's OMAP range to get a good idea of what is  
in many phones. Intel XScale has ben used along with devices from  
Freescale. Also, Samsung have a couple of dev boards that are an  
almost ready to go phone solution.

Whilst I would agree that the intel quad cores are pretty powerful  
beasts, they have a few draw backs. I would rather see a mix, perhaps  
like the chameleon, where one CPU is used for IO and a DSP is used for  
the core audio work. DSPs tend to be good at the MACs in single cycles  
(although pipelined) and the associated memory access with better  
clock to power ratios, although I know general purpose processors are  
getting good at it. Also, dual and quad cores give you more CPU horse  
power, but without a corresponding upgrade to the memory pathways.  
Also, in most cases, adding more processors takes more CPU time to co- 
ordinate and communicate. For much of the kind of work needed in a  
rack module, a management CPU (midi, ADC, DAC, USB? etc) and a  
dedicated DSP would do the trick. In theory, such a device could work  
as a complex sequencing engine or as a complete synthesis module or as  
a CV device and synthesis component. Get the pricing right, and the  
way is open to a rack with several of these mixed with analogue  
modules. Could be a very interesting project.


More information about the Synth-diy mailing list