[sdiy] Function Generator ICs

[Tom Wiltshire]

On 19 Sep 2007, at 09:59, Paul Perry wrote:

> Maybe a micro would be the go today? Anyone tried tio make a  
> general purpose
> voltage controlled audio osc from a PIC?

I have. I tried at first with a 8-bit processor (PIC 16F series chip)  
which is fairly limited, but is still able to copy the 16-bit divider  
DCO of the Roland Juno synths. I wrote firmware for a chip which  
takes a 7-bit digital MIDI note input and generates a square wave of  
the appropriate frequency. The Juno followed this with a waveshaper  
stage to generate ramp and PWM waves.  You could use the PICs 0-5V  
analogue inputs for frequency modulation, although I didn't bother.  
The analogue inputs are only 10-bit, so you have to watch the range  
of the modulation to avoid zippering. You get true analogue  
waveshapes though, not stepped waves from a DAC.

I'm currently working on a dsPIC (16-bit processor) audio osc,  
although this won't be directly voltage controlled. Instead, it will  
have its basic pitch set digitally (either via an SPI connection or  
directly from a MIDI serial stream) and use analogue inputs for  
frequency and waveshape modulation. This gives you digital stability  
and the analogue ability to feed anything you like into it. Output is  
16-bit via a cheap 24-bit/192KHz Cirrus DAC. The DAC is stereo, so I  
hope to be able to have two independent oscillator outputs.

Bruce Duncan/Eric Brombaugh of Modcan Electronics have a commercial  
example of something similar with the VCDO 58B, which also does two  
operator FM synthesis.

There are definitely problems with this approach. I'm vaguely keeping  
in mind that I'd like to build a polysynth, but the current  
consumption of the processor (70-100mA at the speed I'm using) means  
I'm going to be using a big power supply if I put 16 voices in it -  
300+W just for the oscillators! (How does this compare to synths of  
old?)

If anyone has any great ideas about a better way, I'd love to hear  
it. I'm starting to think that Scott Gravenhorst is going in the  
right direction with the almost-custom-hardware and massive  
parallelism of FPGAs.
T.