Lostgallifreyan wrote:
An off-topic question, but very interesting at least to me... Do DDS chips
exist with 128 or even 256 phase accumulators onboard with the step size
adjustment being capable of matching the speed of the stepping itself
(though taking external control of local phase modulations between
accumulators), and allowing mixing of all outputs, perhaps in user-selected
groups based on binary fractions of the total accumulator count?

I ask because if they do it might be possible for me to convert my phase mod
synth code to dedicated hardware without resorting to very fast CPU's...
While the rates are audio only, the huge parallel array gets demanding of CPU
time as it is.

I suspect the answer to all that may be 'no' without custom VLSI chips
because of the relatively complex paths between accumulators needed for a
phase mod synth of N operators per algorithm, but maybe DDS chips come in
enough varieties to surprise me.

You can do this with an FPGA. That is a programmable VLSI chip that
you program yourself to perform the functions you need. It operates
at logic speed, not at processor speed.

Rob wrote in :

You can do this with an FPGA. That is a programmable VLSI chip that
you program yourself to perform the functions you need. It operates
at logic speed, not at processor speed.


Thanks. I'll look into it. I suspect my abilities might be constrained to
running my code as currently, on a PC with at least 1 GHz for the 48+ voice
polyphony I'm getting, but Yamaha hadn't got that when they got 16 voices out
of a DX7 so I do at least have a strong incentive to explore.

Lostgallifreyan wrote:
Rob wrote in :

You can do this with an FPGA. That is a programmable VLSI chip that
you program yourself to perform the functions you need. It operates
at logic speed, not at processor speed.


Thanks. I'll look into it. I suspect my abilities might be constrained to
running my code as currently, on a PC with at least 1 GHz for the 48+ voice
polyphony I'm getting, but Yamaha hadn't got that when they got 16 voices out
of a DX7 so I do at least have a strong incentive to explore.

For the specific function that you want, there also exist soundcards
with dedicated chips that can do that. In the early days they were
hardware designs, I think today's boards more often use a DSP to emulate
what the hardware did in the past.

Also note that when you need a lot of CPU power to do repeated and
duplicated tasks like that, a modern video card has the perfect architecture
for it. On NVIDIA cards, you can download and execute software that you
develop yourself using their CUDA development tools. Maybe other video
card manufacturers offer comparable tools by now.

Rob wrote in
:

For the specific function that you want, there also exist soundcards
with dedicated chips that can do that. In the early days they were
hardware designs, I think today's boards more often use a DSP to emulate
what the hardware did in the past.


Yamaha did a board with six large IC's on it, but it was very expensive, and
not very common. My I/O is cheap now, 30 quid where a grand was the original
cost! Sadly while it has exccelent MIDI and wave portage, its onboard DSP is
only for routing between local audio ports, and I doubt I can get at it.

Also note that when you need a lot of CPU power to do repeated and
duplicated tasks like that, a modern video card has the perfect
architecture for it. On NVIDIA cards, you can download and execute
software that you develop yourself using their CUDA development tools.
Maybe other video card manufacturers offer comparable tools by now.


That video board idea is nice. Presumably they're common and cheap too,
though I'm not sure if thsi will limit my choice of OS a lot. As it is my
existign hardware limits me to W98, and at a pinch WXP though Wine in Linux
might work.

The main problem with all these mthods is that if I have to add any
specialised hardware and methods, the number of buyers who'd say they cannot
use the synthesier would be about 96% of its likely market. Even staying
with W98 (or Wine) and advising people to buy any cheap old machine to run it
on, so long as it can do MIDI as well as audio, might put that percentage
much higher. I ought to learn some ASM after disassembling my own GCC
compiles too, because I suspect a lot might be improved that way in the fast
loop if I stay with i386 hardware.

Normal service may now be resumed bu those who want it. This was a tad
off-topic and long at that... I was just taking the opportunity.. I will look
at the FPGA, for sure, on the offchance that it allows a dramatic reduction
in the speed and complexity of the sontrol system.

Thanks, Rob, for that help. I'll save this thread.