[sdiy] A query about the TR-909 noise generator
tom at electricdruid.net
Sat Dec 21 13:31:48 CET 2019
I’ve been able to simulate the LFSR and I can confirm that it is a maximal length 31-bit sequence. Not 32-bit as Roland claim in the service manual. Sorry Roland!
The only list I have of two-tap sequences for a 31-bit register gives the following options:
We can now add:
It’s clocked at ~300KHz, so it’ll be very flat. I tested the 4070 XOR-based clock on the breadboard to find out how fast it was. I’ve got an old ST HCF4070BE and an RCA CD4070BE hanging about, and the two chips vary by about 3%, plus the variation you get from the 100p cap (could easily be +/-10%). Given the cycle length of the LFSR, we have 119 minutes of noise before the cycle repeats. Plenty long enough!
I agree Richie, it’s an unexceptional design. But you know how people are about anything x0x - it *must* be magic! That leads them to go looking for 4006’s to try and build one, so I wondered if an exact replacement might satisfy a few people.
Synth & Stompbox DIY
> On 21 Dec 2019, at 11:49, Richie Burnett <rburnett at richieburnett.co.uk> wrote:
> ---- Tom Wiltshire wrote ----
>> I can’t find any references to the two taps they’ve used though - could 13 and 31 be a non-maximal sequence for a 32-stage LFSR? Can anyone confirm this?
> I'm not a good enough mathematician to spot an M-sequence polynomial on sight, but you could always simulate it in C or MATLAB or something to determine the sequence length. You know the initial state, so you just need to count the number of iterations before that state comes around again.
> FWIW, the 909 noise source is pretty textbook stuff though. It's clocked pretty fast so it's flat out to above 20kHz and the sequence is long enough that you can't hear the repeats even at such a high clock rate. There's not much quirky or interesting going on there, (in my opinion.)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Synth-diy