[sdiy] Black Magic oscillator cans

[Magnus Danielson]

From: Tim Ressel <madhun2001 at yahoo.com>
Subject: [sdiy] Black Magic oscillator cans
Date: Sun, 20 Jul 2003 02:56:50 -0700 (PDT)

> Yo,
> 
> Does anyone have experince with programmable
> oscillators? ECS and Cardinal make these little
> oscillators that look like normal clock oscillators,
> but are programmable. It would seem you can program
> any frequency into these things you want. However I
> cannot find documentation that admits this feat
> directly. For $6.00 it seems a real deal. I need
> 2.00024 MHz oscillators to go with TOG chip sets, and
> this might be the way to get there.

Epson has them too. Most higher frequency crystal oscillators work the same
way: a fixed crystal oscillator and a PLL with a CMOS VCO. It is easy enought
to program up frequency dividers to get most frequencies of interest.

For the Epson ones you see to require some special programmer, but I wouln't
be supprised if it is a fairly simple single-wire protocol.

Be aware that these programmable PLLs usually have pretty bad cycle-to-cycle
jitter. This comes from a badly designed PLL loop which is pushing the VCO
here and there all the time. If you work with jitter-sensitive applications
there is no choice than to measure the cycle-to-cycle jitter yourself.
Rule of thumb is that the RMS jitter needs to be 1/14 of the bit-clock for
BER values of 10^-12.

However, if you only think of musical use such as TOG I think they are more
then sufficiently stable.

You could actually synthesize the frequency yourself. 2.00024 MHz you say...
There are good CMOS VCOs available which allows you to do synthesis over a
large range of frequencies. However, 2.00024 MHz doesn't work well with divide
and multiply setups, since it is 2000240 = 2^4 * 5 * 11 * 2273. The 2273 factor
is what worries me. It probably requires a stepup of 2273 somewhere in the
line. It's not that it is impossible to do such big stepups, but I do wonder
if the off the shelfs PLLs is able to do it. The 11 factor could come from the
reference itself (say 11 MHz). However, there is other ways to synthesize such
a frequency, by using frequency shifting, but maybe that is a bit too obscure.

Maybe another TOG reference frequency will work almost as well for you, but be
much more easy to acheive. BTW, I can't figure out how you derived 2,00024 MHz.

Cheers,
Magnus