[sdiy] Really Low, LFO

[cheater cheater]

Hi Kyle,

On Fri, Apr 2, 2010 at 23:35, Kyle Stephens <lightburnx at yahoo.com> wrote:
> Accuracy means getting, in this case, a specific frequency. Precision means it may be wrong (not necessarily), but at least it's consistently 'wrong'.
>
> My ears are liable to pick out what "sounds good," independent of any specific value like 5.000Hz, and I'd like to be able to pick more than less the same point again and again when I set the pot to a given position, relatively free of stuff like temp drift or other second order interference.

What's second order interference? I tried finding the definition of
that but was unable to. I've seen this mentioned on the list every now
and then but can't figure out what this is.

> I brought it up as I figured most material on function gens that slow would be scientific equipment where both those factors are important concerns.

You can try looking in weather station designs if you want to. They
have to work on very long cycles of statistic series, high chance
they'll have very slow lfos, plus they have to withstand the weather.

On Fri, Apr 2, 2010 at 23:45,  <jays at aracnet.com> wrote:
> I think I'd agree here, I'd go digital. I do think a DSP would be overkill. Especially with what you get in low end uCs like PICs now days.

Don't take me wrong - I was talking about DSP as in 'digital signal
processing' not 'dsp microcontroller' :-) Definitely something that
can be done on a bread and butter uC.

> When you have timing this long the quality of parts becomes critical. Things like op-amp offsets start to affect symmetry. I now use LT1013s in some of my LFOs because of this. As mentioned issues with caps become critical and so on.....

Interesting tip, will be checking it out!

On Fri, Apr 2, 2010 at 23:49, Tim Parkhurst <tim.parkhurst at gmail.com> wrote:
> On Fri, Apr 2, 2010 at 1:51 PM, Kyle Stephens <lightburnx at yahoo.com> wrote:
>> Can anyone point me to some reading material on infrasonic oscillator design? I want to build a really low, low LFO, with a period of around half an hour (yes, _that_ slow!).
>>
>
> Hey Kyle,
>
> How about dividing a square wave waaaaaay down? You can use a variety
> of oscillator designs, of course, and you could use an integrator
> (with a big honking cap) to smooth the square out. Alternately, you
> could possibly use an R/2R converter and sum several divided down
> squares (higher freq's get attenuated more) to get a pretty fair
> approximation of a slow triangle wave (still might require a smoothing
> cap).

Tim,
you might be looking at problems such as accumulative error from the
frequency of the squarewave, and any fluctuations in pitch will
transfer to the divided-down output. Generally - and this is not a
fully scientific expression - oscillators seem to be specified more
for 'local' pitch than 'global'. For example if you're looking at
local pitch (e.g. the pitch over the last 1 second) and the phase
jumps every now and then it's statistically insignificant in this
scheme so the pitch is still 'within parameters', but the global pitch
(i.e. number of cycles / time elapsed) is going to be off. This 'phase
jump' might happen because of aliasing. Since the pitch of a
divided-down oscillator will depend on more cycles of the original
undivided oscillator, it's statistically more prone to this kind of
problem. But I don't understand the measurement error science there
enough to say for sure if it can be significant enough to be an issue
- by intuition, I think so.

Kyle,
if you're OK with doing this in digital, then there are most probably
single-chip signal generators that will work well for you; give it a
search.

Cheers
D.