[sdiy] digital delay, was ...

[john mahoney]

> The BBD 'is' the delay...
Yeah, I knew that. I worded my question badly.

> ... but it cannot pass a good or repeatable DC level.
> So the idea is to move the voltage information to the frequency domain.

Wonderful, thank you. I wasn't seeing it.

> Before affordable data acquisition systems... tape recorders were used
with
> voltages expressed as frequencies... which the tape could reproduce.

Clever old engineers! (Did I say old? I meant experienced. [Probably some of
'em on the list.Ahem!])

> > > ... If you used sine waves... I think you could
> > > do away with most of the anti-alias filters as well.
> >
> > The need for filters is a function of the sampling rate and the
bandwidth of
> > the devices downstream. Unless you are running square waves thru the
BBD,
> > you may need an a-a filter.
>
> ahhh... no.  you need a filter that cuts off at 1/2 the BBD clock at the
input,

I was talking about the output filter. Is that not also an anti-aliasing
filter? If not, sorry. I haven't talked much about sampling (analog or
digital) in years, until lately.

> The output would still need some filtering.
Yes! Heh heh...

> A square wave at the input would absolutely guarantee that you'd NEED an
> anti-alias
> filter... unless the input frequency was an exact subharmonic of the
clock. BBD
> and square wave inputs are a super-bad idea :^P

Well, maybe it depends on the ratio of input freq to clock frequency, as
well as the purpose of the signal -- you get different types of distortion
from each approach (filter or no filter). With a gate or trigger, maybe it's
better not to filter that leading edge at all, for quickest rise time i.e.
the least timing error; while on the trailing edge there is at most a one
sample (er, bucketload) error, certainly no big deal. But I've been wrong
before, as we've all seen. :-)

Cheers!
--
john