References: <3E244AFE.4F3E@sneakemail.com> <3E26E3E2.email@example.com> <3E2922B4.C3B@sneakemail.com> <3E2A07F4.firstname.lastname@example.org> <3E2A5081.4C97@sneakemail.com>
Subject: Re: Binary Sampler
X-Newsreader: Microsoft Outlook Express 6.00.2800.1106
Date: Sun, 19 Jan 2003 22:04:26 GMT
NNTP-Posting-Date: Sun, 19 Jan 2003 17:04:26 EST
Organization: Cox Communications
"John Larkin" wrote in
> On Sun, 19 Jan 2003 16:47:33 GMT, "Mike" wrote:
> >"Mike Monett" wrote in message
> >> Mike wrote:
> >> [...]
> >> > Ugh. That's not good. Okay, then, what is your sample rate for the
> >> > sine and square waves?
> >> >
> >> > -- Mike --
> >> I'll have to check the software, but I'm pretty certain the sampling
> >> at the same frequency.
> >Okay then, let's try this: you input a sine wave at frequency fo, and you
> >sample at frequency fo-df, and your output frequency is df, where df <<
> >Precisely the same result would be obtained if your input frequency was
> >You get (fo-df)/df samples per cycle of the df frequency sine wave. The
> >samples are passed through a limiter and low pass filtered to average out
> >the noise.
> >-- Mike --
> I *think* that MM is proposing to use a dds to generate a sampling
> timebase for his 'binary sampler'. If the signal frequency is known,
> you can in theory synthesize a slightly different sampling clock and
> thus sample the waveform down to a low, sorta-baseband I guess you
> could call it, frequency. The LeCroy patent I cited above does just
> My observations here are that
> 1. You're adding the signal reprate jitter to the measurement, unlike
> a triggered-timebase scope that ignores this. So if the reprate has
> any jitter, the waveform is trashed. Lots of interesting signals have
> jittery reprates but contain picosecond detail: lasers, radar, most
> any laboratory pulse generator, most anything triggered by an
> affordable crystal oscillator.
> 2. It's very hard to make a DDS with picosecond edge jitter over wide
> time spans; hell, even over narrow time spans (he says, with chagrin.)
> Has to do with the price of perfect dacs and ideal lowpass filters.
Indeed. Analog synthesis doesn't fare much better. Given a reference input,
I can lock to the reference and guarantee a maximum jitter variation
relative to that reference. The only way I'd know of to generate the sample
clock for Mike's scheme is to either generate it from the same source as
that used to generate the input, or to derive the clock from the input, then
use it to sample the input. Those approaches produce a clock at the same
frequency, though. It would be significantly more painful to generate a
sample clock that's offset by a very small frequency.
> 3. It's about a jillion times cheaper and easier to do a conventional
> time-delay-ramp timebase for a sampler.
> LeCroy is the modern Philco: always doing strange stuff.