From: Mike Monett
X-Mailer: Mozilla 2.02 (Win16; I)
Subject: Re: Binary Sampler
References: <3E26E3E2.firstname.lastname@example.org> <3E2922B4.C3B@sneakemail.com> <3E2A07F4.email@example.com> <3E2A5081.4C97@sneakemail.com> <3E2C7D00.C7D@sneakemail.com>
Date: Tue, 21 Jan 2003 11:08:20 -0500
NNTP-Posting-Date: Tue, 21 Jan 2003 11:07:46 EST
Organization: Bell Sympatico
> I suppose it depends on how well you're trying to reconstruct it, and what
> information it contains.
> Since your clocks aren't synchronous, your reconstructed waveform will have
> variations due to both frequency offsets and jitter accumulation. For small
> frequency offsets, you will lose information conveyed by the frequency of
> the aliased sine wave. For large frequency offsets, the bandwidth of your
> filter has to be high, and you lose noise rejection.
> It looks like the only reason your circuit works is that your signal
> frequency and your sample frequency are perfect, and contain no jitter. What
> would happen if the frequencies were offset by 40ppm and the cycle to cycle
> jitter was 0.001UI? I think the output would be much different...
> -- Mike --
The frequency offset gives the time expansion, or effective sweep time.
As you point out, you trade off bandwidth, noise rejection, acquisition
time, and tracking error due to sampler ripple as it alternates above and
below the signal.
Another minor effect is the quality of the integrator. The op amp bias
input bias current causes periodic jumps in the tracking signal, and the
non-ideal behaviour of aluminum electrolytics causes a tiny swing in the
dc output, but less than the tracking error.
The amazing thing is how it acquires the signal after slewing through a
large range. The acquisition takes one sample, just as the simulations
show. There is no overshoot, ringing, or long dribble time as it settles
on the new amplitude. The capture is instantaneous.
Clock jitter causes a reduction in bandwidth. See