From: Mike Monett
X-Mailer: Mozilla 2.02 (Win16; I)
Subject: Re: Binary Sampler
References: <3E244AFE.4F3E@sneakemail.com> <3E26E3E2.firstname.lastname@example.org> <3E2922B4.C3B@sneakemail.com>
Date: Sat, 18 Jan 2003 21:05:40 -0500
NNTP-Posting-Date: Sat, 18 Jan 2003 21:05:06 EST
Organization: Bell Sympatico
> "Mike Monett" wrote in message
> > Mike wrote:
> > I think you meant to say it undersamples. The sample clock can be at or
> > below the fundamental.
> According to your description, you're sampling at the carrier rate. The
> modulating signal, which you're trying to recover, is a much lower
> frequency, and is being oversampled.
Mike, I appreciate your reading the article that carefully. It is really
looking at 100 MHz sine waves and square waves. There is no 1 KHz
modulation on the signal, although that's exactly what I state in the
article. The modulation refers to an internal parameter that sets the
period of the signal. I will modify the article to remove this confusion.
> > Well, that's nice. I can't do that. But I'm not really sure the binary
> > sampler would like all those nice linear equations:)
> My point is that you _already_ do all that, except for calculating the error
> term. The window is a simple description of a low pass filter (it's not
> realizable, but it's an adequate model, and it's very mathematically
> tractable). In your circuit, it's how you get from your DFF output to the
> op-amp output. Look at your Sampler Theory of Operation page.
I'll try to find some time and go over your model and see if I can
understand it better.
> What I meant was that you should subtract the input from the output, so you
> can see the error term by itself.
> -- Mike --
Well, according to the simulation results, as the SNR degrades, the
recovered signal drops in amplitude and shifts in phase. So a simple
subtraction might not tell the whole story.
In some applications, this is of little importance, but it is crucial in
others. So I really need to find out if the hardware does the same, and if
the effect is linear with SNR or does it suddenly collapse somewhere. If it
does, why, and if there's anything that can be done to improve it. Lots of
But I really appreciate your thoughtful comments. Thanks for taking the
time to post.