[time-nuts] Conditioning clock signal paths

[Magnus Danielson]

From: "Stephan Sandenbergh" <stephan at rrsg.ee.uct.ac.za>
Subject: Re: [time-nuts] Conditioning clock signal paths
Date: Wed, 28 Jun 2006 20:02:11 +0200
Message-ID: <006a01c69adc$fad714c0$401c9e89 at Stephan>

> Hi Magnus,

Hi Stephan,

> Thank you for the excellent advice.

Thanks for the kind words.

> From what I understood is this: the higher your slew rate the smaller the
> time epoch onto which you map the sum of your signal and the noise. This
> decreases the jitter in a proportional relationship to you slew rate.

Yes. That is the basic problem with jitter-trigger.

> Am I missing the point completely?

No.

> How does this relate to bandwidth limiting to reduce noise?
> (to me it seems it only applies to sinusoids then?). 

This is a good question actually, since it is a contradiction.

The white noise (if we assume it is flat, just for the sake of simplicity)
will increase in amplitude linear to the bandwidth. This is straight out of the
noise scoolbook.

Slewrate depends linearly with the amplitude and the frequency of a sine
signal. Any signal being bandwidth limited with a filter to its frequency will
essentially have the same slewrate (at least very close to) that of a sine of
the same frequency and amplitude. Base unit for slew-rate is V/s or you can
also express it as VHz. It is really the amplitude of the derivate of the sine
signal, so it becomes Ap*2*pi*f if you are picky, or about 8.89*Arms*f.

Now, if you have something close to a squarewave and filter it down to is
fundamental, you will loose in slewrate. You *will* get more triggering-jitter.
If you instead increase the slewrate, you will get less triggering-jitter.

Does this rule out filtering? No. Filtering is still a valid method to remove
non-harmonic signals (which is also noise compared to the signal). It can bite
you as it will probably lower your slew-rate, but you may improve on that after
the filtering, but now you are sure that it is your intended fundamental to a
higher degree than an unfiltered input. Some form of bandpass filter may be
recommended.

Cheers,
Magnus