[time-nuts] TI Counter Noise Floor/Resolution

Magnus Danielson cfmd at bredband.net
Sun Jul 31 16:47:59 UTC 2005


From: Brian Kirby <kirbybq at bellsouth.net>
Subject: [time-nuts] TI Counter Noise Floor/Resolution
Date: Thu, 28 Jul 2005 15:06:50 -0500
Message-ID: <42E93ADA.7050205 at bellsouth.net>

Brian,

Just back from a 4 day of music festival, so lagging behind somewhat.

> Why we are talking about resolution and noise floor, I am looking for a 
> merit figure to say the data I have measured, is valid.
> 
> As an example, when I worked in satellite communications, we had a 
> measurement called C/kt. This is a carrier to noise measurement, 
> normalized back to a 1 hertz bandwidth. When we made these measurements 
> with a spectrum analyzer, we had to account for the measurement 
> bandwidth of the instrument, the detector used (factors for sine wave 
> response vs. gaussian noise), etc. Basically if the reading was 6 db 
> above the noise floor, then we could believe the measurement.
> 
> In my case, were I have baselined the HP53131A noise floor, I know my 
> readings have to be above this floor, to be valid.
> 
> My question is how much should our readings be above the noise floor to 
> be considered valid, and what units/formulas should we use ? Can we 
> convert Allan variances to a decibel reading, and use similar techniques 
> as I described?

Interesting view on things. A 10*log(AVAR(tau)) would be the appropriate thing
to use, this would be equalent to a 20*log(ADEV(tau)) since ADEV(tau) =
SQRT(AVAR(tau)).

Again, a 6 dB margin is would render the right neighborhood, but then again,
that is a rule-of-thumb rather than exact science.

Cheers,
Magnus




More information about the Time-nuts_lists.febo.com mailing list