[time-nuts] Fast frequency counting question

Magnus Danielson magnus at rubidium.dyndns.org
Sun May 4 19:57:43 EDT 2008

From: Bruce Griffiths <bruce.griffiths at xtra.co.nz>
Subject: Re: [time-nuts] Fast frequency counting question
Date: Mon, 05 May 2008 11:21:42 +1200
Message-ID: <481E4506.1000503 at xtra.co.nz>

> Murray Greenman wrote:
> > Hi all,
> >
> > We have a problem at work where one of our engineers needs to measure
> > the start-up chirp in an oscillator destined for a GPS application.
> >
> > We are looking for ppb level frequency resolution with sample periods
> > around 10 - 20us (yes, that's us, not ms!). The chirp is all over well
> > inside 1ms.
> >
> > The measurement is on a 40MHz oscillator, which is mixed down to about
> > 198kHz to improve resolution. We can count the down-converted frequency
> > in a conventional manner with no problem. The problem is of course that
> > at 198kHz the period is ~5us so even an interpolating counter will
> > struggle.
> >
> > Preliminary tests with a Pendulum CNT90 counter give results that
> > deviate +/- 16Hz when the gate time is reduced to 10us. We are using
> > GPIB to capture multiple repeated samples, and trigger the counter when
> > the oscillator is powered up. We also have an Agilent 53181A counter
> > available.
> >
> >   
> Amplified mixer noise could easily account for this if the post mixer 
> amplifier/filter system isnt appropriately designed.
> Optimising the post mixer amplifier and filter will drop this noise to 
> well below the counter resolution.

The problem here is that 198 kHz is a bit high for the normal low-pass filter
approaches... but bandpass filters should be able to pull it off.

The topology should not need to be that complex thought.

Thinking more about it... yes the mixer approach will give a usefull gain of
200, so if propper signal handling is done, sub ps resolution should be
reachable. That should give about 7 digits of resolution.

The short time-base is limiting the beat-frequency (obviously). However,
keeping the beat frequency that high clearly avoids the 1/f noise problems.


More information about the time-nuts mailing list