[time-nuts] Fast frequency counting question

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sun May 4 17:46:05 EDT 2008

Magnus Danielson wrote:
> Murray,
>> Thanks for your reply. Yes, I believe Leon has 30dB gain post mixer. He
>> is using an FSS 1000E as his mixer. So, we don't think the problem is
>> slew-rate limiting. Your suggestions for checking the performance with
>> changes in level is a good one.
> It should be standard practice... it is simple enought to rule it out.
> A simple -6 dB pad away.
> However, I am starting to consider that you are primarilly single-shot
> resolution limited.
Yes, for 1ppb measurement in 10us you need a resolution/jitter of around 
1ps or so when measuring the 198KHz signal.
Achieving such jitter is easy with a properly designed slope amplifier 
(you can even do this with a 1kHz signal).
You are unlikely to achieve this jitter by just feeding the amplified 
mixer 198 kHz output to the counter unless the amplifier gain 
distribution and associated filtering is correctly chosen.
That just leaves you with the problem of selecting a counter with 1 ps 

Sampling the 40MHz directly with an ADC wont achieve the required timing 
jitter and resolution of about 10fs as there are no ADC's out there with 
such low sampling jitter.
The sampling jitter of available high resolution fast ADC is around 50fs 
at best.
However when mixing down to 1MHz or less the sampling jitter requirement 
is relaxed to 2ps or so, which is easily achieved with an appropriately 
filtered low noise sampling clock source.
>> This is an area of counting I've not considered before. We are working
>> with the chip manufacturers, and they report the problem, which we would
>> like to replicate. Personally I can't see it as an issue, but it has
>> been raised, and we would like to understand how to measure it. The chip
>> manufacturer has a high-speed A-D system and some method (unknown to us)
>> for post-processing the data.
>> I've suggested that Leon reply to you directly, as he has a better grip
>> on what the problem and requirement is. I understand that the problem is
>> relayed to warm-start TTFF performance in a GPS application.
> OK.
> I'd happy to help you out.
> Cheers,
> Magnus


More information about the time-nuts mailing list