[time-nuts] mixers for frequency measurement

Magnus Danielson magnus at rubidium.dyndns.org
Sat Jan 21 11:27:57 EST 2012


On 01/20/2012 02:54 PM, Ulrich Bangert wrote:
> Bob,
>
>> ~2x10^-10 you can do this with a good frequency counter, no
>> mixers needed. ~2x10^-11 you can do this with a very good
>> /hard to find / expensive frequency counter. 1.0x10^-11
>> pretty easy, nothing very fancy required for a single mixer
>> approach. 1.0x10^-12 works fine with an RPD-1 and some care,
>> but not a lot of crazy stuff 1.0x10^-13 you need some
>> attention to detail, and may need a better mixer. 1.0x10^-14
>> can I come live at your house? If you have this sort of
>> stuff, the cost of a fancier test setup should be a minor
>> issue. 1.0x10^-15 indeed people do measure this stuff.
>> Proving accuracy at this level involves a lot of work on
>> secondary effects.
>
> In general I agree to all these numbers. I just want to point to the fact
> that a HP5370 or a SR620 allow for a kind of "high resolution mode". This is
> a mode in which the counter is externally armed to make 1000 TI measurements
> per seconds and display the mean of them. Which gives a SQRT(1000)
> improvement of all counter related non systematic errors. My experiments
> with a SR620 indicate a 6E-13 noise floor for Tau = 1s without any need for
> mixing for two 10 MHz sources. Naturally this works ok only for source
> frequencies>= 1000 Hz ( The higher the frequency the less trigger noise ).
>
> For those of us who have no access to H2-masers or BVA-oscillators as a
> reference this may be not exactly an overkill but quite good to characterize
> HP10811/FTS1000(1200) or the like not to mention anything worse than that.

Such block averages will indeed improve resolution, but one has to 
recall that the block-summing causes a pre-filtering shifting the 
counters effective bandwidth. Once that is considered (really just an 
issue for sources of noise over-shining the instruments contribution) it 
will be fine. The end result will be that the instrument limit slope 
hits the level of the stable source much earlier.

I will try to spend some more effort to see if there are further 
theoretical limitations to consider.

It's more of a practical limitation of getting all those readouts that I 
wonder about. I consider hooking up a FPGA board on the back of the 
HP5372A high speed interface to see what that gives me.

I also have another project on a FPGA ongoing with a DDMTD test, but 
last time I tried things I ended up with a tool problem.

Some of the GPIB interfaces used, can have serious rate limits. So, do 
elaborate some of what you have done.

Cheers,
Magnus




More information about the time-nuts mailing list