[time-nuts] Fwd: OT: 10 MHz data capture, help

Magnus Danielson magnus at rubidium.dyndns.org
Sat Apr 11 14:43:21 UTC 2009


SAIDJACK at aol.com skrev:
> Sorry, my previous post should  have read: 40K individual Time Interval 
> Measurements per second on the  Wavecrest. I think it's the fastest time 
> interval counter out  there?

No. The HP5372A happily gives you a rate at 13,3 MHz sample rate. I 
think the SIA-3000 is higher than 40k too (I think I recall 200k) and I 
know that the Pendulum CNT-90 and followers also passes that rate. It is 
however far from that simple number, as maximum block length can vary 
alot and continous mode rate can be lower than burst-rate due to pure 
I/O concerns. In general will binary output perform better than ASCII 
output. The HP5372A binary output mode is actually pure core data 
values, which means that the CPU is only copying the databits and does 
I/O processing rather than elaborate calculations. The HP5372A also has 
a binary output option which would allow continous stream output at full 
rate as data is hooked  in before the 8192 sample long block memory 
(where 1 or more samples is lost as reference samples, which explains 
the kind of odd varying sample lengths).

A derivate of the HP5372A in VXI form can come up to 80 Msamples/s as I 
recall it. However, that is among the few major spec changes than from 
the HP5372A cases.

So, there is two classes of performance: burst-rate and continous mode.
Burst-rate mode is limited by maximum sample burst rate (essentially 
limited by the time between two sample) and burst length (number of 
samples).
Continous mode is limited by sample rate alone, which really is an 
effect of I/O. Many counters actually lack a dedicated continous mode, 
but can be tricked into it by running them re-triggered and do 
synchronised binary readout. Very few counters have dedicated hardware 
features for accelerated continous mode.

Cheers,
Magnus



More information about the time-nuts mailing list