[time-nuts] time stamping counters

Magnus Danielson magnus at rubidium.dyndns.org
Wed Jan 2 23:15:52 EST 2008


From: "Pete" <peterawson at earthlink.net>
Subject: Re: [time-nuts] time stamping counters
Date: Wed, 2 Jan 2008 18:58:44 -0700
Message-ID: <003c01c84dac$35198e00$0200a8c0 at BASE1>

Pete,

> I was unable to respond to your message earlier; too many visitors
> & too many diversions.

Know the feeling.

> Thanks for the info on the CNT-91; I have read the datasheet
> thoroughly and it would be VERY nice to own one, but too
> many $$ for me.

Indeed. I have a CNT-90 here, or is it CNT-91. Who knows. ;-)

> I have downloaded the evaluation copy of Timeview. It works
> well, but without any capability to save/store results, it's only
> a toy for me. I can't justify the $395 license fee since I'm retired.
> Oh, well.

Indeed. I can understand the viewpoint. I've hacked up my own processing from
the raw-format file.

> I have a licensed copy of HP VEE 5.0 (now quite old), but it
> does allow me to operate my CNT-81 from GPIB and save
> data for later processing. Using the internal (block) function
> I have been able to capture TI data at 6Kpoints/sec, but records
> have to be limited to 4466 points per capture. Still, this is 
> useful.

The Timeview seems to pull data faster. I suspect that just as with the HP
5370, 5371 and 5372 there is a binary mode in there. Don't have a programmers
manual lying around.

> While learning to use my CNT-81 I was surprised to discover
> a minor, but possibly important anomaly in it's performance.
> I was attempting to discover the best case jitter measurements
> which could be taken on stable sources at 5, 10 and 100MHz.
> For the 5 & 10 MHz sources, the jitter consistently reads <
> 22ps for large & small samples. At 100MHz I was expecting
> jitter results as good, or better than those at lower frequencies.
> But, this was not the case. The jitter results were >40ps!
> After consulting the folks at Pendulum, it was confirmed that
> input signals near the internal 100MHz reference cause
> period measurements to exhibit 240ps peak-to-peak jitter at
> 10x the frequency delta of the reference to the input. I have
> screen shots of Timeview analysis if you're interested.

Would be nice to see, but it is to be expected. Certain timing relationships
can be screwed up by another signal. Infact, just hooking another signal up to
an input may cause jitter peaks and values as the relative phase slowly shifts.

This is infact why I got myself a HP 5359A. It allows a bit more arbitrarily
delay setups and I can (and have) verified the delay and jitter using better
equipment than CNT-81 or CNT-90.

Cheers,
Magnus



More information about the time-nuts mailing list