[volt-nuts] More thoughts on the 3458A

Poul-Henning Kamp phk at phk.freebsd.dk
Wed Jan 22 14:41:21 EST 2014

In message <CAFoWNwBDPPnjEF8gBsFWsFPCFc6qhG+dWG=oYEAhuGgU=wHXrw at mail.gmail.com>
, Jan Fredriksson writes:
>One thing that strikes me reading old 3458A product datasheet from HP is
>that they stressed the high measurement speed, 100kHz @ 16 bit, but not
>empasizing much the DC accuracy.

If I had no scruples at all, I would start digitizing audio using
two HP3558A's synchronized to a Cesium frequency standard, and
market it to the audio-homoepathy segment at prices you can not
even imagine :-)

>Now, the 3458A remains state of the art in
>DC accuracy / liearity, but for high speed measurements and of course
>frequency, the 3458 is now nowhere near state of the art any more.

The biggest issue is that you cannot use an external sample-clock,
but only free-wheel the 3458A.  NIST did modify a 3458A during
some of their J-J AC experiments, so they could synchronize the
sampling to the synthetic AC generated by the J-J.

I don't know how "state of the art" it is or isn't, you can obviously
buy chips which do 100 GS/s @ 16 bits these days, but my impression
is that they are nowhere near the HP3458A in absolute precision or
for that matter, reference stability.

The main limitation is that the HP3458A has very little RAM, so to
use the 100kHz sampling-rate, you need to download and run code
which can compress the samples in real-time, preferably into something
GPIB almost can keep up with.  That is easier than it sounds, but
still a hazzle most people would prefer to be without.

Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

More information about the volt-nuts mailing list