[time-nuts] HP 5372A vs. 5370A
Ed Palmer
ed_palmer at sasktel.net
Wed Feb 9 23:00:33 UTC 2011
Magnus Danielson wrote:
> On 09/02/11 22:14, Ed Palmer wrote:
>> I routinely use the 5372A to make multiple 1 sec. measurements of
>> frequency or time interval and then dump the results via GPIB for
>> analysis. Standard stuff for either the 5370 or 5372 except as mentioned
>> elsewhere, these measurements are made with no dead time.
> I love the time deviation view.
>> My unit has Option 040 which adds FFT capabilities. I can do phase noise
>> measurements within the limitations of the 200ps resolution of the box.
>
> Very nifty thing to have in the box. Both the units I use have Option
> 040, but I rarely use it.
>
>> The June 2008 issue of Circuit Cellar had an article ("Let's be Crystal
>> Clear") on crystals and oscillators. The 5372A was used to analyze the
>> startup behaviour of an oscillator. You should be able to get this
>> article through inter-library loan or maybe there's a copy lurking on
>> the net somewhere. There are also various manuals and app notes online
>> at the Agilent site.
>
> The Frequency vs. Time view is lovely.
>
> The 5372A does calculate ADEV but does not provide ADEV plots.
>
Since they don't specify, I'm assuming that it uses the original,
non-overlapping calculation. True?
>> The biggest limitation of the 5372A is that you can only make 8191
>> frequency measurements or 4095 time interval measurements with each
>> measurement lasting no longer than 8 seconds. If you want to use it for
>> making more or longer measurements (e.g. long term Allan Deviation
>> measurements) you have to start doing some GPIB programming, but I think
>> you lose the no-dead-time advantage.
>
> The internal memory can handle 8192 time-stamps. Depending on
> intricate details of various measurement modes one or more reference
> samples is taken, and one or two channels is time-stamped. The event
> and time counters is sampled on each measurement but runs
> continuously, even when measurements is not running. You can program
> bursts such that retrieving the burst can be done such that a new
> burst can be started without loosing either the continuous sample
> series or no-dead-time advantages. However you need to do binary
> transfers.
>
I was wondering about that. Is that info buried in the programming
manual? I haven't gone through the binary programming section hardly at
all.
> Another approach is to use the Fast Port which taps into the hardware
> and would allow longer runs. I have not much documentation on it and
> experience on how it works. Needs to play with it.
>
My unit doesn't have the Fast Port option so that isn't a possibility
for me. You'd obviously need some custom hardware to make it happen.
More complications.
>> If you see a 5372A that's reporting an "Error 160: Out of sensitivity
>> cal." don't worry about it. It just means that the onboard lithium
>> battery is dead and needs to be replaced. The calibration is trivial
>> to do.
>
> I've done the exercise for two 5372A's and it was easy to do.
>
>> The most important thing about either the 5370 or 5372 is to study it.
>> Figure out it's capabilities, limitations, and tricks so that you use it
>> to it's full potential. I'm still working on that!
>
> The HP5372A programmers manual is really an inside-out manual where
> you learn the internal formats and how they are processed for the
> various views and results. Very open in what it does and how you can
> duplicate the measurements from binary format. It's a bit confusing
> initially, but once you spent the time to learn it, it makes sense.
>
> There are a few GUI-hacks I would like to have, for longer
> measurements I would like to have an ETA count-down, but real-time
> update for certain views would be lovely.
I wrote a little GPIB program to automate some measurements. I added an
approximate 'percent completed' printout. It really comes in handy.
Ed
> The CNT-90 outshines the HP5372A in this respect. The post-measurement
> zoom functionality of the HP5372A is not a marvel of functionality but
> provides a basic function, where as the CNT-90 could do more of that.
>
> Cheers.
> Magnus
>
More information about the time-nuts
mailing list