[time-nuts] Using 5335 frequency counter for timing

[Bill Byrom]

First I should disclose that I work for Tektronix, and have done so for
30 years. So I have worked directly with many generations of
oscilloscopes. Good oscilloscopes can make timing measurements of up to
about a millisecond with timing errors of 10's of nanoseconds.

There are a very large range of oscilloscopes using different
technologies over the past few decades, so there are several different
answers to your questions. The range of bandwidths and new instrument
prices range over a factor of roughly 1,000. Bandwidths range from about
30 MHz to 70 GHz, and prices from a few hundred dollars to a few hundred
thousand dollars US.

Many high-end scopes do have external clock inputs (usually 10 MHz), but
this feature is not usually available on lower end or older
oscilloscopes. This only affects the accuracy of longer time interval
measurements (over 1 us), where the internal clock error is usually not
any better than 1 ppm (and 50 ppm for low cost oscilloscopes). Internal
timebase error for a 1 second measurement interval is in the several
10's of ppm for most low-end to mid-range oscilloscopes.

Short-term time measurement error for high-end oscilloscopes can be
under 1 ps, and in equivalent time mode is limited by trigger and
aperture jitter. Lower cost oscilloscopes might have short term timing
errors of several ns, limited by jitter and risetime (bandwidth).

Analog oscilloscopes have analog timebases which have large errors
compared to digital oscilloscopes. They have been obsolete for about
20-25 years.

There are several digital oscilloscope technologies. 
*  The easiest to understand is real time sampling, where you have a
sampler running at a high sampling rate and each sample is converted in
to a digital value in real time. A single trigger event results in a
waveform capture, but the instrument can capture a waveform without a
trigger event. The memory length can be large (1 G sample in some
cases), but the sampling rate is limited by the need for the A/D to
complete the conversion in real time. Using internal A/D interleaving,
waveform sampling rates up to 200 GS/s are available, but the cost is
very high in such cases. Lower cost recent models typically have
sampling rates of 1 to 5 GS/s.
*  Many (but not all) real time oscilloscopes also offer random
equivalent time sampling. In this case the instrument samples at a much
lower sampling rate than would be required for the chosen time/division
setting and waveform length. So, for example, every 10th waveform point
might be filled after each trigger. In this mode, the signal has to
trigger the instrument multiple times (in some cases thousands of
trigger events) before all of the waveform record points are filled. The
sample points are purposely randomly delayed at the start of each
sub-acquisition cycle and the displayed points appropriately skewed so
that all waveform samples are filled without aliasing. This gives you
much better time resolution than with real time sampling, but the
trigger must be extremely accurate and the signal being measured must
have low jitter. Unless you want to get an eye pattern, the signal must
be repetitive (same exact waveform on each trigger). A trigger is
required for this mode to build up a waveform.
*  Sequential sampling oscilloscopes: These instruments are often
referred to as "sampling scopes", although this is now a misnomer since
all digital (and some analog) oscilloscopes use a sampler. This is a
sequential form or equivalent time sampling, and only one sample (or
less) is acquired for each trigger event. So a 1,000 point waveform
requires that 1,000 triggers must be accepted, and since the trigger
processing rate is usually on the order of 200 k/sec only about 1/1000
of the trigger events are actually used if the trigger signal is 200
MHz. An external separate trigger input is usually required. This
instrument has the highest time resolution and short-term timing
accuracy and vertical accuracy of any oscilloscope type, but must be
used for stable repetitive signals.

The waveform update rate of equivalent time mode (repetitive or
sequential) is much slower than real time mode due to the many waveforms
which must be acquired in equivalent time mode to build up a waveform
record. 
--
Bill Byrom N5BB
Tektronix RF Application Engineer

----- Original message -----
From: Hal Murray <hmurray at megapathdsl.net>
To: Discussion of precise time and frequency measurement
<time-nuts at febo.com>
Cc: hmurray at megapathdsl.net
Subject: Re: [time-nuts] Using 5335 frequency counter for timing
Date: Sun, 07 May 2017 20:11:58 -0700


kb8tq at n1k.org said:
> None of them will do as well as a really fast scope.

How accurate is the clock in a scope?  Do the high end scopes have an 
external clock input?

I remember playing with a scope many years ago.  Trigger on a PPS from a
GPS, 
look at the next PPS.  It should be 1 second later.  I think the scope I
was 
using was off by 6 PPM.

I'd expect that there would be a crossover.  For short times, the scope
would 
be better.  For long times, the better crystal (or external input) in
the 
5335 would take over.


-- 
These are my opinions.  I hate spam.



_______________________________________________
time-nuts mailing list -- time-nuts at febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.