[Phys-L] timestamping instrumentation

[John Denker <jsd@av8n.com>]

On 10/20/21 1:22 PM, Tom Van Baak wrote:

> There's been mention of timestamping in this thread. Those of us that
> work with GPS and precise clocks use it a lot. Time stamping,
> sometimes called time tagging, has advantages over traditional period
> / frequency counters or time interval counters.

Yes. Tremendous advantages.

> Most any microcontroller can be used to collect time stamps;
> microprocessors with GPIO pins as well. It's easy to get resolution
> ranging from 100 ns (10 MHz clock) to 1 ns (1 GHz clock).

Yes.

> There are several laboratory-grade instruments that do timestamping
> to tens of ps. Look for the Keysight 53230A, or the Stanford Research
> FS740, or the Pendulum CNT-91. All of these are high-performance
> universal counters, with lots of features, including timestamping [1]
> [2] [3].

> [1]
> https://www.keysight.com/us/en/assets/7018-02959/technical-overviews/5990-7861.pdf
>
>  [2] https://www.thinksrs.com/products/fs740.html
>
> [3]
> https://pendulum-instruments.com/products/frequency-counters-analyzers/cnt-91-91r-advanced-frequency-time-interval-analyzer/

Those are nice, but they cost thousands of dollars.

> There is a student-friendly, hobbyist-grade time stamping board that
> grew out of ham radio circles [4]:
>
> https://tapr.org/product/tapr-ticc/
>
> This is based on an Arduino with the heavy lifting being done by a
> wonderful timing chip originally designed by TI for time-of-flight
> measurement. The TAPR TICC does timestamping to better than 100 ps
> resolution.

Impressive.
 
> On the other end of the performance spectrum you can do timestamping
> using any old laptop and just a serial port. A win32 example is
> here:
>
> http://leapsecond.com/tools/pctsc.c
>
> It has its limitations but it shows that a lowly serial port pin can
> be used as a trigger for time stamp recording. I use this for quick
> 'n dirty low rate Geiger measurements. For more precision and no
> dead time I use dedicated hardware for the counting & timing and the
> PC just for data logging.

1) The corresponding linux program is here:
  https://www.av8n.com/cgit/poissfit/tree/timestamp.c

2) However.... I don't recommend any of the serial-port solutions for
 the radioactivity application. (Other applications maybe.)

 -- Here's one of the many, many issues: Each click of the Geiger counter
  is a very narrow spike. A son of a glitch. That's very unlike anything
  a serial port was designed to handle. If you have 1000 events per second
  with a nice 50/50 duty cycle, the serial port can handle it fine, but
  if you have even a few events per second consisting of spiky clicks
  you're gonna lose events.

 -- You're gonna need a level shifter if you want to drive the serial port
  pins reliably. The hardware cost for that is not large, but it's comparable
  to the entire system-on-chip solution.

 -- Laptops generally don't have serial ports. They haven't for many years,
  unless you spring for the "docking station / port expander". This is not
  a major issue, because colleges often have ten-year-old /desktop/ machines
  being used as doorstops. You can't run up-to-date windoze software on them,
  but you can run linux. It's easy to install and runs fine. So there is some
  value in having the aforementioned linux timestamping program.

 -- Don't even think about using a DB9-to-usb conversion dongle. Those don't
  work reliably even for routine applications, and pulse collecting is far
  from routine. You'd be better off buying a PCI-e serial card for your
  desktop box ... but again, the hardware cost of that is comparable to the
  system-on-chip solution.

 -- Recent linux kernels (at long last) have some usable real-time features,
  which are used (in a rudimentary way) in the program mentioned above.
  Trying to get real-time performance out of a windoze box is masochism.
  It was not designed for that.

 -- Every minute you spend working on a custom solution is two minutes wasted.
  You're much better off finding something that already has a fan base, i.e.
  a user community that has already solved (or at least identified) the problems.
  Something like the tapr-ticc or perhaps the Raspberry teensy solution that
  was mentioned a while back:
    https://www.raspberrypi.org/forums/viewtopic.php?t=300464
    https://github.com/PaulStoffregen/FreqMeasureMulti
    https://www.pjrc.com/teensy/td_libs_FreqMeasure.html