On 19 June 2017 at 15:35, David C. Partridge david.partridge@perdrix.co.uk
wrote:

I can't recall how you step the data, but it must be in the manual.

I wrote a bit of C code that attempts to correct the timebase, if 10 MHz is
put into channel A. It measures the frequency on channel A, then adjust the
crystal frequency until the frequency is 10 MHz. That's probably not the
best way on a TI counter, but seems to work reasonably well - at least any
errors are small compared to the fact that only integer values may be
programmed into the calbytes. Hence it seems impossible to fully correct
this.

One issue is that frequency shift is non-linear with respect to the
calbytes. So the approach I took was to shift the value by 4, see the
effect of the shift, and then use that to try to work out a suitable
correction.

Below is the bit of C-code. If you want the complete set of code, which
reads data from the SR620, I can send you it. I don't think it is well
commented, and in any case has only been tested on Solaris. It will only
work with an NI card, but I would expect it to work under

case SETTIMEBASE:
  output(device_descriptor,"MODE3;"); /* Set frequency mode */
  output(device_descriptor,"SRCE0;");  /* Measure source A */
  do {
    calbyte_4=read_calbyte(device_descriptor,4);
    frequency=read_frequency(device_descriptor);
    frequency_error=frequency-1e7;
    if (fabs(frequency_error) < 0.0008)  {
      printf("GOOD ENNOUGH frequency error=%lf\n\n",frequency_error);
      exit(0);
    }
    printf("cb4=%d f=%lf\n",calbyte_4, frequency);

    if(frequency_error > 0)
      set_calbyte(device_descriptor, 4, calbyte_4+3);
    else
      set_calbyte(device_descriptor, 4, calbyte_4-3);

    printf("Changed calbyte 4 a little cb4=%d f=%lf\n",calbyte_4,

frequency);
new_frequency=read_frequency(device_descriptor);
new_frequency_error=new_frequency-1e7;

    if(frequency_error > 0)
      frequency_change_per_calbyte=(new_frequency-frequency)/3; /* in

Hz/calbyte /
else
frequency_change_per_calbyte=-(new_frequency-frequency)/3; / in
Hz/calbyte */

printf("frequency_change_per_calbyte=%lf\n",frequency_change_per_calbyte);
/* Do a correction of half the value calculated */
calbyte_change=-(int)
((new_frequency_error/frequency_change_per_calbyte)+0.5);

    /* Attempt to correct the frequency */
    set_calbyte(device_descriptor, 4, calbyte_4+calbyte_change);


    /* Read the frequency, then exit if within an acceptable tollerence

*/
frequency=read_frequency(device_descriptor);
frequency_error=frequency-1e7;
calbyte_4=read_calbyte(device_descriptor,4);
printf("new cb4=%d frequency error=%lf\n\n",calbyte_4,
frequency_error);
} while (fabs(frequency_error) > 0.0008);
exit(0);
break;

Yes please. ...

-----Original Message-----
From: time-nuts [mailto:time-nuts-bounces@febo.com] On Behalf Of Dr. David Kirkby (Kirkby Microwave Ltd)
Sent: 19 June 2017 21:03
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] SR620 TCXO calibration

On 19 June 2017 at 15:35, David C. Partridge david.partridge@perdrix.co.uk
wrote:

I can't recall how you step the data, but it must be in the manual.

I wrote a bit of C code that attempts to correct the timebase, if 10 MHz is put into channel A. It measures the frequency on channel A, then adjust the crystal frequency until the frequency is 10 MHz. That's probably not the best way on a TI counter, but seems to work reasonably well - at least any errors are small compared to the fact that only integer values may be programmed into the calbytes. Hence it seems impossible to fully correct this.

One issue is that frequency shift is non-linear with respect to the calbytes. So the approach I took was to shift the value by 4, see the effect of the shift, and then use that to try to work out a suitable correction.

Below is the bit of C-code. If you want the complete set of code, which reads data from the SR620, I can send you it. I don't think it is well commented, and in any case has only been tested on Solaris. It will only work with an NI card, but I would expect it to work under

case SETTIMEBASE:
  output(device_descriptor,"MODE3;"); /* Set frequency mode */
  output(device_descriptor,"SRCE0;");  /* Measure source A */
  do {
    calbyte_4=read_calbyte(device_descriptor,4);
    frequency=read_frequency(device_descriptor);
    frequency_error=frequency-1e7;
    if (fabs(frequency_error) < 0.0008)  {
      printf("GOOD ENNOUGH frequency error=%lf\n\n",frequency_error);
      exit(0);
    }
    printf("cb4=%d f=%lf\n",calbyte_4, frequency);

    if(frequency_error > 0)
      set_calbyte(device_descriptor, 4, calbyte_4+3);
    else
      set_calbyte(device_descriptor, 4, calbyte_4-3);

    printf("Changed calbyte 4 a little cb4=%d f=%lf\n",calbyte_4, frequency);
    new_frequency=read_frequency(device_descriptor);
    new_frequency_error=new_frequency-1e7;

    if(frequency_error > 0)
      frequency_change_per_calbyte=(new_frequency-frequency)/3; /* in Hz/calbyte */
    else
      frequency_change_per_calbyte=-(new_frequency-frequency)/3; /* in Hz/calbyte */

printf("frequency_change_per_calbyte=%lf\n",frequency_change_per_calbyte);
/* Do a correction of half the value calculated */
calbyte_change=-(int)
((new_frequency_error/frequency_change_per_calbyte)+0.5);

    /* Attempt to correct the frequency */
    set_calbyte(device_descriptor, 4, calbyte_4+calbyte_change);


    /* Read the frequency, then exit if within an acceptable tollerence */
    frequency=read_frequency(device_descriptor);
    frequency_error=frequency-1e7;
    calbyte_4=read_calbyte(device_descriptor,4);
    printf("new cb4=%d frequency error=%lf\n\n",calbyte_4, frequency_error);
  } while (fabs(frequency_error) > 0.0008);
  exit(0);
break;

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

On 19 June 2017 at 21:43, David C. Partridge david.partridge@perdrix.co.uk
wrote:

OK, the code is here

http://www.kirkbymicrowave.co.uk/tmp/srs-0.02.tar.gz

Note

1. It has only be tested on Solaris SPARC, with a National Instruments GPIB
   card.

2. I would expect it to run under Linux, with few if any changes. If any
   are needed for an NI card, please let me know. I don't mind trying to get
   it to run under Linux, or any Unix system.

3. I've no idea how easy it would be to run under Windoze, and have no
   intention of trying to help getting it to run under Windoze.

4. There's no man page, but there is some rudimentary in-built help, if you
   run it with the wrong number of arguments. Below I run it with none, so it
   generates a semi-helpful help message. It needs at least one option, and a
   GPIB address. So something like

srs --time 12

would run it time-internval mode, and expect to find the SR620 on address
12.

I've NOT distributed this before, and have made no attempt to clean it up
to consider it in a fit state to distribute. But feel free to try it. Send
any comments directly to me.

Consider it released under the GPL version 2, or at your option any later
version.

drkirkby@buzzard:~/srs-0.02/src$ ./srs
srs - Version 0.02, by David Kirkby, G8WRB, drkirkby@kirkbymicrowave.co.uk
Dumps data from a Stanford Research SR620 time-interval counter
Usage: srs [options] GPIB_address
MODE
--time
--width
--risefall
--frequency (or use --freq)
--phase
--count

SOURCE (or start)
--A
--B
--reference (or use --ref)
--ratio

SAMPLE SIZE
--size n (n=1 to 1000000)

DISPLAY
--mean
--rel
--jitter
--max
--min
--trig
--dvm
OTHERS
--autocal
--readcalbyte n
--setcalbyte n m (sets calbye n to the value m)
--settimebase (Put 10 MHz into input A before running this, to correct
the timebase)
--help
-h, --help
-v, --version
-V, --verbose