[time-nuts] FE-5650A option 58 tuning word for 10 MHz output

[Mathias Weyland]

Hello guys

I'm new to this list. I got myself a FE-5650A Rubidium Standard off of 
ebay.
It's the "option 58" 1 pps output variant, hence I have to modify the 
tuning
word used in the DDS phase accumulator to get 10 MHz out. I found a 
vast amount
of awesome descriptions on how to do that on the web and in particular 
on this
list. One write-up that stood out was this one by Mark Sims:

http://www.mail-archive.com/time-nuts@febo.com/msg13486.html

I think I can pull this off since everything is documented so nicely. 
However,
I'm having trouble calculating the right tuning word and this is why: 
Mark notes
that the reference frequency reported by the unit is the one with the 
C-field
pot at the lowest frequency position. He gives a number of suggestions 
on how to
deal with that. Since I didn't get that hydrogen maser for Christmas, 
the best
approach seems to be "to calculate the true reference frequency from 
the saved
(minimum C-field) R=reference frequency and F=divisor word and use that 
value to
calculate divisor words." I don't understand how the saved minimum 
C-field
reference ties into this calculation.

My approach would have been to calculate the true reference frequency 
from the
saved divisor alone, ignoring the minimum C-field calculation. I don't 
see how
the minimum C-field reference frequency would help me since the C-field 
pot is
not in the min position anymore due to factory tweaking. To be 
specific, this is
what I would do:

The unit returns the following string upon 'S':

OK50255055.760840Hz F=2ABB5046B34A2E00

Now based on this, the tuning word should be coded in the first 8 
characters, of
F, i.e. '2ABB5046'. I'm a bit confused about the remaining characters 
being
non-zero. Any documentation I came across has a number that ends in 8 
zeroes...
In any case, 0x2ABB5046 is 716918854 in decimal and the resolution 
would
therefore be

2^23 / 716918854 = approx. 0.0117 Hz which makes sense.

The physics package would then output a frequency of

f_ref = (2^23 / 716918854) * 2^32 = approx. 50255055.809934 Hz

This is higher than the reference given in the 'S' output, which is in 
line with
what Mark wrote. However, scaling this with the average correction 
factor he
gave yields

f_ref * 1.000000002150 = approx. 50255055.917982 Hz

Which is higher than what I would expect. Then again I'm not entirely 
sure what
I would expect because various errors add up in the above calculation. 
I'd be
interested in what people with more experience think about those 
results.

I would then use

M = 10000000/(2^23/716918854) = approx. 854633872.509003

to find the 10 MHz tuning word, which I would then round up 
(unfortunately it's
smack in the middle between two integers...) and convert to hex, 
yielding
0x32F0AD91. This does in fact result in a 10.000 MHz output waveform 
but I have
no means to check its accuracy (yet?). I'd appreciate any hints about 
where
things could have potentially gone wrong, especially with respect to 
the minimum
C-field reference frequency that I ended up not using.

On a slightly related note, I have cooked up a small PCB with a local 5 
V
regulator and status LEDs that mates with the amphenol connector used 
on this
standard. I have to complete the write-up on it and will probably put 
up a
video about the mod on my youtube channel; once this is done I'll be 
sitting on
9 spare boards since I got 10 boards done. If there is interest, I 
could send
off the spares without profit, i.e. for about 5 bucks or so. I imagine 
this
could be of use to those who have the same standard. The board doesn't 
do
anything funky, it is just neat. In any case I'd like to ask if it 
would be OK
to formally place this offer on the list once I got everything ready.

Thanks a lot and best regards!

Matt