Hi Magnus,
On 5/5/2014 3:30 PM, Magnus Danielson wrote:
Ed,
On 05/05/2014 10:56 PM, Ed Palmer wrote:
Hi Joe,
I didn't realize that the info in the 5061A manual was somewhat
'generic'. Thanks for the pointer.
On 5/5/2014 6:20 AM, J. L. Trantham wrote:
Ed,
If I have the math correct, and you are measuring the voltage to
ground through a 10 MegOhm input impedance DMM, you have about 7.5
nA beam current which seems a bit low compared to what I remember of
the HP 5061A. However, you still have a definable 'peak' with a
'peak to valley' voltage of
about 60 mV or a 'useful signal current' of about 6 nA. If your
unit's circuitry can properly amplify that and keep it a clean
signal, it should work. However, I would recommend setting the OCXO
precisely on frequency with a GPSDO before trying to close the loop
and 'locking' the signal to the CS tube. It will dramatically lower
the work load of getting everything adjusted properly, particularly
in a setting of low beam current.
Easier said than done. This oscillator has no manual frequency
adjustment and if there's an electrical one, I haven't found it yet.
But the oscillator's response seems very vague as though it's just
kind of wandering around so I think there's something not right
there. It's likely related to the bad synthesizer chip.
The 4065 use the STEL DDS to jump around in an interesting patten to
measure the Zeeman position of the main lobe and of the first
side-lobes to steer the C-field in a separate control-loop. This means
that the C-field is completely under CPU control, and no manual
trimming. I think you can disable this and steer it manual over the
serial interface, but I don't recall the details.
The 6 side-lobes separate with heavy dependence on the C-field, so
servo-loop on their position relative the main lobe works really well.
With the C-field servo-controlled like this, the C-field shift of the
main lobe, which is a much weaker dependence and also one of the major
aspects in selecting Cesiums in the first place, a more accurate
Si-realization becomes possible, as well as with much better repeat in
realization as the arbitrary C-field shift can be almost completely
canceled.
I was talking about adjusting the EFC for the OCXO, not the C-field,
although I agree with what you've said. I hope that there's some kind
of adjustment to compensate for crystal aging, but maybe not. Obviously,
this tuning loop is also under CPU control and it appears that the
failed DDS chip results in no RF to the tube at the right frequency so
no signal for the main tuning loop to use to discipline the OCXO.
Hence, the OCXO kind of wanders around. Even replacing the DDS with a
fixed synthesizer wasn't enough to coax the CPU to discipline the OCXO.
As the Zeeman steering and centering is done, locking to Rabi feature
is done in a similar sense, by measuring multiple points. In fact,
this is where it spends most of it's time and the for some of the TDM
slots jump around for the Zeeman measuring points every once in a while.
All that is described in a patent, which then describes this being
driven by the 1802 CPU, but that was later replaced by the 6800-series.
Precision time with whopping 8-bitters. It gives the same warm and
fuzzy feeling as rubidium and cesiums with 741s in PSUs and
control-loops :)
So, this core looks pretty brainless with not as much fun trimmers as
the 5061A for instance, but it has quite better brains, it's just that
the faded STEL 1173 drivers makes it blind.
The STEL 1173 DDS functionality isn't very hard to copy good enough,
but it is still some work do be done. I don't remember where the
datasheet went for it.
It wasn't hard to find the datasheet. I was surprised to find that
Intel took over the part. I don't know how long they carried it, but
the datasheet was from them.
Somehow, the value of 40 nA sticks in my mind from the 5061A. The
5061A manual says end of life of the HP CS tube is a peak beam
current of 8 nA or less. However, I have units with less current
and they still lock. The HP manual also says to measure the voltage
at the output of the tube with a 100 MegOhm or higher input
impedance DMM. If yours is less, that may artificially lower your
values.
I reconfigured my test setup to use a 100 Mohm meter and converted
the results into nanoamps. The 10 Mohm meter was loading down the
tube a bit. The new graph looks somewhat better.
Great.
EOL of the tube is a multifactor issue, including Signal to Noise
ratio and the 'useful signal current' to 'background current' ratio.
The 'background current' is what you see with no RF signal applied
to the tube. Have you measured that? A ratio of 1 is EOL per the
HP manual. If yours is about 4.5 nA, as suggested by the 'off peak'
values shown, or less, you still have a useful signal and,
hopefully, a useful tube.
The ratio of signal level to background level looks quite good. The
background current is only 1.5 na for a ratio of 6:1. I checked that
measurement twice because it looked suspiciously good.
Great.
I'd recommend continuing with the repair.
I will, but it could be world's slowest repair as I try to find one
of those synthesizer chips. I've seen places that claim to have
them, but only in the 48 pin DIP package rather than the 44 pin
PLCC. I REALLY
don't want to do that conversion!
In any case, I have many other toys to amuse me! :)
While you wait, don't waste that tube.
I won't! I'm already paranoid about using up the Cesium. I plan on
doing as much as I can with the tube unpowered - except maybe for the
ion pump, although this tube doesn't seem to need much of an ion pump.
When I bought the unit, I saw a picture showing an ion pump reading of
010, but all I've ever seen is 000. But the supply is working and the
system does display the reading correctly if I use a high voltage probe
to 'load' the supply. I'll have to watch it and see if I get a reading
after it sits unused for a few weeks. I've read messages in the past
that said that FTS/Datum tubes are really good at holding vacuum.
Thanks for all the tips!
I wish I had a 44 pin PLCC variant of the STEL 1173 lying around to
help you out.
So do I! :)
Ed
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
