[time-nuts] Sorry for the unedited post: Counter Internal Oscillator Importance with ExternalReference?
magnus at rubidium.dyndns.org
Wed May 10 16:08:10 EDT 2017
On 05/10/2017 07:12 PM, Tom Knox wrote:
> Hi All;
> I used my speech recognition software last night and reading my post
> this morning I was embarrassed. Hopefully this email will make
> sense. Thanks everyone for their input. Tom, i have a question; was
> there any difference between the two 53230A oscillators when locked
> to an external reference? I have also experienced poor results with
> my 53230A HS phase locked to an external reference. My issue is
> constant pop-ups showing reference errors. I wonder if that is
> related to issues you encountered. I would like to compare oscillator
> options in the same counter model and I have several each 53230A's
> and MCA3027's but sadly they have the same oscillator option in each
> respectively so I cannot directly compare PLL oscillator
I won't comment on the 53230A's as I haven't dug into them (yet).
> My thinking is a phase locked internal reference may not be the best
> option for those of us with ULPN, Ultra Stable house references so I
> am trying to reverse engineer the Tektronix FCA/MCA counter to
> determine if there is away to directly feed an external reference
> instead of phase locking the internal reference. My plan is to find a
> way (if possible and it is encouraging) to do this while kee ping the
> rest of the functionality intact, Which does appear possible on the
> Tektronix, Fluke, Pendulum instruments (All the same base units
> branded differently). It appears depending on the oscillator option
> U11B sends an off/on signals to switch between internal oscillator
> options; Std Pin3, Oven Pin5, or Rubidium Pin4. This switching then
> has the installed reference option provides 10MHz signals to U11B
> pin204 RB, Pin205 Std, Pin206 Oven respectively, most likely changing
> PLL parameters at the same time. Whether this is auto sensed or
> manually controlled I have yet to determine. The selection may be in
> the menu. My thinking is this may allow simply adding a connector and
> cable to the rear panel to apply an external reference directly. In
> these counters U9A seems related to the PLL functions including
> control, sensing, and switching using the standard External Reference
> input. If my idea pans out this feature would remain unchanged so the
> internal oscillator can still be used ( When the direct feed is not
> appropriate) to clean up a dirtier external reference. My plan is
> (Unless someone else has already looked into this is to next see if
> the same is also possible with the 53230A. The BIG question does this
> improve performance. Any thoughts? I spoke with a friend Fred Walls
> today and asked if he though the internal reference was much of a
> factor when Lock to a good external reference and he did not thin so.
> So I will keep anyone interested informed with my progress since I am
> sure many of you have reference that may benefit from this
> modification if it does improve performance.
Now, there is a PLL in the CNT-90/91 (Fluke IDs are PM6690/6691 and
Tektronix IDs are FCA-3000/3100) and that is to lock the 100 MHz
oscillator to whatever 10 MHz reference you got muxed in. The PLL chip
is U9A which is an ADF4001 and the 100 MHz oscillator is around U47A,
L3, C171, C177-C180 and D23. You need this setup, so you cannot bypass.
The FPGA does not to anything magic, there is first a mux to select
whatever internal 10 MHz reference there is, according to option. Then,
if it has external 10 MHz selected, it then selects that and outputs.
The ADF4001 is used as a phase-comparator and needed divide by 10 to
compare things. The Charge-Pump output has the traditional capacitor and
capitor/resistor chain, so there is your PI control-loop, which is
exactly what you would like to see there. It would track in due to first
degree thermal shift and I'd expect it to be sufficiently high bandwidth
not to cause any concern of interest.
The external clock input includes a crystal filter, which helps to shave
of disturbing sidebands.
I'm skeptical that there is much gain to be found in modifying that.
I know other designs have had problems, but I don't see that this one
should have any major reasons for concern. I could do some measures once
I'm back home.
If you have high-quality, low jitter/phase-noise 100 MHz, then you could
feed that into the counter too, but it may need the 10 MHz for some
things, I haven't checked all details yet to rule it out.
Still, I wonder if it is worth the trouble.
The designers knew where their main problems where, and this never
popped up as a big concern for them.
> Thanks again for everyone's input.
Thanks for keeping us out of trouble. :)
More information about the time-nuts