[time-nuts] PRS-10 PPS output

Tue Apr 7 13:47:45 UTC 2020

Hi

The main point is that these Rb’s (not just the PRS-10) are not really
set up for what you are trying to do. A proper GPSDO will do a much
better job than something like this. ( GPS into PRS-10  != GPSDO ). 

There are a number of sub-variants of the PRS-10 (and the other telecom
Rb’s) so knowing exactly what this or that example will or will not do is 
tough.  Even firmware versions are not perfect indicators since there are
hardware differences as well as firmware …..

======

Assuming the 5 ps is at the ~ 2 seconds or less level, that’s about 5x10^-12. 
The spec on the PRS-10 (just sitting there) is significantly worse than this (parts
in 10^-11) . Playing with a disciplining input that is *much* worse than that level 
will only degrade things further. 

https://www.thinksrs.com/downloads/pdfs/catalog/FS740c.pdf <https://www.thinksrs.com/downloads/pdfs/catalog/FS740c.pdf>

Page 2 has a pretty good graph showing (green line) what the PRS-10 does
close in. The units are ADEV so they may or may not map directly to your
“5 ps” number. 

Bob

> On Apr 7, 2020, at 9:24 AM, Mike Ingle <finndmike62 at gmail.com> wrote:
> 
> Hi Bob,
> 
> If I am understanding you correctly, it is normal operation that the time
> tagging stops working, and that the PPS out should not phase lock to the
> PPS-in?  It is perfectly fine in my application that the PPS out has a
> fixed non zero offet to the average GPS PPS signal (obviously any
> individual GPS PPS signal has an fairly large uncertainty in the 10s of
> ns).  I had hoped that the unit would lock to the GPS  and act as a jitter
> cleaner for testing our IRIG time stamping ADC boards.  I am new to the
> time-nuts group, but have become interested, since creating our IRIG time
> stamp circuit.  It certainly seems like a never needing rabbit hole of
> needing a better reference source to be able to test the equipment one is
> making.  For example on our IRIG time stamp system I measure 14ps
> uncertainty using the XL-AK as a 10MHz reference, and measuring a fixed
> cable length.  But our customer gets 5ps using a better reference.  I
> wanted to explore the precision of the IRIG time stamp system, and am
> trying to upgrade my lab as inexpensively as possible to find the limits.
> 
> Anyway, everyone's feedback is appreciated.
> 
> --mike
> 
> On Tue, Apr 7, 2020 at 2:53 PM Bob kb8tq <kb8tq at n1k.org> wrote:
> 
>> Hi
>> 
>> Backing up a little bit …. the PPS in on these telecom Rb’s is designed to
>> easily get the part set on frequency. It’s not designed for a GPSDO
>> application.
>> What you are seeing is consistent with that “application target”.
>> 
>> Bob
>> 
>>> On Apr 7, 2020, at 6:32 AM, Mike Ingle <finndmike62 at gmail.com> wrote:
>>> 
>>> Hi all,
>>> 
>>> Thank you for your feedback.  I found the schematics for the PRS-10
>> online,
>>> and the output is 5V HCMOS.
>>> 
>>> signal :
>>> 
>>> locked -->  hc08 in    hc08 out  -->  hc14 in   hc14 out --> 4x 240 ohm
>>> resistors -> output pin
>>> PPS    -->  hc08 in
>>> 
>>> Which brings me to my next question.  I have been running this used
>> PRS-10
>>> FW version 3.21 with apparently (from lady heather) 126000 hours
>> lifetime.
>>> Since Friday when it came in the mail.  It has the bench heat sinks and
>> the
>>> accessory output board.  It is being disciplined by a blox5T PPS signal.
>>> Occasionally the PPS is missing from the PRS-10 (1 in 1000 or so) .  I
>>> presume due to loss of lock, as the and gate would inhibit PPS when lock
>> is
>>> zero.
>>> 
>>> First some over-all results, then some background on my testing
>> methodology.
>>> 
>>> The PRS-10 seems to lock to PPS-in such that the PRS-10 PPS out is 1.84us
>>> after PPS in.  The +1.84us happens to be where the time tag (TT) goes to
>>> zero.  I suspect that something might be wrong with the Time Tag setup.
>>> The unit steers itself until the TT=0ns, at which point the TT no longer
>>> returns data.  The PPS out from this point seems to wander by about 60ns
>> or
>>> so (delay from GPS PPS in to PPS out).
>>> 
>>> Why do I think that I am missing PPS from the PRS-10?
>>> 
>>> Test 1.
>>> 
>>> GPS PPS -> scope ch2 (1M input on BNC T) -> PRS-10 PPS in
>>> PRS-10 PPS out -> scope ch4
>>> 
>>> scope setup to trigger on ch2 PPS in and look 1.84us (200 ns/div) later
>> at
>>> the PRS-10 PPS out.  On infinite persistence, one clearly sees that some
>>> PPS out times  remain at ground.
>>> 
>>> OK so is the problem that the GPS had outliers of more than +/- 1.2us
>> (the
>>> scope display at 200ns/div), or that the PRS-10 did not create a PPS?
>>> 
>>> Test 2.
>>> 
>>> Trigger on PRS-10 PPS-out and set up the scope delay to 1s, to observe
>> the
>>> next PPS after the trigger.  Set up a mask test.
>>> This also indicates a failure.
>>> 
>>> My goal was to get a good low noise and accurate reference oscillator.  I
>>> bought the PRS-10 for 550 euro, and probably should have just purchased a
>>> new one, but live and learn...
>>> It probably does not need to be GPS disciplined, but I wanted to get it
>>> calibrated before setting it to free-running.
>>> 
>>> At this point, to get further, I probably need a time interval counter,
>>> which I can set up using my 4ch 14 bit 500MSPS  A/D boards pulse
>> stretching
>>> input, which should give 5ps RMS time stamping.  Unfortunately, right
>> now I
>>> only have one "good" quality timebase, the PRS-10.  I do have an old
>>> TrueTime XL-AK GPSDO, but the 10MHz out is noisy, with strong 100Hz spurs
>>> (2X line freq in Germany), and my boards on-board 50Mhz tcxo.
>>> 
>>> The two time bases I neglected to mention, are the 10M ref out from my
>>> Rigol DG4062, and the 10M ref out from my Siglent SSA 3021X.  I cannot
>>> comment on their respective quality.
>>> 
>>> Possible test config 1
>>> prs-10 10MHz -> ref in -> LMX2581 (creates 500MHz sample clk for AD9684
>> 14
>>> bit 500MSPS A/D)
>>> PRS-10 pps out -> pulse stretcher ->  ch0 500MSPS A/D
>>> ublox pps out -> ch1 500MSPS  ( does not require pulse stretcher, as the
>>> leading edge is 48ns, and should interpolate to sub 50ps easily)
>>> 
>>> possible test config 2
>>> 50MHz tcxo -> LMX2581 (creates 500MHz sample clk for AD9684 14 bit
>> 500MSPS
>>> A/D)
>>> PRS-10 pps out -> pulse stretcher ->  ch0 500MSPS A/D
>>> ublox pps out -> ch1 500MSPS  ( does not require pulse stretcher, as the
>>> leading edge is 48ns, and should interpolate to sub 50ps easily)
>>> PRS-10 10MHz -> ch2 500MSPS
>>> 
>>> I think that test config 2 is probably the best until I can get a good
>> free
>>> running 10MHz reference.
>>> With the shown setup I can store the arrival time/ rising edge zero
>>> crossing of all of the inputs to a file for analysis.
>>> 
>>> Any further suggestions on test setup?
>>> 
>>> --mike
>>> 
>>> 
>>> 
>>> 
>>> [image: pps_fail.png]
>>> 
>>> 
>>> 
>>> On Sun, Apr 5, 2020 at 6:43 PM Joseph Gwinn <joegwinn at comcast.net>
>> wrote:
>>> 
>>>> On Sun, 05 Apr 2020 12:00:02 -0400, time-nuts-request at lists.febo.com
>>>> wrote:
>>>> Re: time-nuts Digest, Vol 189, Issue 9
>>>> [snip]
>>>>> ----------------------------------------------------------------------
>>>>> 
>>>>> Message: 2
>>>>> Date: Sat, 4 Apr 2020 14:08:02 -0400
>>>>> From: Bob kb8tq <kb8tq at n1k.org>
>>>>> To: Discussion of precise time and frequency measurement
>>>>>     <time-nuts at lists.febo.com>
>>>>> Subject: Re: [time-nuts] PRS-10 PPS output
>>>>> Message-ID: <1131AF5E-3444-4E48-B5D9-ECB36457D014 at n1k.org>
>>>>> Content-Type: text/plain;     charset=utf-8
>>>>> 
>>>>> Hi
>>>>> 
>>>>> Pulling the output of a normal CMOS gate to ground through a small
>>>>> resistor is not generally a good idea. That?s what an attenuator or far
>>>>> end termination of the coax is doing. It?s not specifically outlawed in
>>>> the
>>>>> spec, but it's still not what they are designed to do. Also the output
>>>> level
>>>>> is going to be pretty wimpy run through an attenuator.
>>>>> 
>>>>> One way to 'fix' the problem is with a 50 ohm series resistor at the
>>>> source
>>>>> end. That only works to the degree that the output impedance of the
>> gate
>>>>> is very low when in saturation. How true this is?. that depends.
>>>> 
>>>> Well, the coax 1PPS outputs I've had to deal with are all claim in
>>>> their datasheets to be able to drive a 50-ohm load, so I didn't worry
>>>> about overloading the output circuitry with a standard attenuator.
>>>> Their source impedance seemed to be closer to 200 ohms, as I recall.
>>>> 
>>>> The PRS-10 datasheet (page 59) says:  "The LOCK/1PPS function may be
>>>> configured via RS-232. The factory default is a low level to indicate
>>>> lock, with a 10μs pulse to +5V at 1PPS, with the leading edge being
>>>> defined as the 1PPS timing reference. This BNC output is a CMOS logic
>>>> output via a 1kΩ resistor."
>>>> 
>>>> So, a 50-ohm load (or even a dead short) should not hurt anything.  And
>>>> we can predict the peak voltage over a 50-ohm load driving a 50-ohm
>>>> coax is 5(25/1000)= 0.125 Vpeak.
>>>> 
>>>> For the TTL outputs in the days of yore, the series 50-ohm resistor was
>>>> standard practice.
>>>> 
>>>> Joe Gwinn
>>>> 
>>>> 
>>>> Context:
>>>>> 
>>>>>> I've [JMG] had this issue with coax 1PPS outputs across the board.
>>>>>> What I generally do is to attach a coaxial 50-ohm attenuator between
>>>>>> instrument coax output connector and the 50-ohm cable it will drive.
>>>>>> The attenuator matches the output impedance to the coax impedance.  I
>>>>>> use attenuators between 3 dB and 10 db.  The actual attenuation is not
>>>>>> as marked, because of the mismatch between the ~1K output impedance
>> and
>>>>>> the 50-ohm input of the attenuator.  I suppose that a 75-ohm
>> attenuator
>>>>>> may work as well or better to drive 50-ohm cable, but have not tried
>>>>>> it.  All the ringing et al are suppressed because the length of the
>>>>>> mismatched part of the path is maybe an inch or two.
>>>>>> 
>>>> 
>>>> _______________________________________________
>>>> time-nuts mailing list -- time-nuts at lists.febo.com
>>>> To unsubscribe, go to
>>>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>>>> and follow the instructions there.
>>>> 
>>> <pps_fail.png>_______________________________________________
>>> time-nuts mailing list -- time-nuts at lists.febo.com
>>> To unsubscribe, go to
>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>>> and follow the instructions there.
>> 
>> 
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at lists.febo.com
>> To unsubscribe, go to
>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>> and follow the instructions there.
>> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.