Does anyone recognize this board, is it a gpsdo?

[grumpydoc]

I won't be able to get to this until the middle of next week, but what I'll do is repeat my tests and then do the tests that you can do to see if anything is visible with just the 1992.

I should have thought of this long ago.

That would be useful. I might even dig the ISA NEC card out and see if it will play.

But the more I think about this the more I think that the OCXO would have to be fairly broken to demonstrate anything except measurement noise with a 1992. I found a fairly nice document discussing OCXO phase noise and jitter here http://sss-mag.com/pdf/vcxophasejitter.pdf and also a couple of OCXO data sheets eg http://www.wild-pc.co.uk/docs/OCXOVT.pdf (or just Google a few) to get an idea of typical phase noise specs (since the data sheet for the TCO-6920A doesn't give us details).

For the OCXO's I've looked at the largest deviation is generally 10kHz (± 100ps), a couple give values at 100kHz (± 1ns). So if we assume something like -120dB/Hz from 1-10kHz that's maybe -80dB in that band or 1/1E8 cycles with periods in the range ±10ps to 100ps which means (I guess) that the other 99,999,999 samples per 100M should be better than that (I think, I struggle a bit with  maths at this level).

Looking at the signal with a time domain instrument with a resolution in the order 100ps to 1ns should really only show measurement noise - it's too blunt an instrument. If it genuinely shows cycles more than 100ps out the phase noise/jitter is going to be hugely out of spec.

It would, I think, be better to look in the frequency domain - a decent SA should be able to show the phase noise directly, as long as the RBW can go down to the 10's of Hz.

I don't have an SA, decent or otherwise so the only way I can look at the frequency domain is to run an FFT on my 'scope - which is not the fastest or most modern instrument (a Lecroy 9354). The best I can do is a FFT over 1M points captured at 1Gs/s. That shows the 2nd harmonic at about 30dB down which seems about right and a bandwidth for the fundamental of 2-3kHz with no particular "shoulder" - translating ±1.5kHz into jitter means ±15ps. Plausible but given the sampling interval of 1ns I'm not sure whether it's a real measurement or fantasy.

[edpalmer42]

Okay, now I've got a third test to do  . I've got an Advantest R3465 SA, but it only has RBW settings down to 300 Hz so I don't know what I'll see.  Timelab can drive it so it's a trivial test - two or three mouse clicks, sit back and watch.

Ed

[grumpydoc]

Okay, now I've got a third test to do  . I've got an Advantest R3465 SA, but it only has RBW settings down to 300 Hz so I don't know what I'll see.  Timelab can drive it so it's a trivial test - two or three mouse clicks, sit back and watch.

If you are really getting jitter of ±3ns then it should show up as a lot of spectral content up to a couple of hundred kHz away from the nominal 10MHz signal so 300Hz RBW should be OK.

If it is working properly and you wanted to measure the phase noise it might be a bit limited.

One thought - with your original measurements did you look at the OCXO output directly or the LVDS? It would be worth repeating all measurements with both signals.

Disclaimer: I'm making this up as I go along (but hopefully learning as I do so - I certainly understand the FFT on the 'scope a whole lot better).

[edpalmer42]

My original measurements were made directly off the oscillator via a x10 probe.  I haven't looked at the LVDS signals at all.

Making things up is a great way to figure things out.  I do it all the time! 

Ed

[grumpydoc]

My original measurements were made directly off the oscillator via a x10 probe.  I haven't looked at the LVDS signals at all.

In which case it might be worth considering trigger jitter as a source of error, especially as the signal isn't especially high after a 1:10 attenuation so noise will affect the triggering more.

If there is a significant difference between the sine and square wave outputs then that might be the cause of some of the error.

That said the Lecroy is hopeless at triggering cleanly but even it only reports about ± 0.5ns variability in period.

[edpalmer42]

My original measurements were made directly off the oscillator via a x10 probe.  I haven't looked at the LVDS signals at all.

In which case it might be worth considering trigger jitter as a source of error, especially as the signal isn't especially high after a 1:10 attenuation so noise will affect the triggering more.

If there is a significant difference between the sine and square wave outputs then that might be the cause of some of the error.

That said the Lecroy is hopeless at triggering cleanly but even it only reports about ± 0.5ns variability in period.

Although trigger jitter is a consideration, the results were so bad that it's unlikely that it was a significant factor.  I also made measurements on a couple of really high quality oscillators.  Now those results might have been degraded by trigger jitter but they were much, much better than the results for the Epson OCXO.  I'll include the high quality oscillators when I post in a day or two.

Be warned, it'll be a long post! 

Ed

[grumpydoc]

Although trigger jitter is a consideration, the results were so bad that it's unlikely that it was a significant factor.  I also made measurements on a couple of really high quality oscillators.  Now those results might have been degraded by trigger jitter but they were much, much better than the results for the Epson OCXO.  I'll include the high quality oscillators when I post in a day or two.

Ah, that gives a lot more confidence that measurements are valid.

[grumpydoc]

Be warned, it'll be a long post! 

Looking forward to it.

In the meantime I pulled a few FFT pics off the LeCroy

Spectrum for the OCXO output (click on the picture for full size) - span here is 0-100Mhz and 10dBm per division. 10MHz fundamental with 1st harmonic approx 30dB down


Zoomed in on the fundamental, centre 10Mhz and span just 9.990Mhz to 10.01Mhz (10MHz ±10kHz)


Rather crude but if we take it at face value we have ±1kHz at 10dB down and anything more than ±2kHz more than 70dB down. It should really be a narrower trace but the effective RBW is 1kHz so it's a pretty blunt tool.

[Vgkid]

Looking forward to edpalmers update.

[grumpydoc]

Looking forward to edpalmers update.

Likewise - it's a pity I can't make the same measurements as Ed or that I don't have some more appropriate test equipment

I'm fairly sure that whatever the 'scope FFT shows it is not phase noise with a 1ns sampling period. Possibly a bit of AM due to noise.

I'm not even that sure about that 2nd harmonic, given that I will only get 100 points per cycle.

There doesn't seem to be anything which would correspond to 3ns of jitter though.

[edpalmer42]

I've got some measurements!

I compared my NEC GPSDO, which I suspect has a bad OCXO, to my HP Z3801A GPSDO and my Efratom FRT Rubidium Standard.  I used both the Racal-Dana 1992 and the HP 5372A to show the capabilities of both.  During the measurements, the NEC was monitored with u-center by eavesdropping on the LEA-5T output to the micro.  The Z3801A was monitored by Z38xx.  Both units were properly locked to GPS.

The 1992 measurements were Time Interval measurements.  Input A was a 'scope probe set to x1 to give a larger signal and reduce trigger jitter.  It was connected directly to the OCXO output on the NEC or via a BNC probe adapter to the terminated output of the FRT.  Input B was connected to the Z3801A.

The first picture shows the results.  Both graphs are together for time intervals below 100 sec.  This is the result of the resolution of the 1992.  Note that at one second, it's just under 1e-9 which is the resolution of the 1992.  After the 100 second point on the graph, the blue line turns and flattens.  This shows the two GPSDOs being disciplined and tracking each other as they should.  The level that they're at (~6e-12) shows the composite noise for both of them.  It could be one or the other or both.  You can't tell.  The black line shows the same test for the Z3801A vs. the FRT.  Since it continues down further than the blue line, you can conclude that this combination has lower noise than the NEC.  Therefore, the 6e-12 noise is due to the NEC.  If I had run the test longer, the black line would have turned and followed the 'GPS line' which is the approximate performance limit of GPS.

The remaining pictures show my 5372A tests.  This particular test does very fast measurements of the signal's period and increments whichever counter is valid for that measurement.  Each of these tests consisted of 10 million measurements.  The first picture shows the Z3801A.  Out of 10M measurements, 9.2M were in the 100 ns bin.  Next is the FRT.  Even better at 9.7M measurements.  Finally, we have the NEC.  Well....  I think you can see the problem.  I didn't bother to add these up.  There were a screen above and below with the outliers.

One disadvantage to using something like the 1992 is that you have to trade resolution for time.  It takes longer for the graph to dig down to the low levels you want to look at and that gives more time for things to go wrong.  Temperature changes, GPS changes, and noisy connectors will have you cursing.  It took me two runs to get the Z3801A-FRT graph because a connection went noisy and spoiled the data.

To be honest, I expected the 1992 results to be more pronounced than they were.  The 5372A results were so obvious that the relatively good results of the 1992 tests was a bit of a surprise.  But it still shows that the 1992 is capable of showing the difference between my NEC and the other units.  The question for Grumpy is whether his 5680 is quiet enough to show the difference.

This message is big enough.  I'll post the Spectrum Analyzer pictures in a new message.  I have to get set up to take those measurements so it'll be later today.

Ed

[grumpydoc]

I'm looking forward to the SA data as confirmation but the OCXO does, indeed, look as though it is completely broken.

To the precision that I can measure I'm not seeing the same for mine - but Ed has an instrument with 5x the time domain resolution, and I don't have any instrument to look directly at the frequency domain.

I'm still waiting for the GPIB adapter which might be a while.

I might see if I can use the 'scope as a glorified data logger, especially with the interleaved sampling as that potentially gets me down to 100ps resolution (though I'm not sure whether the act of interleaving samples will destroy the data I'm trying to see).

[edpalmer42]

Here are the spectrum plots, but they don't show anything obvious (to me at least).  The levels for the NEC are low because I used a low-impedance, high-bandwidth x10 probe to prevent loading the OCXO down in case it's internally terminated in 50 ohms.  The probe looks like 500 ohms so it should be okay.  The other plots were a direct 50 ohm connection.

The phase noise plots are giving me a bit of trouble.  The mention of FFT reminded me that my 5372A has an FFT option that lets me make phase noise measurements within the unit's 150 ps resolution.  If I'm willing to wait long enough, I can make measurements 1 Hz from the carrier.  Whether the numbers are valid is another matter!  I'm trying to get some consistency between the HP 5372A and the Advantest R3465 spectrum analyzer.  Give me a day or two to work through it and make sure that the measurements make sense.

Ed

Edit:  The plot labelled 'FRK' is the FRT Rb Standard.  FRT is the name of the unit.  FRK is the Rb Standard inside the unit.

I also tried zooming in on the carrier to look for any difference between the NEC and the others.  At RBW=300 Hz and VBW=1 Hz they looked the same.

[grumpydoc]

The phase noise plots are giving me a bit of trouble.  The mention of FFT reminded me that my 5372A has an FFT option that lets me make phase noise measurements within the unit's 150 ps resolution.  If I'm willing to wait long enough, I can make measurements 1 Hz from the carrier.  Whether the numbers are valid is another matter!  I'm trying to get some consistency between the HP 5372A and the Advantest R3465 spectrum analyzer.  Give me a day or two to work through it and make sure that the measurements make sense.

You can probably resolve AM 1Hz from the carrier if you get enough samples. I don't see how you can resolve phase noise/modulation that close with 100ps of time resolution because the difference in period between 10Mhz and 10.000001MhH is only 10 femtoseconds 

You'd need to be looking at 10kHz deviation before the difference in period was > 100ps.

Thinks I have a Marconi 2305 modulation meter somewhere that I picked up at a rally for peanuts. I wonder if that is any use in looking at the phase noise.

[edpalmer42]

I think you're probably right.  I haven't attempted to sort through the math to see what I could expect in terms of offset frequency and dBc/Hz values.  But here's a couple of things I noticed.

Device      'Alleged' phase noise @ 1 Hz offset via the HP 5372A
Z3801A      -56 dBc/Hz
FRT            -55
NEC           +52  Plus?  Really?  Is that even possible??

So I figured that this was some kind of noise floor in the mid -50s.  Fair enough.  The Z3801A and the FRT are rated better than this.  Then I measured a Datum FTS-1200B OCXO which has specs that are much better than the Z3801A or the FRT.  I got a value of -64 dBc/Hz which is also much worse than the spec.  Now I don't know what I'm measuring.

Ed

[edpalmer42]

Well, the phase noise tests were pretty much a bust.

As you can see in the first picture, below 3 KHz, all I'm seeing is the resolution limit of the R3465.  Above 3 KHz, the NEC, Z3801A, and FRT all measure about -105 dBc/Hz, in other words, the noise floor of the R3465.  Above about 40 KHz, the NEC noise floor is high enough to measure while the others are good enough that I'm just measuring the R3465 again.

The HP 5372A noise floor was around -90 to -95 dBc/Hz so it didn't show anything.

And now, after all this screwing around, it looks like the NEC isn't as bad as I thought.  It turns out that there's a relatively strong (-30 dBc) 20 MHz component on the OCXO output.  This was confusing the 5372A and making the OCXO look much worse than it really is.  When I filter that out, the results in the second picture look fairly good.  It didn't affect the phase noise measurements because they're so close to the 10 MHz carrier that they didn't see the other signal.

Note to self:  Always look at the bloody signal first!!

 

Ed

[edpalmer42]

In an attempt to redeem myself after wasting everyone's time, I've had a bit of a breakthrough in talking to the unit and getting it to respond.  Connect to the MCU port on the main connector at 115200,8,N,1.  I used a MAX232 equivalent to do the interfacing.  Remember that it's only 3V3 so don't use a 5V chip.

It looks like it's speaking TL1 https://en.wikipedia.org/wiki/Transaction_Language_1.  Lukier's post that showed the output during bootup gave me the clue.  Those semicolons at the end of each line gave it away.  Unfortunately, that means that it's going to be very difficult to figure out what the commands are and what the responses mean because each company can do whatever they want. 

The only command I've found so far is 'status;'.  It accepts upper or lower case.  All commands must end with a semicolon.  The response was 'STATUS=2,A,W;' with no antenna attached, 'STATUS=4,A,W;' a few minutes after connecting the antenna and 'STATUS=3,O,T;' after about 30 minutes.

Another valid command might be 'restart;'.  It results in a 'SYNTAX_ERROR;' response instead of 'UNKNOWN_CMD;' so maybe there's something missing.

It looks like the command interface times out after 5 or 10 minutes with no input.  Disconnect and reconnect to reestablish communications.

Ed

[grumpydoc]

Good discovery regarding the TL1.

I don't think that the measurements etc were a waste of time, you probably now know your test equipment better now at the very least.

The phase noise plot is reassuring and, as you say the harmonic content is easily filtered - in fact it looks like there is a filter on the board which I've always assumed is a low pass filter to clean up the output from the OCXO before converting to logic levels.

But -30dBc harmonic content is not unusual for an OCXO - nor indeed for a signal generator which are usually something like -25dBc to -30dBc, and the Z8301 is not all that much better if you looks at the spectrums  on the SA but gave a much tighter spread of timings on the 5372A - basically everything was +/- the precision of the instrument which is what you'd expect. Even with some filtering the NEC .was not as tight.

Still something slightly odd here I think.

Did you look at the square wave LVDS output on the 5372A?

[edpalmer42]

More testing showed that the problem wasn't really the 20 MHz harmonic after all.  I wasn't being careful enough when I was probing the signal.  External noise (mainly local FM broadcast signals) was visible on the signal.  When I used my low impedance probe and better grounding on the probe, those signals disappeared and the numbers reported by the 5372A improved to the point where there's no significant difference between the NEC when probing at the oscillator and the FRT or the Z3801A.

As for the LVDS, it was a fussy, finicky, PITA measurement that required 3 hands, but I did it.  If you can use that output, it's definitely the way to go.  I looked at the + signal on one of the 10 MHz outputs with my low impedance, 700 MHz probe.  The rise and fall times are too fast for my poor 100 MHz scope - all I saw was the scope's limit of ~ 3.5 ns.  The signal was 0.4 Vp-p with a DC offset of ~1V2.  I don't have a picture, because that would have required yet another hand!  It was a nice, clean square wave with minimal overshoot or other aberrations.  But I was able to get the 5372A measurement.  See below.

Ed

[Velund]

The only command I've found so far is 'status;'.  It accepts upper or lower case.  All commands must end with a semicolon.  The response was 'STATUS=2,A,W;' with no antenna attached, 'STATUS=4,A,W;' a few minutes after connecting the antenna and 'STATUS=3,O,T;' after about 30 minutes.

Good find. Maybe it will be possible to make a simple and meaningful front panel status indication using some PIC and few LEDs...

[jpb]

I see someone is looking to make a good profit on one of these - though to be fair, the price is reasonable if it works ok - it is just more than 4 times the original sellers price:

http://www.ebay.co.uk/itm/10MHz-GPS-Disciplined-Oscillator-GPSDO-/151847766917?hash=item235ad50b85:g:7WQAAOSwLqFV9Zp8

[Velund]

The only command I've found so far is 'status;'.  It accepts upper or lower case.  All commands must end with a semicolon.  The response was 'STATUS=2,A,W;' with no antenna attached, 'STATUS=4,A,W;' a few minutes after connecting the antenna and 'STATUS=3,O,T;' after about 30 minutes.

Another valid command might be 'restart;'.  It results in a 'SYNTAX_ERROR;' response instead of 'UNKNOWN_CMD;' so maybe there's something missing.

It looks like the command interface times out after 5 or 10 minutes with no input.  Disconnect and reconnect to reestablish communications.

Ed

I suspected that some of fields in responce to status command actually related to OCXO, and left my unit without GPS antenna for a few hours (just to see if W means Warmup), but no, it still was 'STATUS=2,A,W;' after 3 hours. BTW, I do not see any timeouts, left terminal open for 20 mins without input and it was still responsive

Playing with commands, I found another one... It is 'debug'...

If entered without parameters, it returns two 16-bit hex values alternately

debug;

0000;
debug;

00A9;
debug;

0000;
debug;

00A9;
debug;

0000;

Tried to enter some parameter and got some 'stable' responce...

debug 0;

00A9;
debug 0;

00A9;

Looks like it accept any hex value as a parameter...

debug 0000;

00A9;
debug 0001;

3373;
debug 0002;

3235;
debug 0003;

FFFF;
debug 0004;

412F;
debug 0005;

0432;
debug 0006;

0150;
debug 0007;

0000;
debug 0008;

0000;
debug ffff;

FFFF;
debug fffe;

FFFF;
debug 8000;

FFFF;
debug 012A;

1000;
debug 012b;

8A80;
debug dead;

FFFF;
debug deadbeef;

FFFF;

If it is simply a dump of words from MCU address space, maybe there is a chance to 'milk' flash contents and look for the rest of commands?

[Velund]

Experimented with status command little bit. Started with unit that stayed power on for about 6 hours without GPS antenna connected. Programmed 'status;<cr><lf>' to a hotkey to send command by one keypress...

Before connection of antenna , X3 and X4 blinked and status returnng  'STATUS=2,A,W;'.
about 20 sec after antenna connection LEDS changed to steady X3 and status started to return  'STATUS=4,A,W;'
Observed brief blink of X3 and pressed a key immediately, got  'STATUS=4,A,I;'... In a second another keypress returned "A.W" again.
Started here to poll it every second or even faster. In a few minutes seen once more switch to "A,I" but again momentarily, just a single reply between "4,A,W"'s...
Got a phone call and was forced to switch my attention to more important things for a 20 min or so.
When returned to this stuff, observed blinking X4 and status switched to 'STATUS=3,O,T;'.
Tried to disconnect antenna. Status almost immediately switched to 'STATUS=2,O,H;'. X3 and X4 started to blink.
Reconnected antenna. Status in a few moments switched to 'STATUS=1,O,H;' for a few seconds, then returned to 'STATUS=3,O,T;'.
Disconnected antenna again. Status almost immediately switched to 'STATUS=2,O,H;'. X3 and X4 started to blink.
Connected another antenna that was indoors and wrapped by aluminium foil (no GPS signal, but normal antenna current). Status almost immediately switched to 'STATUS=1,O,H;'. X3 and X4 continue to blink.
Now it stays in '1,O,H' for about 10 minutes, will look what will happen later.

So, for now I observed following values:
First field - 1,2,3,4... Yet to figure out what it really mean, but seems 1 is for no lock at all, 2 for antenna problems, 3 for timing lock and 4 for position-only lock or maybe position survey mode?
Second field - characters "A" and "O". Can someone imagine that "O" means "Operational"?
Third field - characters "W", "I", "T", "H". Maybe "W" is for "Warmup" or "Waiting", "T" for "Timing", and "H" is for "Holdover"? What may momentary "I" mean then?

[PA0PBZ]

Dit you see this?

https://www.eevblog.com/forum/testgear/bg7tbl-gpsdo-master-reference/msg741902/#msg741902

Same STATUS=3,O,T so what about the other commands there like inv; ?

[Velund]

Dit you see this?

https://www.eevblog.com/forum/testgear/bg7tbl-gpsdo-master-reference/msg741902/#msg741902

Same STATUS=3,O,T so what about the other commands there like inv; ?

Thank you for reference. Yes, it is the same command set... Below is responces of unit in holdover for 1.5 hours, seening no sats but with antenna (in foil) connected...


inv;

INV=NECGPS,057706,003xxx,00,057706,0172,,,6920N02442,0169;

Based on information from another thread:
INV=a,b,c,d,e,f,g,h,i;<cr><lf>
a: Name of the module (Max 12 char.) : NECGPS
b: Article number (Max. 6 characters) : 057706
c: Serial Number (Max: 6 characters) : 003xxx (last digits removed here)
d: Hardware version (Max. 2 characters): 00
e: Firmware article number (6 characters) : 057706
f: Firmware version (4 characters) : 0172
g: Date of test (format : DD/MM/YYYY) : <missing>
h: Version of test system (Max. 4 characters) : <missing>
i : Oscillator's type ( Max. 10 characters) : 6920N02442
j: FPGA version (4 characters) : 0169

conf;

CONF=0,0,A,+00:00,+0;
alarm;

ALARM=2;
gps_time;

GPS_TIME=1880,139398,01.07.15,00:00:00,00,17;
info_track_sat;

INFO_TRACK_SAT=0;
status;

STATUS=1,O,H;
temperature;

TEMPERATURE=+39.67;
type;

TYPE=4500,OEM_01;
atdc_status;

ATDC_STATUS=0,0;
hold_perf_status;

HOLD_PERF_STATUS=0;
info_vis_sat;

INFO_VIS_SAT=0;