Author Topic: bg7tbl gpsdo master reference  (Read 367635 times)

0 Members and 2 Guests are viewing this topic.

Online edpalmer42

  • Super Contributor
  • ***
  • Posts: 2260
  • Country: ca
Re: bg7tbl gpsdo master reference
« Reply #75 on: August 24, 2015, 07:22:10 am »
Time for a sanity check people!

Moto has confirmed that the frequency is off by 1.5e-11.  That's 0.000 146 Hz on a 10 MHz signal.

A Rubidium oscillator straight from the factory is typically adjusted to within 5.0e-11.  Lightages will see his FE-5680B move this much just from temperature changes.  A Rubidium's monthly drift will be that much or more.

An HP/Symmetricom 5071A Cesium standard when averaged over a 1 sec. period is only stable to 1.2e-11.

It's one thing to look at two units and say that one is better because the 'bad' unit has this error.  It's damned annoying to find that something that's supposed to be dead on 10 MHz isn't!  It would probably cause you to choose a different unit if you're shopping for one.  But it doesn't mean that the unit is trash.  As I've said before, the BG7TBL unit will provide good service for the vast majority of users.  If you've already got one it'd be hard to find a situation where this unit can't do the job.

If I had one of these units I'd file this info away in the back of my head and carry on.  If necessary, I'd make a correction in any measurements to compensate for the error.  But such a correction would be very rare.  I have many other things I'd rather do than try to reverse engineer something like this to correct such a small error.  Don't even consider it unless you're a PLL guru who can design digital PID controllers in your sleep!

Ed
 

Offline Lightages

  • Supporter
  • ****
  • Posts: 4313
  • Country: ca
  • Canadian po
Re: bg7tbl gpsdo master reference
« Reply #76 on: August 24, 2015, 07:38:08 am »
Oh I wasn't going to do anything to mine. Just like you said if I had known before I purchased it I would have chosen the other. I can't use better accuracy than it has even with the error.

And yes, I was playing around today and I can move the frequency of the FE5680B with cooling fan speed. They are so close in frequency that I can make it faster or slower just by blowing different amounts of air over the case of the 5680.
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #77 on: August 24, 2015, 04:01:43 pm »
Time for a sanity check people!

You are right, of course. I guess this is why they call the mailing list "time nuts". I still think it would be interesting to develop the firmware for one of these things. Maybe BG7TBL could even be persuaded to give up the source code. Of course, I need another "project" like I need a hole in the head.

BTW, is there any good reference on now these units work? Is it just counting the 10M osc cycles in a 1 PPS interval from the GPS, subtracting that from the target value (10M) to derive an error signal that drives a PID loop controlling the EFC voltage?

I bet that Altera MAX CPLD is doing that count of 10M cycles in 1 PPS intervals. I wonder how one deals with metastability issues.

Or maybe it is working more like a PLL, with a phase comparator...
« Last Edit: August 24, 2015, 04:15:14 pm by motocoder »
 

Online edpalmer42

  • Super Contributor
  • ***
  • Posts: 2260
  • Country: ca
Re: bg7tbl gpsdo master reference
« Reply #78 on: August 24, 2015, 06:22:11 pm »
Time for a sanity check people!

You are right, of course. I guess this is why they call the mailing list "time nuts". I still think it would be interesting to develop the firmware for one of these things. Maybe BG7TBL could even be persuaded to give up the source code. Of course, I need another "project" like I need a hole in the head.

BTW, is there any good reference on now these units work? Is it just counting the 10M osc cycles in a 1 PPS interval from the GPS, subtracting that from the target value (10M) to derive an error signal that drives a PID loop controlling the EFC voltage?

I bet that Altera MAX CPLD is doing that count of 10M cycles in 1 PPS intervals. I wonder how one deals with metastability issues.

Or maybe it is working more like a PLL, with a phase comparator...

We know nothing about Mr. BG7TBL.  He's obviously very talented.  We don't know if he just made a mistake or if he's operating at the edge of his knowledge level.  Is it significant that he's started producing units that reuse existing GPSDOs rather than his own?  Or was that just because those other units are available?  No way to tell.

I don't remember a specific low-level writeup on how GPSDOs work.  There are lots of amateur projects around.  Some use digital technology, others use analog.  Google for "Brooks Shera".  His was one of the first hobbyist GPSDOs that came out.  Maybe it was the first.  He passed away a few years ago, but his source code was saved and has been released.  James Miller came up with a very well known GPSDO that's based on very simple analog technology and performs very well.  I would think that any training material on FLL or PLL systems would be useful.

A few years ago I had an eerily similar situation where a commercial GPS receiver had a 10 MHz output that was about 1.5e-11 off frequency.  It turned out that the firmware implemented an FLL rather than a PLL.  The vendor had a PLL version in his archives that solved the problem.

Ed
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #79 on: August 24, 2015, 07:26:17 pm »
Time for a sanity check people!

You are right, of course. I guess this is why they call the mailing list "time nuts". I still think it would be interesting to develop the firmware for one of these things. Maybe BG7TBL could even be persuaded to give up the source code. Of course, I need another "project" like I need a hole in the head.

BTW, is there any good reference on now these units work? Is it just counting the 10M osc cycles in a 1 PPS interval from the GPS, subtracting that from the target value (10M) to derive an error signal that drives a PID loop controlling the EFC voltage?

I bet that Altera MAX CPLD is doing that count of 10M cycles in 1 PPS intervals. I wonder how one deals with metastability issues.

Or maybe it is working more like a PLL, with a phase comparator...

We know nothing about Mr. BG7TBL.  He's obviously very talented.  We don't know if he just made a mistake or if he's operating at the edge of his knowledge level.  Is it significant that he's started producing units that reuse existing GPSDOs rather than his own?  Or was that just because those other units are available?  No way to tell.

I don't remember a specific low-level writeup on how GPSDOs work.  There are lots of amateur projects around.  Some use digital technology, others use analog.  Google for "Brooks Shera".  His was one of the first hobbyist GPSDOs that came out.  Maybe it was the first.  He passed away a few years ago, but his source code was saved and has been released.  James Miller came up with a very well known GPSDO that's based on very simple analog technology and performs very well.  I would think that any training material on FLL or PLL systems would be useful.

A few years ago I had an eerily similar situation where a commercial GPS receiver had a 10 MHz output that was about 1.5e-11 off frequency.  It turned out that the firmware implemented an FLL rather than a PLL.  The vendor had a PLL version in his archives that solved the problem.

Ed

One of my coworkers has emailed and received responses from BG7TBL. I'll batch up my questions and see if he can get some answers.

I found the Shera articles on my own after posting this; he has an article on his website as well as a link to a PDF of a QST article. Good articles. I see now how it is able to detect phase differences < 1 cycle. At least for Shera, he has something running at a frequency higher than 10MHz to do that. I guess you could also do it with the usual XOR or R/S flip-flop and some sort of filter to turn that into an analog voltage, and then sample that at whatever rate suits your fancy.

BTW - I have been reading up on the UBlox modules, and they actually have a module that will discipline an external frequency reference directly. It has an output you can hook to a DAC, and a frequency input, and some configuration you write to tell it what the frequency is, what the range and gain is of the frequency adjustment, what the stability of the reference is, etc. I wonder how well that works? Might make for an extremely compact GPSDO.
« Last Edit: August 25, 2015, 05:14:09 am by motocoder »
 

Offline usagiTopic starter

  • Frequent Contributor
  • **
  • Posts: 390
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #80 on: August 25, 2015, 08:47:38 am »
i received some documentation from ocilloquartz for star4. looks like there we should have rs232 access to the controller. i'm going to verify if it's ok to redistribute.

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #81 on: August 25, 2015, 06:06:49 pm »
I was off in my calculation on the error of the BG7TBL/2014-12-09. I misinterpreted the Timelab graph as showing degrees, when in fact it was showing seconds. So it is actually slipping a cycle every 1800 seconds. This works out to an effective frequency of 9,999,999.999444

Not a huge difference, but I wanted to correct the record.
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #82 on: August 25, 2015, 06:31:19 pm »
Here are some Timelab graphs showing the Lucent KS-24361 versus the latest BG7TBL attempt as well as the two 3rd-party units he is putting his logo on.
 

Offline dadler

  • Supporter
  • ****
  • Posts: 851
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #83 on: August 25, 2015, 06:39:12 pm »
Hmm-I bought the BG7TBL/2014-12-09 because it was the most-recommended.

What is the now-most-recommended? I haven't even had a chance to power on the BG7TBL/2014-12-09 unit  :-\
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #84 on: August 25, 2015, 07:31:05 pm »
Hmm-I bought the BG7TBL/2014-12-09 because it was the most-recommended.

What is the now-most-recommended? I haven't even had a chance to power on the BG7TBL/2014-12-09 unit  :-\

Depends on your use for it. There's nothing wrong with it unless  a frequency difference of 0.0006 Hz causes you problems. FOr 99.99999% of the people out there, this is a non-issue. It is a nice compact unit that works just fine.
 

Offline dadler

  • Supporter
  • ****
  • Posts: 851
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #85 on: August 25, 2015, 08:10:30 pm »
Hmm-I bought the BG7TBL/2014-12-09 because it was the most-recommended.

What is the now-most-recommended? I haven't even had a chance to power on the BG7TBL/2014-12-09 unit  :-\

Depends on your use for it. There's nothing wrong with it unless  a frequency difference of 0.0006 Hz causes you problems. FOr 99.99999% of the people out there, this is a non-issue. It is a nice compact unit that works just fine.

This is certainly a want vs need situation. My use is only in calibrating my frequency counters and rubidium oscillators, and synchronizing my other equipment. In that regard, The Lucent I have and the BG7TBL/2014-12-09 are likely perfectly fine for this use case. However, I want the "best" (a balanced combination of most accurate/stable within a reasonable price bracket).
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #86 on: August 25, 2015, 08:17:45 pm »
Hmm-I bought the BG7TBL/2014-12-09 because it was the most-recommended.

What is the now-most-recommended? I haven't even had a chance to power on the BG7TBL/2014-12-09 unit  :-\

Depends on your use for it. There's nothing wrong with it unless  a frequency difference of 0.0006 Hz causes you problems. FOr 99.99999% of the people out there, this is a non-issue. It is a nice compact unit that works just fine.

This is certainly a want vs need situation. My use is only in calibrating my frequency counters and rubidium oscillators, and synchronizing my other equipment. In that regard, The Lucent I have and the BG7TBL/2014-12-09 are likely perfectly fine for this use case. However, I want the "best" (a balanced combination of most accurate/stable within a reasonable price bracket).

I personally like the Huawei (Star4) unit. There is some updated info coming on that one shortly that I think will make it even more attractive.
 

Online Vgkid

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #87 on: August 25, 2015, 08:26:18 pm »
^^^ Speaking of that...
Usagi, have you heard back from them. I'm still waiting.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline usagiTopic starter

  • Frequent Contributor
  • **
  • Posts: 390
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #88 on: August 25, 2015, 10:36:24 pm »
oscilloquartz sent me some basic technical documentation, and gave the OK to release it.

this product is discontinued, this document is provided as-is with no warranty. oscilloquartz cannot provide technical support for this product, so please don't bug them  8)

Offline radioFlash

  • Regular Contributor
  • *
  • Posts: 163
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #89 on: August 26, 2015, 02:19:38 am »
I have the 2015-07-06 / Huawei version.  I've tried the serial connection through Putty at 4800/8/n/1 as well as other baud rates, but don't get anything intelligible (looks like binary data, not ASCII). What are the correct serial port settings, or should I not expect ASCII?


Datasheet for the Ublox receiver:
http://www.amtechs.co.jp/2_gps/pdf/LEA-5T.pdf
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #90 on: August 26, 2015, 02:20:01 am »
oscilloquartz sent me some basic technical documentation, and gave the OK to release it.

this product is discontinued, this document is provided as-is with no warranty. oscilloquartz cannot provide technical support for this product, so please don't bug them  8)

Thank you for working with them to get this. I can confirm that the TOD output works - 4800/8/N/2, Pin 6B on the connector just as it says in the docs. I need to switch one wire over and I can check the management interface.
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #91 on: August 26, 2015, 02:23:50 am »
I have the 2015-07-06 / Huawei version.  I've tried the serial connection through Putty at 4800/8/n/1 as well as other baud rates, but don't get anything intelligible (looks like binary data, not ASCII). What are the correct serial port settings, or should I not expect ASCII?


Datasheet for the Ublox receiver:
http://www.amtechs.co.jp/2_gps/pdf/LEA-5T.pdf

radioFlash - the default wiring on that unit is to expose the serial transmit from the UBlox GPS chip on the board. You can install the u-center application from UBlox to view that (it is binary) - it's using 9600/8/N/1. However, it's a read-only interface, which u-center isn't expecting, so some things don't work right.

If you want to do a little soldering, you can swap out that wire for one of the connections to the 26-pin connector on the Oscilloquartz board inside. This will let you either receive the TOD output at 4800 baud, or (in theory - not tested yet) communicate with the board's management interface (9600/8/N/2). This connector and the two serial interfaces are discussed in the manual that usagi attached to his post a few up from this one.
 

Offline radioFlash

  • Regular Contributor
  • *
  • Posts: 163
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #92 on: August 26, 2015, 02:30:03 am »

radioFlash - the default wiring on that unit is to expose the serial transmit from the UBlox GPS chip on the board. You can install the u-center application from UBlox to view that (it is binary) - it's using 9600/8/N/1. However, it's a read-only interface, which u-center isn't expecting, so some things don't work right.

If you want to do a little soldering, you can swap out that wire for one of the connections to the 26-pin connector on the Oscilloquartz board inside. This will let you either receive the TOD output at 4800 baud, or (in theory - not tested yet) communicate with the board's management interface (9600/8/N/2). This connector and the two serial interfaces are discussed in the manual that usagi attached to his post a few up from this one.

Thanks for clearing it up--I'll have a try with u-center for now.
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #93 on: August 26, 2015, 02:44:23 am »
Management interface works too. It's very picky about end-of-line characters, which is a PITA when using Putty because it does not give you control over what it sends. You need to send "<command>;<cr><lf>".

Here's the response to the INV command:

INV=GPS STAR 4+,015880,1057,01,015881,0149,11022010,0010,8663-XS,0152;
This is:


INV=a,b,c,d,e,f,g,h,i;<cr><lf>
a: Name of the module (Max 12 char.) : GPS STAR 4+
b: Article number (Max. 6 characters) : 015880
c: Serial Number (Max: 6 characters) : 1057
d: Hardware version (Max. 2 characters): 01
e: Firmware article number (6 characters) : 015881
f: Firmware version (4 characters) : 0149
g: Date of test (format : DD/MM/YYYY) : 11/02/2010
h: Version of test system (Max. 4 characters) : 0010
i : Oscillator's type ( Max. 10 characters) : 8663-XS
j: FPGA version (4 characters) : 0152


Please post your units response to this command if you wire up the management interface. I'd like to see what range of firmware/hardware versions are out there.
« Last Edit: August 26, 2015, 02:54:16 am by motocoder »
 

Online Vgkid

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: bg7tbl gpsdo master reference
« Reply #94 on: August 26, 2015, 02:59:50 am »
Thanks for the info usagi et al.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #95 on: August 26, 2015, 03:01:43 am »
Another useful command: TEMPERATURE

This is probably the temperature from the FPGAs internal sensor. Very useful to see if things are overheating inside that non-ventilated BG7TBL enclosure. Mine is out of the enclosure at the moment, but I'll run it later when it's all buttoned up and see what it says. Out of the enclosure, it's reading +40.72.
 

Online edpalmer42

  • Super Contributor
  • ***
  • Posts: 2260
  • Country: ca
Re: bg7tbl gpsdo master reference
« Reply #96 on: August 26, 2015, 04:43:32 am »
Here are some Timelab graphs showing the Lucent KS-24361 versus the latest BG7TBL attempt as well as the two 3rd-party units he is putting his logo on.

Very interesting!

If you ran the Allen Deviation graphs longer you'd see more of the behaviour of the units.  Notice how they all line up on the left side of the graph?  That's the effect of your 53131A counter.  It's only when they deviate from that line that you see the interaction of the GPSDOs.

The phase graph really shows the frequency error doesn't it?  If you hit the 'r' key, TimeLab will take a linear best fit of each graph and then subtract it from the graph.  The frequency offset will disappear and you'll see the short-term comparison of the units.  This would give you an idea what would happen if you used the 10 MHz as a reference for your counter, generator, etc.

Ed
 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #97 on: August 26, 2015, 05:36:13 am »
Here are some Timelab graphs showing the Lucent KS-24361 versus the latest BG7TBL attempt as well as the two 3rd-party units he is putting his logo on.

Very interesting!

If you ran the Allen Deviation graphs longer you'd see more of the behaviour of the units. 

I tried to run the Allan Deviation for longer, but for some reason Timelab just stops capturing after some time. It's not the value I set in the Acquisition settings, it stops even if I set it to only stop on Manual Termination.

Notice how they all line up on the left side of the graph?  That's the effect of your 53131A counter.  It's only when they deviate from that line that you see the interaction of the GPSDOs.

I suspected this. I even tried calibrating some of the other aspects of the counter like  offset, gain, and the Time Interval calibration. I think the offset and gain calibrations were fine, but the time interval calibration made things worse. This counter has a "quick" mode of calibration of that which can be done without any fancy equipment, but the fine mode requires things I don't have. So by doing the quick cal, I think I actually made it worse.

The phase graph really shows the frequency error doesn't it?  If you hit the 'r' key, TimeLab will take a linear best fit of each graph and then subtract it from the graph.  The frequency offset will disappear and you'll see the short-term comparison of the units.  This would give you an idea what would happen if you used the 10 MHz as a reference for your counter, generator, etc.
Ed

Take a gander at the updates on the Huawei/Star4+ unit above. This is looking like a really good option now that we have a management interface working.

 

Offline motocoder

  • Supporter
  • ****
  • Posts: 769
  • Country: us
  • Electrical Engineer
Re: bg7tbl gpsdo master reference
« Reply #98 on: August 26, 2015, 06:16:43 am »
Reported temperature of the Huawei, in the enclosure, sitting on top of the heat-generating Lucent, is 51~52C.
 

Offline TSL

  • Regular Contributor
  • *
  • Posts: 243
  • Country: au
Re: bg7tbl gpsdo master reference
« Reply #99 on: August 26, 2015, 06:30:29 am »

I don't remember a specific low-level writeup on how GPSDOs work. 

The Trimble Thunderbolt manual has some reasonable light info on how it works and mentions Allan Variance and Kalman Filtering in section 5 of its op manual here...

ftp://ftp.trimble.com/pub/sct/embedded/bin/Manuals/ThunderBoltBook2003.pdf

A much deeper discussion can be found from NIST here...

"The Use of GPS Disciplined Oscillators as Primary Frequency Standards for Calibration and Metrology Laboratories"

http://tf.nist.gov/general/pdf/2297.pdf

regards

Tim
VK2XAX :: QF56if :: BMARC :: WIA :: AMSATVK
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf