($40 GPSDO working well now, my mistake)

[tkamiya]

'Average' of what? Compared against what?

With two, I can show which one is running faster than the other.  There is nothing to base my decision on which one is more accurate.  With three, I can average.  Unless one is an obvious out-lyre truth is likely to be in the middle somewhere.  It's average of all 3.  Compared against each other.

[cdev]

How could one show if one is running 'faster'? How do you derive that info?  (Lady Heather? But it connects to the GPSDO via the UART?)


There are still major gaps in my knowledge here.

Even the largest excursions of jitter that I know must be occurring are still so small I haven't figured out how to visualize or measure them yet.

Maybe get a really really good free running oscillator like a rubidium frequency standard that had recently been calibrated - using a better frequency standard than itself, and compare them to that?

I would need some hardware which I don't currently have, plus the oscillator, to do that.

[tkamiya]

I had both GPSDO connected to a scope.   Synch was on channel 2.  Over few hours, it moved with some recognizable amount.  Since Channel 2 was steady (trigger) and channel1 was moving to the left, I could deduce frequency on channel 1 was lower.

This "just trust us.  it's accurate" thing bothers me.  I want to quantify it.

I know there is a way to do this on counters, and I have one that is capable.  I have a significant lack of knowledge.  This is an interesting challenge.  I have a few "signal analyzer" type equipment and pen chart (really pen and paper).  My goal is to keep logging short term changes for long term.

I understand rubidium does NOT do better than OCXO on short term drift.  Maybe we just have to buy us Cesium clock for Christmas. 

[cdev]

How long has they been on, and were they both connected to the same antenna?

How do you divide the signals?

Do you mean one 10 MHz output shifted it's phase relative to another?

[tkamiya]

Antenna is in attic.  Short Andrew Helix connects it to 4 channel Spectracom distribution unity gain amp.  (so basically no signal loss)  Port 1 goes to Nortel GPSTM.  Port 2 goes to another GPSTM.  Port 3 goes to this board in discussion.  Port 4 is terminated.  Earlier, I posted pictures of my setup.

Yes that's exactly what I mean.  Since signal 1 is a triggering signal, it is always on the same spot.  Signal 2 is what the signal was happened to be doing when 1 got triggered.  That signal moves to left.  Wavelength is shortening := frequency is higher.

My local cal lab is not able to cal the time standard.  I'm going to have to find a way to get this done.  With too many assumptions and supposed-to-bes, I am getting confused.

[cdev]

It should become stable after the GPSDO has been on for a while. I would connect one to Y and one to y, so you had lissajous figures. Since both are getting the same signal the phase shouldn't shift after its finished warming up. This may take a day or so but if you are getting a decent signal not that long.

Thats a nice power supply you are using in your unit. Do you mind if I asked where you got it and also if it generates a lot of heat?  I want a dial 12v 5 volt unit in a low form factor.

[tkamiya]

Problem with Lissajous is you can't tell which one is ahead.  So I briefly tried it and switched back to y(t).

The power supply is a little unit I bought on Amazon.  "Mean Well" RS25-12.  12V 2Amp unit and voltage is actually adjustable.  8 dollars I think.  I think the meaning of the words kind of got lost in translation though.  In the box, I also have DC/DC converter from Amazon for 6 bucks.  I did this with intention of feeding as little voltage as possible to the Arduino board.  It has a 3 terminal onboard regulator and it and the CPU gets pretty hot if you feed 12v directly.  I might just take that out because I don't need it with Lady Heather.

It's hard to tell from from images but Mean Well directly attached to the bottom cover which is 1.5mm aluminum.  With few hours of running I could not tell location of the power supply by touch alone.

Funny you are asking me these question because I was JUST thinking about the same thing.  I don't like external power supply solution, so that was OUT.  But my Oven is running 50 degrees C.  I doubt very little external heat reaches the element.

What I'm going to do is to purchase a very short rack, in size of foot rest, and stick it under the desk.  I can see quite a bit of activity when I open my door and when A/C kicks in.  So its current location is not ideal.

[tkamiya]

I left the comparison running overnight.  Phase shifted for about 60 degrees and stopped.  Lock seems to be solid as well.  Total of 121 seconds of hold over, which I think happened in earlier part of test.

Few nights ago, I've found a symmetricon distribution amp with redundant input on eBay.  Port A is a master, and if it fails, switches to port B.  Bought that.  I think it will be a very nice overkill for my top secret government lab.

[LapTop006]

Few nights ago, I've found a symmetricon distribution amp with redundant input on eBay.  Port A is a master, and if it fails, switches to port B.  Bought that.

I saw that one go past.

I think it will be a very nice overkill for my top secret government lab.

That sort of thinking is why I'm building my own custom pods for the Spectracom 8140 distribution system.

[tkamiya]

That's what I'm using now:  Symmetricon 8140.  What kind of customization did you do on remote modules?

[cdev]

The Extron video distribution amps will amplify 3x10 Mhz (R,G+B) per output as well as a sync "TTL" (1PPS) signal.

Mine works well with the GPSDO 10 Mhz fed into one of the three channels and the PPS fed into the "S" So that gives me three 10 MHz outputs and three 1PPS outputs.

I have not tried splitting the 10 MHz to send to all three channels (I will probably use a CATV splitter to do that) at the same time although I don't see any reasons it would not work except maybe the output being lower.

Its cheap, built well and it does what's required of the job so far.

------

I still am undecided about how I want to handle the power for my TruePosition.
The problem as I see it is the thermal influence on the GPSDO of additional heat sources in the same box.

The GPSDO's own voltage regulation is supposed to be quite good. But because its a valuable and fun piece of gear for me, I want my case implementation to protect it as best as I can and provide an optimum thermal environment for it as far as stability goes. Certainly a far better one than a metal box sitting on a telephone pole.

I want to use a 12v voltage regulator to protect it from overvoltage. Normally this will likely be working with just slightly over 12v so I am not expecting it to get hot, but I still want it's heatsink to be oversized and separated from the GPSDO thermally as best as I can. Ideally the heatsink would be on the outside of the box but another alternative is putting the gpsdo in a box inside of the case and doing the same with the power supply and experimenting with what fan setup would give the best balance between the various thermal factors. Having a horizontally oriented case where there isnt much convection would seem to indicate a need for venting for intake somewhere and a small exhaust fan on the power supply side.

My house probably varies more in temperature throughout the year than most, intentionally, but its still nothing near as extreme as a metal box on a pole so I think I will be okay, as long as the temperature variations are within reason and consistent, I think the TP should be able to handle it with its learning algorithms.

I just don't want to confuse it with additional heat.

[CJay]

This "just trust us.  it's accurate" thing bothers me.  I want to quantify it.
I understand rubidium does NOT do better than OCXO on short term drift.  Maybe we just have to buy us Cesium clock for Christmas.

I think what you need is this:

http://www.wriley.com/3-CornHat.htm

Texaspyro will know better than I do.

However, if you want to buy me a Cesium clock, I can PM my address for delivery 

[metrologist]

Antenna is in attic.  Short Andrew Helix connects it to 4 channel Spectracom distribution unity gain amp.  (so basically no signal loss)  Port 1 goes to Nortel GPSTM.  Port 2 goes to another GPSTM.  Port 3 goes to this board in discussion.  Port 4 is terminated.  Earlier, I posted pictures of my setup.

Yes that's exactly what I mean.  Since signal 1 is a triggering signal, it is always on the same spot.  Signal 2 is what the signal was happened to be doing when 1 got triggered.  That signal moves to left.  Wavelength is shortening := frequency is higher.

My local cal lab is not able to cal the time standard.  I'm going to have to find a way to get this done.  With too many assumptions and supposed-to-bes, I am getting confused.

This is my experience with my two UCCM units, only the drift was systematic. I do not recall now if the cycle was twice/day or every two hours... Over a period of a few minutes, one wave would shift ~+180° and then track back to ~-180° and then go back in phase until the cycle would repeat. Since I only have two units, I could not determine which one is moving, if they are both doing some sort of fix at the same time. I'm on a quest just to verify that, but one of my units does not show the same oscillator info and lock indicator in LH. I am planning on building two more units based on Lars DIY GPSDO.

[LapTop006]

That's what I'm using now:  Symmetricon 8140.  What kind of customization did you do on remote modules?

Spectracom, not Symmetricom.

Haven't built them yet, ongoing project.

The TL;DR is I'm building some multi-output pods using modern components, since I have a few pieces of kit that do nasty things on their reference input (ex, permanently terminated, or the ref is combo in/out and may reset unexpectedly) and while the pods are cheap, by the time you add shipping to Australia they stop being so.

If I had a working kicad I'd have already ordered some test boards for the bias tees for circuit testing.

[cdev]

I wonder what the expected lifetime of a telecom GPSDO is, and whether some GPSDO units are removed from service because of failing, or appearing to soon be in danger of failing, some calibration tests?

They could probably tell exactly how much aging has occurred and how much longer they will remain in spec.

Once they go outside of the range they are designed to correct I would suspect that one might need to replace the OCXO in them with an identical newer unit to get them working again.

This is my experience with my two UCCM units, only the drift was systematic. I do not recall now if the cycle was twice/day or every two hours... Over a period of a few minutes, one wave would shift ~+180° and then track back to ~-180° and then go back in phase until the cycle would repeat. Since I only have two units, I could not determine which one is moving, if they are both doing some sort of fix at the same time. I'm on a quest just to verify that, but one of my units does not show the same oscillator info and lock indicator in LH. I am planning on building two more units based on Lars DIY GPSDO.

[tkamiya]

I'd imagine our type of use and environment is not what designer had in mind.  I'm sure we won't survive long in a telecom cabinet. 

[edpalmer42]

I wonder what the expected lifetime of a telecom GPSDO is, and whether some GPSDO units are removed from service because of failing, or appearing to soon be in danger of failing, some calibration tests?

They could probably tell exactly how much aging has occurred and how much longer they will remain in spec.

If you look at the OCXO tuning voltage, you can get an idea of how much 'life' remains in the OCXO.  Measure it - either with a DMM or by talking to the unit and plot it on a graph vs. time.  Maybe once a day for a few weeks.  Do curve-fitting and see approx. how long it will take to get to the limit.

Once they go outside of the range they are designed to correct I would suspect that one might need to replace the OCXO in them with an identical newer unit to get them working again.

If your GPSDO lets you change the PLL parameters you can 'tune' it to support a wide variety of OCXOs.  Most don't have this ability.  One I know of that does is the Trimble Thunderbolt.

This is my experience with my two UCCM units, only the drift was systematic. I do not recall now if the cycle was twice/day or every two hours... Over a period of a few minutes, one wave would shift ~+180° and then track back to ~-180° and then go back in phase until the cycle would repeat. Since I only have two units, I could not determine which one is moving, if they are both doing some sort of fix at the same time. I'm on a quest just to verify that, but one of my units does not show the same oscillator info and lock indicator in LH. I am planning on building two more units based on Lars DIY GPSDO.

As mentioned by CJay, the Three-Cornered Hat measurement will help to determine what's really going on.  However, to get usable data, you have to make three measurements simultaneously.  That's okay.  You wanted to buy more hardware didn't you?   

[cdev]

It seems from what I have read about the unique properties of the PIC on leapsecond.com that this could be done in a very affordable manner for almost nothing - using the synchronous timing function of the PIC without introducing any additional error.

Caveat is I am considering a PIC hardware solution that runs using one clock (like a 10MHz GPSDO reference) and then can compare it to two additional devices. So total of three. Am I correct in that that is enough? (reference oscillator and two additional DUTs)

As mentioned by CJay, the Three-Cornered Hat measurement will help to determine what's really going on.  However, to get usable data, you have to make three measurements simultaneously.  That's okay.  You wanted to buy more hardware didn't you?   

[edpalmer42]

It seems from what I have read about the unique properties of the PIC on leapsecond.com that this could be done in a very affordable manner for almost nothing - using the synchronous timing function of the PIC without introducing any additional error.

Caveat is I am considering a PIC hardware solution that runs using one clock (like a 10MHz GPSDO reference) and then can compare it to two additional devices. So total of three. Am I correct in that that is enough? (reference oscillator and two additional DUTs)

That wouldn't give you enough data.  Normally, you'd use three TICs to measure time intervals between units A-B, B-C, and A-C and then process the results.  You need all three comparisons to solve the equations.  One (or two) units could be a Rb standard or a good, stable OCXO.

I'd be looking at something inexpensive like a PICTIC II or an updated version of it.  Even that would cost $150 - $200 for three units.  Anything beyond that is likely > $500.  Of course, if you have or can borrow one or more commercial TICs, that makes it easier to get the three units.  Higher performance is nice, but I wouldn't worry too much about that.  The PICTIC II has a resolution of < 1ns.  You can make some nice measurements at that level.

[cdev]

Thanks!

Why is the PIC based hardware so expensive?

[edpalmer42]

Thanks!

Why is the PIC based hardware so expensive?

You mean for the PICTIC II?  The cost of $150 - $200 is for all three units.  When it was originally designed, the goal was to keep the price <=$50 per unit, but that was back in 2010/2011.  Recently the concept was updated with the creation of the TICC from TAPR.  But the performance was also improved.  End result is a very nice unit with a minimum resolution of 60 ps and a cost of $190 each.

[cdev]

That's what I'm using now:  Symmetricon 8140.  What kind of customization did you do on remote modules?

Spectracom, not Symmetricom.

Haven't built them yet, ongoing project.

The TL;DR is I'm building some multi-output pods using modern components, since I have a few pieces of kit that do nasty things on their reference input (ex, permanently terminated, or the ref is combo in/out and may reset unexpectedly) and while the pods are cheap, by the time you add shipping to Australia they stop being so.

If I had a working kicad I'd have already ordered some test boards for the bias tees for circuit testing.

If you can make a delicate conical inductor out of thin enameled wire and gently attach the wire coming out of its tip to the signal path trace you can make a great bias tee that can handle a very wide range of frequencies in less than half an hour. Because a conical inductor is not resonant at any single frequency. Try it!

[cdev]

Thats great and I'm sure its competitive with pro gear and support TAPR etc. But what if I wanted to make the exact same thing using my own PCB that I had made at OSH or similar. What would all the parts cost if I bought them from LCSC when appropriate or here, and put a lot of energy into not spending even an extra penny that wasnt needed? And used my own Arduino clone from ebay. Or something else. Whatever worked.

Thanks!

Why is the PIC based hardware so expensive?

You mean for the PICTIC II?  The cost of $150 - $200 is for all three units.  When it was originally designed, the goal was to keep the price <=$50 per unit, but that was back in 2010/2011.  Recently the concept was updated with the creation of the TICC from TAPR.  But the performance was also improved.  End result is a very nice unit with a minimum resolution of 60 ps and a cost of $190 each.

[edpalmer42]

Thats great and I'm sure its competitive with pro gear and support TAPR etc. But what if I wanted to make the exact same thing using my own PCB that I had made at OSH or similar. What would all the parts cost if I bought them from LCSC when appropriate or here, and put a lot of energy into not spending even an extra penny that wasnt needed? And used my own Arduino clone from ebay. Or something else. Whatever worked.

Thanks!

Why is the PIC based hardware so expensive?

You mean for the PICTIC II?  The cost of $150 - $200 is for all three units.  When it was originally designed, the goal was to keep the price <=$50 per unit, but that was back in 2010/2011.  Recently the concept was updated with the creation of the TICC from TAPR.  But the performance was also improved.  End result is a very nice unit with a minimum resolution of 60 ps and a cost of $190 each.

I'm getting confused about which unit you're talking about.  But I don't know how much it would cost to build either from scratch.  All the info is available for the PICTIC II.  It's just jelly bean components.

Most of the info is available for the TICC.  I looked around, but didn't find the board layout.  Maybe I didn't look far enough.

You'll have to figure out your costs.

For anyone playing along at home, here are the links to the PICTIC II and the TAPR TICC.

http://www.ko4bb.com/doku2015/doku.php?id=precision_timing:pictic  NOTE:  This page includes info on the PICTIC and the improved PICTIC II.  Make sure you're looking at the right info.

https://www.tapr.org/kits_ticc.html

[cdev]

Major cost reductions in as many essential or game changing things as possible - as well as the basics, shouldn't just be an obsession or a target, it should become a principle we promote and implement with a fervor- like we implement aggressive cost inflation now.

Our goal should be making it possible for everybody to have a decent standard of living and security on as little money as possible. Not make poor people die earlier from stress.

How many lives does it save to make taxpayer developed cancer drugs that don't actually improve on existing ones at all, more expensive by evergreening patents and eliminating alternatives? None. It kills poor people. It steals away lives and children's educations.