New or used GPS Frequency Standard?

[FriedMule]

My reason for writing "below" $1000-$1500 is that this is my upper limit. Why that high price? Since I have no idea on what to get or what to look for and I have seen $30,000 units would I make sure it did not go way out of what I can afford. Just take a look at a bench DMM 5½ is fairly cheap, 6½ start to hurt a lot and 7½ / 8½ would kill me. :-) For the same reason did I think a 10E6 would be cheap a 10E12 maybe start to hurt and 10E14 would cost every one of my organs.

This is what I can get for about $250 and it may be fine to a start, but why trow out $250 if I find out I have to buy another unit 6 months from now, because it's not even as accurate as the crystals in my scope? :-)

[tkamiya]

How can I get you to see you won't actually get 10^-14?  To get anywhere close to 10^-13 level, you'll have to measure for 24 hours, log the data, and take a mean value of it.  Then maybe you'll get that kind of accuracy.  It's far more realistic to think, measure for 10 minutes and take an average, and get 10^-12ish, if that.  If you just hook up and take a single shot measurement, you could get 10^-10ish. 

Be careful with HP branded ones from China at rock bottom price.  They are sometimes FrankenGPSDO.  Just because someone upgraded OXCO or GPS board, that won't give GPSDO more performance.  There's a lot of tuning involved, which is more like an art than science.  Also, they are almost certainly a surplus unit from GPS cell sites.  They didn't come from a clean cal lab.  Some of them were awfully abused. 

With all that said, it's your money and your lab.  You should get what you are satisfied with.  You asked our opinion.  Most of us have opinions the direction you are trying to go may be waste of money.  Most of us start like you, then quickly realize that's not the way it works.  It's interesting to play with various kind of GPSDOs since they are cheap.

In my lab, my GPSDO references are Nortel GPSTM for house standard, and T-bolt for secondary unit.  If I need exceptional short term stability, I calibrate an independent OCXO unit with GPSDO for few hours, and use OCXO as reference.  I also have Rubidium and Cesium, but later is rarely used. 

Just be forewarned....  you are about to get sucked into a black hole that will create a giant worm hole on your wallet. 

[MIS42N]

This is what I can get for about $250 and it may be fine to a start, but why trow out $250 if I find out I have to buy another unit 6 months from now, because it's not even as accurate as the crystals in my scope? :-)

I don't think you realise how good even a cheap GPSDO can be. There are two variables to deal with - stability and accuracy. A good crystal is stable but not necessarily accurate. GPS signals are accurate but not (short term) stable. The job of the GPSDO is to transfer the accuracy of GPS signals to an oscillator while trying not to transfer any short term instability. The quality of the local oscillator is the main determining factor.

If you are not spending thousands of dollars on each piece of equipment, almost certainly the GPSDO will be hundreds of times more accurate than any internal oscillator. I don't know the particular item in your picture, but it will definitely be better than anything you have except if you have another GPSDO.

As a quick example, I built a really cheap GPSDO (it has just the oscillator, a PIC16F1455 processor for disciplining, a 74HC04 to drive the output, a Ublox Neo-6M GPS module, and a handful of resistors and capacitors). It was tested earlier this week and shown to provide a 3V p-p square wave into 50 ohm with a 50/50 duty cycle and accurate to 1 part in 10E-9. It was able to show to 0.01Hz the inaccuracy of the laboratory frequency counter. Because of the distance from decent GPS reception I added a line driver/receiver so I can position the antenna and GPS 16 meters from the GPSDO (tested to 30 meters). And this is just a home brew system. Just about any commercial unit will be as good or better.

Regarding attaining an accuracy of 1 part in 10E14, I read that the best cesium beams can do this. I worked in a standards laboratory many years ago, IIRC the cesium beam is very stable in the medium term but can be slightly inaccurate due to environmental effects. The beam is measured by a signal derived from an ultra stable DOXCO which also provides the user output. Using the GPS over a period of weeks it is possible to measure the inaccuracy of the cesium beam and tune it, and also tune the DOXCO to match the beam. Part of the tuning of the DOXCO is to measure and compensate for ageing effects which, after a few months, are predictable.

Because the GPS constellations are locked to the NIST standard, very long term comparisons could theoretically bring an oscillator to better than 1 part in 10E14 accuracy. But in practice 10E14 is good enough for most people. 

[FriedMule]

i am aware of 10E14 being unrealistic for me, just like a 8½ bench DMM is fare outside anything I will ever get my hands on.
I know that the persute of the best accuracy can be a black hole but also super fun.

Some of the sugested units looks very interesting and cheap other models sounds as if they are best to avoid! :-)

[bingo600]

I can't add any relevant info that hasn't already been mentioned wrt. gpsdo choice.
Hope you get what you want, and will be happy with your choice.

/Bingo

[FriedMule]

I'd go for a Used Samsung like this one, good critic on time-nuts & here
https://www.ebay.com/itm/254609356856
https://www.ebay.com/itm/293878344715

Or a Trimble Thunderbolt (Time nut favorite)
Needs 5v, +12v og -12v
https://www.ebay.com/itm/332390729098

I use this 26dB Timing antenna  (N conn)
https://www.ebay.com/itm/402232021010

Since the suggested units is so cheap do I plan on buying the both of the above units. To learn and compare. :-)
But it looks like the Trimble does not have a power supply, any suggestion one known to good for the unit?

The Trimble antenna has a great pole mounting bracket, do any of you know what pole diameter fits?

[bingo600]

http://www.leapsecond.com/pages/tbolt/power.htm
http://www.leapsecond.com/tbolt-faq.htm
https://prc68.com/I/ThunderBolt.shtml

PSU Something like this - Meanwell is one of the better switchers
https://protosupplies.com/product/mean-well-power-supply5v-12v-12v-50w/



Edit :
The Tbolt has F-connector for the antenna input , and it was designed for cheap 75Ohm Antenna cable (use quad shielded , not cheap Harald cable)
In fact i bought a few N to F adaptors for the GPS antenna , and some F-connectors (diy for sattelite).
As F-connectors just "Screw on the cable" , it was super easy to pull & connect the antenna cable from antenna pos to the Tbolt.


N-F   for Antenna  (I'd get 2)
https://www.ebay.com/itm/192922995695

sma - F  - To get the Tbolt ant cable to fit on the Samsung (for test)
https://www.ebay.com/itm/193000694585


And now you'd want to have 2 x USB-->RS232-DB9 adapters , Must be RS232 level w. db9 , not the TTL adapters.
So you can run Lady Heather and gather from both units.


And in 2 weeks you want a gps antenna splitter , and in 4 a better freq counter , and in 3..6 month .. You have your first rubidium 
And you'll dream about an : Agilent 5071a   .... (Do you really need that new car ....)
You ( And your wallet) are destined for a Dark..dark wormhole , and there's no way out. 
Welcome to the time-nuts


/Bingo

[bingo600]

Here's an antenna w. pole  - If customs come after you from uk , i'd prob get one of the below (first is cheapest)

$$
https://www.ebay.com/itm/303869579117

$$$
https://www.ebay.com/itm/283468530873

[tkamiya]

As to size of the pole, there are several.  While those flange mount look all the same, they aren't.  My suggestion is to wait until you get the antenna then look for a pole.  Even if you have a picture, there is no guarantee what you'll actually get is the identical one to the picture.

[FriedMule]

About the black-frequency-hole, I know of a YouTuber who did go to a low-voltage-black-hole, I think there have to be a lot of black holes out there:-)
you are all of amazing help, I have tried to search for different types of coaxial cables and i think i may need at least one 50 Ohm N-male to N-male but then is the question is what do you use? I have to use about 10 meters or 30 feet.
Would you invest in a lightning "catcher"?

[tkamiya]

I use 15 meters of TimesMicrowave LMR-400 with N connectors on both ends.  It doesn't need to be anything this good actually.  RG58 or RG6 works equally well.  I have used even longer RG58 and it worked just fine.  I already had an LMR400 in place so I used it.  I don't have a lightening arrestor because the antenna is installed only 4 meters or so above ground and peak of the house is nearly 8 meters high. 

The requirement or choice will depend on your particular case.  Keep in mind, while connector on antennas are almost always N type, receiver side will vary.  Thunderbolt is F connector.  Some board type will use anything from SMA, SMB, to some sub micro stuff.  My suggestion will be to use N and use short piece of coax as converter to whatever receivers require.

[FriedMule]

Please take a look at this box https://i.ebayimg.com/images/g/oOEAAOSw07lbh4ki/s-l1600.jpg is that the F-female-connector?

I do only need about 30 feet (10 meters) but since my antenna is about the same as my air vent do I maybe need a lightening arrestor. no matter what is a lightening arrestor cheaper to have wasted money on than not have spent the money and find out I did need it:-)

[tkamiya]

One on far left for antenna IS the female F connector. 

If you are going to invest in lightening arrestor, don't forget you will also need a GOOD ground. 

[Uky]

I was adviced (and acknowledged) that when GPS-diciplining an HP 10811 OCXO that has been laying on the shelf for an extended period of time, it can take several months before the unit will reach an acceptable stability. Since 10811's are not hermetically sealed, they need to "dry up".

[MIS42N]

I was adviced (and acknowledged) that when GPS-diciplining an HP 10811 OCXO that has been laying on the shelf for an extended period of time, it can take several months before the unit will reach an acceptable stability. Since 10811's are not hermetically sealed, they need to "dry up".

It depends on what you mean by 'acceptable'. It takes a while for an OCXO to reach a predictable ageing rate. GPS disciplining will keep it on frequency in a shorter time. If you aim for 10e-12 then maybe months. If you are happy with 10e-9 then a day or two should be adequate.

[tkamiya]

By the way....

It's a good idea to have more than one standard running at any given time.  When they start to diverge, you will know something is wrong.  (they will never be exactly the same, by the way...)  I run two GPSDO and one Rubidium for this reason. 

If you only have one, even if they are wildly wrong, you'll never know because that is your standard.

If you are going to have two GPS, you'll need two antennas or one proper splitter designed for this purpose.  You can't just use a tee.  Coax also acts as power line for the antenna.  One by HP/Symmetricom is pretty much industry standard.  There are two channel kind (HP 58535A) or four channel (HP58356A).  I'd recommend 4 channel one as you'd want more very very soon.   

[MIS42N]

If you only have one, even if they are wildly wrong, you'll never know because that is your standard.

A deficiency of some methods.

If you have a GPSDO, you already have two standards, the GPS and the OCXO. The GPSDOs I am designing (built three, starting on the fourth) use the OCXO output as the clock for the microprocessor. The arrival time of the 1pps is timed to the nearest 25ns (quarter of a cycle). Just watching the arrival times gives a good indication if the OCXO is doing OK. The second thing is the microprocessor adjusts the "DAC" (I use PWM so it isn't really a DAC) at long intervals (minimum 20 minutes in one of them) so you can see the rate of divergence. The microprocessor does a calibration run when first set up, this determines the (control volts)/(frequency change) ratio. The OCXO I use typically 0.1V/Hz, so a change in control volts can be interpreted as a change in frequency. Since all frequency changes are aimed at bringing the GPS and OCXO into agreement, the frequency change is always larger than the frequency error. So (after a change) I can say with some confidence the GPSDO was within x Hz of accurate. If the changes are consistently below some value, one can also have confidence it is still that accurate (and if it really matters, wait for the next change for confirmation).

The question is if short term errors (periods of a few seconds) in arrival time are due to the OCXO wandering, or the GPS wandering. It is more likely to be the GPS, because in an intermediate period (a few minutes) they converge again. If the OCXO was at fault it wouldn't do that consistently.

But yes, the old adage - a man with a clock knows the time. A man with two clocks is never sure.

[tkamiya]

New adage for time nuts:

"a man with a clock knows the time. A man with two clocks is never sure.  A man with a dozen clocks has lots more fun."

I actually do what I suggested.  I didn't dream it up in vaccum.  Say I measure an oscillator and the result is not what I expected.  I then use another instrument to see if it's the same or pretty close.  If I get wildly off results, then I'll know I need to find out the reason.  Is one of the instruments off, two of them are off, or DUT is drifting wildly.  At one time when I was getting wildly different results every time I measure, issue was in a coax connector.  Clamp type BNC lost pressure and with a quick yank, it came right off....   Basic mistake I know.  But having multiple source gave me a pretty quick way to know where the problems might be.

It depends on resolution limit of instruments, too.  With a ten digit frequency counter, GPS reading looks dead stable against another GPSDO or Rb, or Cs.  But if you involve DMTD, it will never show a stable number.  It will drift back and forth in seemingly random fashion. 

It is fairly obvious that OP is new to the field.  Really, how is he supposed to know he got what he thinks he got?  When I got my very first one up and running, the first question I had was exactly that....  How do I know if it's accurate?  Well, that's how my time nuttery started.  Now I have about a dozen GPSDO, half a dozen Rb, one old Cs and lots and lots of OCXO.  I'm not even going to count number of frequency counters I have, how much money I spent.  Yes...  I have an OCD.   

[YetAnotherTechie]

I'm on a similar place and need some help.
The primary goal for me to own a gpsdo would be to adjust a rubidium oscilator.
In that case what is better, one of the mentioned samsung units or Leo Bodnars?
I don't care about 1pps, adjustable frequency or lady heather connection.
Should i care about the LH connection and would that make up for the samsung unit disadvantages?

[FriedLogic]

  Simply measuring the phase difference between the 1pps from a GPS receiver and your rubidium is a good way to measure the frequency difference.
  For long measurements where the phase of the rubidium could change by more than one cycle, adding a divider like this can be useful -  http://www.leapsecond.com/pic/picdiv.htm

[tkamiya]

If you are going to get a GPSDO, you might as well get one with PPS.  I like routing 10MHz out to ch1 of a scope and DUT to ch2, then trigger on ch2.  That way, trace ch1 will show me it's either behind or ahead of the DUT.

Here's something to think about...  GPSDO is only good for 10^-9 on second-to-second basis.  So each and every reading can jump around quite a bit.  How stable it will be will depend on how well GPS is locked to the signal, how stable OCXO/TCXO inside is, and how locking loop is set up.  LH is a great thing to have to see trend your GPS is seeing. 

When I tune Rubidium, I usually use HP53132A (ext ref. hooked up to GPSDO) set to 1 second gate time, and do a rough tuning.  Then do 10 second reading for an hour or so, then average the reading.  I can do this easily using PC and Excel.  The procedure will also depend on how accurate you want it to 10MHz.  Stability will really depend on Rb itself.  If you have a PRS-10, you can automate all this.  Exactly on how to do it is on users' manual.

[FransW]

My advise is to read the meaning of numerical dispersion:
"
In computational mathematics, numerical dispersion is a difficulty with computer simulations of continua (such as fluids) wherein the simulated medium exhibits a higher dispersivity than the true medium. This phenomenon can be particularly egregious when the system should not be dispersive at all, for example a fluid acquiring some spurious dispersion in a numerical model. It occurs whenever the dispersion relation for the finite difference approximation is nonlinear. For these reasons, it is often seen as a numerical error. Numerical dispersion is often identified, linked and compared with numerical diffusion, another artifact of similar origin.

In simulations, time and space are divided into discrete grids and the continuous differential equations of motion (such as the Navier–Stokes equation) are discretized into finite-difference equations; these discrete equations are in general un-identical to the original differential equations, so the simulated system behaves differently than the intended physical system. The amount and character of the difference depends on the system being simulated and the type of discretization that is used."

Source: https://en.wikipedia.org/wiki/Numerical_dispersion

Summary: it all depends on the required usage of the equipment. Without a valid cause it makes no sense.
Pointless conclusions are outside the common sense environment.

Regards, Frans

[MIS42N]

Here's something to think about...  GPSDO is only good for 10^-9 on second-to-second basis.  So each and every reading can jump around quite a bit.  How stable it will be will depend on how well GPS is locked to the signal, how stable OCXO/TCXO inside is, and how locking loop is set up.  LH is a great thing to have to see trend your GPS is seeing. 

I think a GPSDO should deliver way better than 10^-9. My own design is still in development, I am writing it up here: https://www.eevblog.com/forum/projects/budget-gpsdo-a-work-in-progress/. Here is data from nearly two days. The design is such that the control voltage is not altered unless the OCXO has wandered an estimated 1 cycle of 10MHz from synchronization with 1pps, at which time the alteration is designed to bring the OCXO into synchronization. This is done in case the 10MHz is used for long duration comparisons. Because the control voltage is not from a DAC, it is more precise. The variations below of control voltage are less than would happen with a 1 bit change in a 16 bit DAC (assuming the DAC is delivering 0-5V). Note that on average changes are done less than once an hour. They actually occur in pairs - set a voltage and wait to hit the one cycle limit (more than an hour). When it hits the limit, apply a change to bring it back to synchronisation. The return is usually 30-40 minutes due to limitations in the control algorithm and don't need to be that quick. In all cases the OCXO is running better than 0.1ppb - i.e. 1 part in 10E-10 from 10MHz.

Because the OCXO is running with a fixed control voltage most of the time, there is no second to second variation. The quality of output is dependent on the OCXO, and short term GPS variations have no effect.

The voltages assume a full scale of 5V. If the 5V were accurate, all but the last digit of the control voltage would be valid. The ppb (parts per billion or parts in 10E-9) changes are calculated from (V1-V2)x(OCXO V/Hz pulling) - which is known and for the particular OCXO about 0.1V/Hz.

Time 122416 UTC. Ctrl 2.0495204 0.045 ppb
Time 154903 UTC. Ctrl 2.0494727 -0.045 ppb
Time 162310 UTC. Ctrl 2.0495081 0.034 ppb
Time 190933 UTC. Ctrl 2.0494477 -0.057 ppb
Time 193924 UTC. Ctrl 2.0494944 0.044 ppb
Time 231243 UTC. Ctrl 2.0495665 0.069 ppb
Time 235106 UTC. Ctrl 2.0495210 -0.043 ppb
Time 022025 UTC. Ctrl 2.0495842 0.060 ppb
Time 030304 UTC. Ctrl 2.0495420 -0.040 ppb
Time 072735 UTC. Ctrl 2.0496064 0.061 ppb
Time 075726 UTC. Ctrl 2.0495530 -0.051 ppb
Time 095237 UTC. Ctrl 2.0494917 -0.058 ppb
Time 104348 UTC. Ctrl 2.0495207 0.028 ppb
Time 135547 UTC. Ctrl 2.0495781 0.055 ppb
Time 143410 UTC. Ctrl 2.0495326 -0.043 ppb
Time 160801 UTC. Ctrl 2.0494782 -0.052 ppb
Time 165456 UTC. Ctrl 2.0495115 0.032 ppb
Time 181143 UTC. Ctrl 2.0494492 -0.059 ppb
Time 190254 UTC. Ctrl 2.0494761 0.026 ppb
Time 232725 UTC. Ctrl 2.0495579 0.078 ppb
Time 235716 UTC. Ctrl 2.0494993 -0.056 ppb
Time 055955 UTC. Ctrl 2.0494480 -0.049 ppb
Time 063402 UTC. Ctrl 2.0494947 0.044 ppb
Time 084617 UTC. Ctrl 2.0494370 -0.055 ppb

[forrestc]

The primary goal for me to own a gpsdo would be to adjust a rubidium oscilator.

So, to adjust a rubidium oscillator, you can use pretty much any cheap GPS receiver with a reasonable 1PPS output.  No GPSDO required.

A common method is to divide the 10Mhz from the Rubidium down to 1PPS and then use a counter/timer to measure the relative time between the 1PPS from the GPS and the 1PPS divided down.   If the relative phase stays the same, then you can be sure the Rubidium is adjusted.  A common tool to do this is a 53131A.  or 53132A.

Alternatively, you can use a timestamping counter such as a TAPR TICC, and clock it directly from the rubidium and feed the 1PPS into one of the channels.   This will output a series of timestamps every time the 1PPS is triggered, and if the rubidium is calibrated correctly, the fractional part of the timestamps won't drift over a long period of time.   This is my preferred method.

Note that the reason why this works is that over a fairly long period, the +-100ns jitter out of a typical consumer-grade GPS receiver becomes so insignificant that it doesn't matter to your measurement.

Let me state this a bit differently:  Effectively what you're doing is creating a time period that you can use to measure the Rubidium that has 100ns jitter no matter the actual time period.   Over a second, you get 1 part per 10,000 error.   Over 10 seconds you get 1 part per 100,000.   Over 100 seconds you get 1 part per million, and over 100,000 seconds you get 1 part per billion.

So, over a time period of just over a day you can measure your rubidium to 1 part per billion.

Typically what I do with the TICC is I will adjust it until it doesn't look like it's drifting, and then check on it as time goes on.  If it's drifted more than expected I'll trim it up. 

There is a bit of finesse needed here in that if you don't pay attention to the data over a long time it may drift enough that you've gained or lost more than a second.   In addition, you'll need to handle the occasional missed 1PPS (which you can just ignore/filter out, as a general rule).

[usagi]

If anyone is looking for genuine Oscilloquartz Star4 boards, I have one with 8664-XS i'll be willing to sell.