Suitable GPS for GPSDO OCXO

[cdev]

The inherent bandwidth of a ceramic patch antenna is significantly more narrow than its equivalent in air. Also the ceramic antenna is substantially smaller. 

Coverage of other constellations varies a lot from antenna to antenna. Ceramic antennas are designed to be narrow banded, they have a specific asymmetrical shape which is how they implement circular polarization. The machines that make them have to trim them individually while measuring their RF characteristics.

That makes them much more frequency specific.  Also size does matter.  The bigger a patch antenna is the more likely it is to "work" on, say Glonass, but is it a good antenna for surveying or precise measurement of position? That will be dependent on a number of things, specifically phase center variation and rejection of multipath reflections which tend to be reversed polarization. Ideally the electrical center of the antenna should be the same irregardless of frequency.

Good broadband antennas often are definable simply by their geometry, not their size.

[cdev]

The best circular polarization antenna designs leverage the circular polarization to extract an extra 3 db of gain out of the signals, and also give you an additional 3 db rejection of the opposite sense signal (reflections) which matters the most at low angles.. (where the patch antennas tend to lose their circular polarization)

 Rejection of multipath reflections can really make a huge difference with a GPS signal.

In order to improve the crappy pattern at low angles, you can set up what amounts to a trap for signals that they go into and never come out of. But this trap which looks a bit like a hollowed out wedding cake with stacked rings, can only work on one narrow band of frequencies.

You were talking about a quadrifilar helix. Most QFHs we see are fed at the top but they can also be fed at the bottom and use a 90 degree hybrid to create the needed phase shift. The best GPS antennas have four or more arms to make the pattern completely symmetrical. They look a bit like a pinwheel. You can also implement interesting antennas that wrap the arms around a sphere. Those antennas should be able to reject LHCP even at low angles. But they are big.

[cdev]

RTKlib is a good way to learn about GPS antennas. Its a very rich toolkit that utilizes a substantial number of different technologies, really a marvel of engineering and science.

[cdev]

Look at Vaisala radiosondes

[metrologist]

I'll have to followup with the posts later. This is what I was doing tonight...

[CJay]

Look at Vaisala radiosondes

Those Vaisala turnstile antennae also have a nice little LNA on the board, I highly recommend powering one up before tearing it down so measurements can be made and the LNA can be used too, they're realy rather useful.

[suicidaleggroll]

Let's see based on at least one module series'  datasheets:
BEIDOU: 1561.098 +/- 3
GPS L1: 1575.42 +/- 2
GLONASS: 1602+/- 4

So for all three that'd range from 1558...1606 with a center around 1582, 47.9 MHz desired "flat" BW.
For only GPS L1 + GLONASS that'd be 1573...1606 with a center around 1589.71, 32.58MHz desired "flat" BW.

I guess maybe all of those frequencies are so close that propagation differences might not be significantly evened out by referring to one vs. the others for better timing stability unless geometry or something else was a significant reason for a benefit.

No those are too close, but even if they weren't it wouldn't matter.  To make a real difference the two frequencies need to be coming from the same satellite, not two different satellites.

[cdev]

The transmitting/receiving antennas on GPS and remote sensing satellites that use GPS are usually QFH or QFH-like  multi-arm antennas.

Example in the image below


It should be mentioned that for precise measurement, antenna's performance has to be characterized by calibration.

Professional antennas go to great lengths to minimize multipath. Once you do that then the problem becomes phase center variations.  Basically the speed of light in different directions is modified by the structure of the antenna such that it performs differently depending on the angle a signal is being received from.

Professional antennas for geodesy are marked so you can position the antenna such that the calibration information - which is not that complicated, is accurate. 

People who want high levels of accuracy end up spending a lot more money on their antenna than you would for a non-calibrated GNSS antenna.

If you are making an antenna commercially in the US you may be able to get help with calibration.

Multi-constellation adds a new element into the mix and substantially increases the bandwidth required of an antenna for example, optimal GLONASS performance requires an antenna maintain its resistance to multipath reflections (which have inverted polarization) at low angles at 1600 MHz.

New GNSS systems that are in the pipeline will increase that bandwidth requirement still more.

[metrologist]

This is what I got. I think the decoupling capacitor cracked. It resonates at 1585MHz.

[cdev]

The two loops in a QFH antenna that is built (non-optimally) without a 90 degree hybrid resonate at two frequencies, one is above the range its supposed to work at, the other below.

I don't own any kind of network analyzer so I have no idea how this would resolve on a network analyzer. I suspect you would see a combination of the two, but the signals in the two individual loops could be analyzed separately to determine their sizes, one a bit too big, one a bit too small. 

[cdev]

Do the cheap Trimble units have any signals, such as 10 KHz, 1 MHz, etc. that might be useful for a GPSDO (besides the timing pulse)?

[metrologist]

I do not think there is any other timing signals from the cheap Trimble units.

Here is the GPS spectrum with my antenna sitting inside the middle of a typical office building. At least it hears something. I'm a little hopeful it will work, but I may build another. I cut all of the elements a little short, about 0.5mm. The disc sander I was using to trim the ends nice and square got a little hungry 

[Gyro]

Do the cheap Trimble units have any signals, such as 10 KHz, 1 MHz, etc. that might be useful for a GPSDO (besides the timing pulse)?

According to the product flyer it's an 8 channel receiver with 1pps (+/-95ns) only. Pretty jittery by modern standards, it's not intended as a timing receiver.

http://www.dpie.com/datasheets/gps/11547e%20Lassen%20LP%200205.pdf


P.S. The manual, 1999 vintage:

http://www.dpie.com/manuals/gps/lassenlpman.pdf

[metrologist]

Is your helix a Sarantel antenna?  They do a size reduced GPS QFH thats pretty cool.

It might be. It does look like S1203A from google images. The antenna is unmarked, however. It measures as an open circuit though, so maybe it is no good or is already decoupled. I thought I once measured it as a short, but now it is definitely DC open.

[metrologist]

Do we think this receiver will be relatively deaf? I know my Jupiter struggles with a puck antenna in the window. The receivers I can use to compare are old Garmin portable units, an 12XL (and why are those still so expensive on ebay?) and a GPS60. These work fine outdoors with clear sky, but not indoors very well. The puck antenna t the window works well with them.

So, I am not to expect this receiver to work well enough with my OCXO's? I'd be happy if I could keep it limited to it's hour or so drift and keep it in cycle. Isn't that GPSDO's do anyway and are ultimately limited by the oscillator for their short term stability? I'm not quite clear why such a highly stable pps clock reference is needed? OK, 250ns is more than two clock cycles, but wouldn't we be able to average that out over minutes?

[MosherIV]

So, I am not to expect this receiver to work well enough with my OCXO's? I'd be happy if I could keep it limited to it's hour or so drift and keep it in cycle. Isn't that GPSDO's do anyway and are ultimately limited by the oscillator for their short term stability?

I did not think OCXO drifted that badly over such a short term. I would expect OCXO to take hours before they even drift by 1 cycle. Xtals are suppose to have better stability in the short term than the PPS of GPS, I suppose that is if they are kept in stable temp.
Bear in mind that they need to warm up to full operating temp first, usually takes an hour. They need to be housed in some case which further insulates them. They need to be left on.
Have you run any experiments with your OCXO to verify what the drift is?

I'm not quite clear why such a highly stable pps clock reference is needed? OK, 250ns is more than two clock cycles, but wouldn't we be able to average that out over minutes?

With such a bad jitter in the PPS, it may take 10s of hours possibly even days to average out.
You would really be better off investing in a better GPS. Even a position one with lower jitter would be better. From the sound of it, a newer more sensitve one may even help with your antena problems.

[cdev]

Its not just simply a matter of getting four satellites so you get a 3D fix, the quality of the satellite signals will determine the stability of the timing output.

[cdev]

You can buy the newer Ublox modules on ebay for less than $10 now. You can use them, or use the Navspark Mini (also very inexpensive)
or any decent GPS with a reasonably precise 1PPS to discipline the clock of a raspberry pi using NTP and use the RPI GPIOs to generate an arbitrary frequency. Hams do it all the time using "WSPrry Pi"   https://github.com/JamesP6000/WsprryPi  . Note that I am just pointing to that as an example of GPIO PI use.. Unless you are a ham, your emissions need to be confined to the insides of coaxial cable to be legal.

[metrologist]

So, I am not to expect this receiver to work well enough with my OCXO's? I'd be happy if I could keep it limited to it's hour or so drift and keep it in cycle. Isn't that GPSDO's do anyway and are ultimately limited by the oscillator for their short term stability?

I did not think OCXO drifted that badly over such a short term. I would expect OCXO to take hours before they even drift by 1 cycle. Xtals are suppose to have better stability in the short term than the PPS of GPS, I suppose that is if they are kept in stable temp.
Bear in mind that they need to warm up to full operating temp first, usually takes an hour. They need to be housed in some case which further insulates them. They need to be left on.
Have you run any experiments with your OCXO to verify what the drift is?

I'm not quite clear why such a highly stable pps clock reference is needed? OK, 250ns is more than two clock cycles, but wouldn't we be able to average that out over minutes?

With such a bad jitter in the PPS, it may take 10s of hours possibly even days to average out.
You would really be better off investing in a better GPS. Even a position one with lower jitter would be better. From the sound of it, a newer more sensitve one may even help with your antena problems.

I have no way to verify my oscillators, but are spec'd < 5mHz/day. This GPS timing accuracy is stated as +/- 100ns in the manual and accuracy is spec'd +/-95ns. Their 100ns figure is the 1 sigma value of 10 sets of 100 samples over 20 minutes. They say "not allowing for SA" but I'm not sure if that is with SA on or off.

[cdev]

That would be fine for a stratum 1 NTP server. When a GPS produces other signals that are locked to the GPS, they typically arent stable enough to be used as a frequency standard unless their output is averaged over a long period of time in various ways to remove the jitter.

[metrologist]

(not allowing for SA).

That means with SA turned off, the way it is now?

My QFH antenna seems to work well on my Jupiter receiver sitting on my desk in the house. I'll compare that with these Trimble receivers when they arrive.

[metrologist]

Is it passive with no LNA in your tests or did you find a LNA to plumb in to it?

Thanks for the insights about the Trimble. I was wondering how as well.

No LNA on the antenna. About 1 foot of RG174 to an MCX connector. I should remeasure it because the decoupling capacitor was definitely cracked.
I could not get a lock with the commercial QFH antenna but should give it another try as I might have had a coax issue.

[edpalmer42]

So what kind of performance might one expect if one took a reasonably decent but still inexpensive VCTCXO, put it in a circuit with minimized (maybe simple oven) thermal variation, reasonably lowered vibration / PCB mounting stress, clean voltage supply, good well designed oscillator topology given what the crystal is suited for, and disciplined that?
EDIT: I realized I was mixing topologies.  If it is VCTCXO it already has an oscillator circuit, so one relies on that or if it is otherwise better select a good inexpensive AT cut or whatever crystal and make a simple VCOCXO or VCTXCO with environmental stabilization based on the crystal.   Either way. 

Clearly you'd get zero average intermediate /  long term error due to the locking.  I guess the question would be from T=0s to T=1k seconds how the locked condition error would be, I assume the local XO would still be able to provide significant improvement over the jittery GPS timing signal for t <= 1000s.  Somewhere in there there would have to be the compromise cross over point between the PLL / loop ADDED jitter due to disciplining the XO by the LPF filtered GPS time signal and the XO's intrinsic drifty / slightly noisy free running output quality.

If you didn'r really need unlocked "hold over" accuracy / stability but just wanted cleaner jitter over T=0 to T=10k seconds and basically zero frequency error over T=X*100s to T=10k+ seconds, is it a reasonable idea?

Here's the manual for a commercial unit called a GPSTCXO.  On page 24 it shows the Allan Deviation.  Better than a bare GPS receiver, but maybe not as good as an OCXO, although it would be a lot cheaper to buy if you were comparing prices for new units from the manufacturers.  I actually have one of these.  It's performance is pretty much identical to the graph.

http://www.jackson-labs.com/assets/uploads/main/gpstcxo_manual.pdf

Note that the part of the Allan Deviation graph that slopes down and to the right is the 'signature' of the performance limit of GPS.  I call it the 'GPS Line'.  Every GPS receiver and GPSDO will more or less follow that line.  On a bare GPS receiver, the line continues in a straight line up and to the left.  The OCXO or TCXO comes in from the left and intersects the line at some point, producing a graph similar to the one shown in the manual.

Ed

[suicidaleggroll]

But descending a level from those high quality VCOCXO devices there are VCTCXO devices which are a couple of orders of magnitude sometimes less expensive and are still considered "good enough" for many telecommunications clocking, GPS RX, and so on applications.

So what kind of performance might one expect if one took a reasonably decent but still inexpensive VCTCXO, put it in a circuit with minimized (maybe simple oven) thermal variation, reasonably lowered vibration / PCB mounting stress, clean voltage supply, good well designed oscillator topology given what the crystal is suited for, and disciplined that?
EDIT: I realized I was mixing topologies.  If it is VCTCXO it already has an oscillator circuit, so one relies on that or if it is otherwise better select a good inexpensive AT cut or whatever crystal and make a simple VCOCXO or VCTXCO with environmental stabilization based on the crystal.   Either way. 

Clearly you'd get zero average intermediate /  long term error due to the locking.  I guess the question would be from T=0s to T=1k seconds how the locked condition error would be, I assume the local XO would still be able to provide significant improvement over the jittery GPS timing signal for t <= 1000s.  Somewhere in there there would have to be the compromise cross over point between the PLL / loop ADDED jitter due to disciplining the XO by the LPF filtered GPS time signal and the XO's intrinsic drifty / slightly noisy free running output quality.

If you didn'r really need unlocked "hold over" accuracy / stability but just wanted cleaner jitter over T=0 to T=10k seconds and basically zero frequency error over T=X*100s to T=10k+ seconds, is it a reasonable idea?

See the attached figure.  This is from a GPS unit that uses a 10 MHz Vectron TCXO to run the clock.  The TCXO is NOT disciplined, it's free-running, what you see in the graph is the frequency error of the clock in Hz as a function of time for a day.  You can ignore the large spikes every few hours, those are caused by the GPS doing a clock fixup which gives erroneous results for the clock velocity when it happens.  This TCXO isn't temperature controlled, but it is in an enclosure with a relatively constant temperature, inside a room with a relatively constant temperature.

[metrologist]

I wish I had taken notes when I was calibrating my HP5314A counter with the high stability TCXO oscillator. I sort of remember it fluctuating through 0.8Hz each day for the week I had it set up at work monitoring the GPS 10MHz reference. Its oscillator spec is 1E7/month and 1E6 0C to 50C, but the room was temperature controlled. I could be remembering monitoring something else with that counter though. Can't wait for my GPSDO to come in.