GPSDO vs Rubidium reference

[mskobier]

All,
    Along with working on the Tektronix oscilloscopes, I have also been dabbling in time references. In particular a surplus 10mhz rubidium reference, and a china made GPS disciplined clock with 10Mhz sine wave output. Having a fair amount of measuring equipment, but nothing that approaches the theoretical accuracy of eith the rubidium reference or the GPS disciplined clock. So today, I hooked both up , let them warm up for a few hours, and input the output of both references to my 2445B, and took a look at the display. Both had a nice sine wave display, but the displays were shifting position in reference to each other. It takes right at 5 minutes ( as best I can measure) for the displays to shift one cycle. That works out to .00333 hz difference. The question I am asking myself, is which one is actually the most accurate? The stand alone surplus rubidium reference or the GPS disciplined reference. Normally I would believe the GPS one is the most accurate, but it is one of the relatively inexpensive ones from China that they are selling on -bay. One of the interesting things about the GPS reference, is that when I opened it up to see what was inside, the OCXO was made in the USA. I have searched the net and can find no reviews on these GPS based references. Do any of you guys (and gals) have one of these units (GPS disciplined Clock, BG7TL 2014-12-09), and just how accurate do you think it is, and how did you test it?

Mitch

[Wuerstchenhund]

I bought a BG7TBL GPSDO (awaiting delivery), but it seems that this device doesn't really output 10MHz:

http://www.ke5fx.com/gpscomp.htm

BTw, since you have one, how are you satisfied with it? How's the build quality?

[Electro Fan]

+1 for same and similar questions

ie, is either a GPS or rubidium reference inherently more accurate and/or stable?

What is the lowest cost generally available solution for getting an accurate and stable 10MHz signal?

- saw this - any comments on it?
http://www.ebay.com/itm/GPS-DISCiPLINED-CLOCK-GPSDO-10MHz-10M-OUTPUT-SQUARE-WAVE-RS232-OUTPUT-GPS-NMEA-/281649576001?pt=LH_DefaultDomain_0&hash=item41939f7841

(EDIT:  I was writing as Wuerstchenhund was posting - so maybe similar/redundant question/interests)

Any reason to think that a Trimble Thunderbolt would be better?

What type of antenna (how long and where does it need to be placed inside/outside) is required for a GPS?

How much difference does the power supply for a GPS make, if any?

Will cable length between the GPS and the test equipment have an impact on performance?

Can a 10MHz clock drive multiple test equipment devices at the same time (or will that degrade performance)?

Thanks! 

[G0HZU]

ie, is either a GPS or rubidium reference inherently more accurate and/or stable?

Both will be fallible in the real world so there won't be a definite answer to this but I don't think it matters unless you are working in something like a high end research lab or a calibration house.

[Mr Simpleton]

I would put my money on the GPS diciplined XO...
It is locked against GPS which relies on caesium clocks. This has excellent long time stability but sucks for short time, and phase noise. Now the XO will have best of both worlds. The rubidium keeps its frequency a lot better than the XO but over time and restarts it will not be as exact as the locked XO, especially if you only had it on for a few hours. Most likely it needs 24+ hours to reach its final frequency.

[Electro Fan]

ie, is either a GPS or rubidium reference inherently more accurate and/or stable?

Both will be fallible in the real world so there won't be a definite answer to this but I don't think it matters unless you are working in something like a high end research lab or a calibration house.

Hi G0HZU,

Thanks.  Ok - if both are "fallible but comparable and sufficient" which would be most cost-effective/practical/easiest to use/live with?  Do you have a preference?

Also, I tripped across one of these:
http://www.ebay.com/itm/EXTRON-ADA-3-80-DISTRIBUTION-AMPLIFIER-/371204748687

- would this enable the clock to drive multiple units, or can simple T-Connectors do the job (I'm guessing something might be impedance sensitive and it can't be that easy).

Thanks, EF

[Electro Fan]

I would put my money on the GPS diciplined XO...
It is locked against GPS which relies on caesium clocks. This has excellent long time stability but sucks for short time, and phase noise. Now the XO will have best of both worlds. The rubidium keeps its frequency a lot better than the XO but over time and restarts it will not be as exact as the locked XO, especially if you only had it on for a few hours. Most likely it needs 24+ hours to reach its final frequency.

Mr. S.,

Just to clarify, are you saying that the GPS disciplined XO will or won't have phase noise issues?  Thanks, EF

[G0HZU]

I'm the wrong person to tell you which to buy because I would say neither. I don't have either of them here.

However, A year or two ago I did borrow the company GPS disciplined OCXO made by Quartzlock and spent some time trying to see how reliable it was and I wan't impressed with it (in terms of how it would suit me)

http://quartzlock.com/product/frequency-reference/gps/E8-Y

It's easy to get the E8-Y to show 'locked' even if it is nowhere near the quoted specs way after it shows locked. This is due to the quality of reception from the various satellites it can detect. I ended up connecting to it via RS-232 via some Motorola diagnostic SW to help me put the antenna somewhere where it could receive lots of satellites cleanly. The SW gave me qty and signal quality for every satellite it could detect. Then it worked a lot better but I could easily make it misbehave by moving or twisting the unit. It also took about 30 minutes to lock.

The main reasons I don't have either is that they take a long time to stabilise from a cold start and they are usually too noisy to use as a precision reference clock for decent GHz capable test gear (although the E8-Y above is pretty good in this respect) The typical Rb standard is quite power hungry too and I really don't see the point of having one of these running all the time in my workroom at home. For me, the high stability OCXOs in my RF test gear are good enough and I rarely need to lock all my test gear to just one OCXO to sync them up better. 

[chicken]

I bought a BG7TBL GPSDO (awaiting delivery), but it seems that this device doesn't really output 10MHz:
http://www.ke5fx.com/gpscomp.htm

Not sure what you mean with "doesn't really output 10MHz". From the linked article:

It is effectively drift-free as would be expected from a GPS-disciplined oscillator, but its frequency is about 2E-11 too low -- or in terms of lost time, perhaps 75 ns per hour. The Thunderbolt is right on target vis-a-vis the HP 5071A, but the 10 MHz output from the BG7TBL unit is actually running at about 9.999 999 999 800 Hz. It's possible that there's a rounding bias or arithmetic precision issue in its software PLL code. Few users would be likely to notice this discrepancy, since it would amount to an error of only 2 Hz at 100 GHz!

Sounds close enough for me.

[mskobier]

All,
    The build quality looked as good as the pictures on -bay showed. I did not see any issues , and it appears to be a nice compact design. The GPS antenna that comes with it is quite long, approx. 20ft (6m), and the unit runs on 12VDC.

Thanks for the link to the KE5FX web page. That answered a lot of questions I had. I can live with a 2E-11 frequency error. Far better than I can even think of measuring. I do not plan on operating at 100ghz, but for me, a 2hz error at that frequency is acceptable. Just trying to calibrate a frequency counter that goes to 7 decimal places, and have a good reliable frequency reference.

For <$150.00 delivered, (paid over $100.00 for the surplus rubidium reference) I think it is reasonably priced.

Looking at the performance of the Trimble Thunderbolt it appears to be a bit more accurate, even if it is limited in the number of satellites it can see at any given time.

Mitch

[Wuerstchenhund]

Not sure what you mean with "doesn't really output 10MHz".

9.999 999 999 800 Hz != 10MHz .

Now, that wasn't difficult. was it? 

[Mr Simpleton]

What I am saying is that a XO do have better phase noise than a caesium oscillator, and depending on offset from carrier can be better than a rubidium too.

Long term, short term stability and phase noise are not that easy to compare

If you really need/want accurate frequency and time you really have to run the reference all the time... Rubidium or OCXO! Now a GPS diciplined oscillator may do a bit better but most high performance uints do take more than one hour to settle.

Most non-timenuts are just as good with a calibrated double oven crystal oscillator, or if you are in a hurry and have the ref swtiched off most of the time a TCXO is a lot faster.

I ended up with a used Trimble Thunderbolt which I plan to run against our Symmetricom GPS-Rb clocks (swtiched on 24/7) at the office, just for fun. And once verified this will help me to calibrate a few HP reference oscillators I have.

[Mr Simpleton]

A GPS locked crystal oscillator can have excellent performance and still be pretty simple...

http://www.jrmiller.demon.co.uk/projects/ministd/frqstd.htm

http://www.jrmiller.demon.co.uk/projects/ministd/stab.htm

Given its price, the performance is not bad at all! And consumes 3W, so can be left on 24/7 even if you care about the environment

[Electro Fan]

Just throwing this question out there for discussion in case anyone has any thoughts - might be others who have a similar question or interest....

Given the choice of either a square wave or sine wave output from a GPS 10MHz clock, what would be the various reasons for choosing one or the other?  (My primary reason for the clock would be to gain accuracy and consistency across various pieces of test equipment, but there might be other purposes/applications and considerations.)  Thx

[johnny_canuck]

A square wave contains (theoretically) infinite harmonics. This may or may not matter to the instrument that's using the 10 MHz signal as it's
timebase. All of my gear (frequency counter, RF sig generator, spectrum analyzer) is looking for a 10 MHz sine wave. I use a Trimble Thunderbolt GPSDO, a DownEast Microwaves 4-port distribution amp and a very clean linear power supply for the lot.

Ken

[Electro Fan]

All of my gear (frequency counter, RF sig generator, spectrum analyzer) is looking for a 10 MHz sine wave.

Ken

What spec would show (how would you know) that a particular piece of equipment is looking for a sine wave vs a square wave (or vice versa)?  Thx

[G0HZU]

In my experience wrt RF test gear the ext ref input is AC coupled and will accept either a low level sine wave and most will also accept a square wave. However, some of my lab grade RF test gear outputs a 2V pkpk square wave biased at 2.5V as a master 10MHz reference output.

But there is risk from spurious effects from the harmonics from a square wave for some types of test gear.
The advantage with a square wave is you can buffer/feed it directly into a divider if you need to divide down to 5MHz or 2MHz or 1MHz to suit some types of test gear. But I think a sine wave would be better if you wanted to daisy chain 'lots' of test gear to a common 10MHz reference.

If you have decent RF test gear then you also run the risk of degrading the spectral purity of the test gear if you feed an external reference that has poor spectral purity to it. Some of the cheap Rb references on ebay would be totally unsuitable for this and these are best used as a frequency reference for calibration rather than a master clock for a whole load of test gear.

Most counters have an AC coupled input for the external reference and will accept either a square wave or sine wave as an external reference in my experience.

[johnny_canuck]

Perhaps I should clarify. I'm using the GPSDO sinewave output as the external timebase reference input for my instruments. The manuals for
these instruments specify a certain RMS voltage for that input. Doesn't that suggest a sinewave?

[G0HZU]

Maybe... but we are in danger of opening a real can of worms here because not everyone will interpret things the same way.

Most test gear is tolerant of square wave or sinewave input for the external reference in my experience.

However, I have some RF test gear here that asks for a particular power level  range at the ext reference input. eg 0dBm to 10dBm at 10MHz. So this implies 50R and a sine wave to me.

But I have other high end RF test gear that asks for about 1.8V rms at the input. The same item of test gear outputs a 5V pkpk square wave as its internal 10MHz reference output.

In my experience most RF test gear is much more tolerant of sine/square wave/drive level changes than the datasheet describes. But I usually just tie sig gens together via the external reference input or feed external reference signals to frequency counters.

I rarely feed an external reference to anything else. If I need to sync up a spectrum analyser to a sig gen (or two) I would use the analyser OCXO reference as the master and this has always worked very well for me.

However, I can't imagine a (realistic) scenario where I would want to do this with a GPSDO or Rb standard to sync them all up instead of using the analyser OCXO as a master. I can't see where the added value comes from. I can forsee the potential downsides of attempting this though...

[TSL]

Perhaps I should clarify. I'm using the GPSDO sinewave output as the external timebase reference input for my instruments. The manuals for
these instruments specify a certain RMS voltage for that input. Doesn't that suggest a sinewave?

No... RMS can be used for nearly any wave shape.

For the math of it for square, pulse, triangular, go here...

http://masteringelectronicsdesign.com/how-to-derive-the-rms-value-of-pulse-and-square-waveforms/

regards

Tim

[TSL]

+1 for what G0HZU said.

Locking a range of instrumentation to a single master standard must be done with consideration of what each instrument requires as a reference.

While most require 10Mhz of some sort, some have 50 ohm impedance others might have 10k. This in it self can lead to mismatches and signal reflections so its a good idea to feed instrumentation with a fully isolating distribution amplifier so that no one instrument can affect the other.

Additionally that standard must meet or exceed the phase noise and harmonic content of your best equipment lest you muddy the results. i.e. My HP8566 manual has this to say...

"Analyzer performance will be degraded unless frequency reference phase noise and spurious signals are < - 140dBc single sideband (1 Hz) referred to 10 MHz at a 100 Hz to 10 kHz offset"

So like G0HZU , if an instrument needs to be in sync with the SA, the SA becomes the master oscillator.

Normally my sig gens and counters are all attached via a distribution amp from my always on GPSDO standard using double shielded equal length coaxes.

As to the question of why sine and not square, johnny_canuck's statement of infinite harmonics is an important consideration especially if its analog/RF stuff your doing.

The LPR manual also has this to say...

"Transmitting rf output signals over long distances is less of an EMI issue for the user when
the signal is a sine wave instead of a square wave because a sine wave lacks harmonics. In addition,
the power consumption of the sine wave driver into 50 ohms is lower than for a square wave driver
into 50 ohms, especially when providing short circuit protection."

regards

Tim

[johnny_canuck]

The Thunderbolt frequency accuracy is specified as 1.16 E-12 (one day average). No OCXO is going to come even close to that. After a couple of hours from a cold start, the T-bolt monitor software shows the 10 MHz output flickering around +/- .05 ppb. That's accurate! That's why they were used to sync up mobile phone sites in the days of CDMA. Of course, you need an outside aerial ($5 on e-Bay) and a run of cheap cable-tv coax to pick up the GPS satellites. And it never has to be sent out for calibration. Phase noise is spec'd as -135 dBc/Hz at 100 Hz offset.

[G0HZU]

The Thunderbolt frequency accuracy is specified as 1.16 E-12 (one day average). No OCXO is going to come even close to that. After a couple of hours from a cold start, the T-bolt monitor software shows the 10 MHz output flickering around +/- .05 ppb. That's accurate!

Sure, but if you sync all three bits of test gear to the analyser OCXO then they are all going to be sync'd to each other.
The added value of having sync is the goal here and not the absolute accuracy that GPS offers.

A typical requirement for me would be to sync up two sig gens so I can investigate phase related phenomena in a system. Typically this would be with harmonically related signals. I can adjust the relative phase of the two signals because they are in sync via a common master reference.

Having a GPDSO doesn't add any real world value here because sync is sync. The OCXO in a decent spectrum analyser will be good enough as a master for this type of work. There will be people out there in industry who 'need' to do dev work with RF test gear that demands the accuracy that GPSDO can offer but I'm suggesting that there aren't many other people outside this group. However, there will be plenty of people who 'want' this capability even though they don't 'need' it. Hence the abundant sales of stuff like this on ebay to keen hobbyists.

[Electro Fan]

Hi G0HZU,

Just checking to see if you could help me sort out a basic voltage (amplitude) measurement question?

On a HP 8648C the dial that sets amplitude can be configured (button pushes) to read mV or Vemf or dBm or dBuVemf or dBuV.  Is it reasonable to assume that some such setting (like mV or Vemf) would produce a voltage that would match the voltage reading on an oscilloscope?  For some reason (might have to do with AC vs DC coupling or impedance, but I've tried those) I can't get a voltage output from the sig gen to match what I'm seeing on the scope (not peak, or peak to peak, etc.).

It's probably just operator error but I'm slightly wondering about the sig gen.

Thanks!  EF

[G0HZU]

Hi EF
You should maybe start a thread about your sig gen but if you set it to +10dBm (10mW) at maybe 10MHz and feed this via some RG58 coax to a scope and terminate the scope in a 50R load then you should see 2Vpkpk on the scope.

You should see the same level of 2Vpkpk on the scope if you set the sig gen to 1.414V emf because the emf is the open circuit rms voltage but you are terminating the output in 50R so the (terminated) rms voltage will be 0.707V (which corresponds to 2V pkpk)

The Vemf reading confuses a lot of people because it refers to the rms voltage BEFORE you terminate the output in 50R. My Marconi sig gen here has the option to select Vemf or Vpd and Vpd is the  rms voltage when terminated in 50R. I've set mine to Vpd mode because I don't like to use Vemf. So to get 2Vpkpk with this sig gen I would select 0.707Vpd. But you would select 1.414Vemf for the same result on your HP8648C when fed into a 50R load. This is the same as selecting a power of +10dBm using dBm instead of Vemf etc.

Also, try entering 707mV   (and not 707mVemf) and you may still see 2Vpkpk across the 50R load at the scope because the mV mode is likely to be the same as my mVpd mode. Both of my Agilent sig gens offer mV or mVemf options and maybe yours does too?

Hope this makes sense.