BG7TBL FA1 and FA2 frequency analyzers

[Jarl]

I get similar results as bnz when measuring Noise Floor with an external oscillator versus measuring Noise Floor with the internal oscillator.

I guess that it might be due to the short rise time of the 4 V square wave output from the CTS OCXO within the FA-2 - a rise time which could be much shorter than the rise time and noise content of the signal from the external oscillator input after it's internal processing within the FA-2.

Since the lack of long term stability of the internal oscillator seems to prevent its use for ADEV measurements beyond 100 s or so you might consider adding a phaselock circuit to the FA-2, disciplining the internal OCXO to an external, more stable reference in order to exploit the advantages of the internal OCXO for longer test periods.

This could be made with just 3 IC's and find it's place within the FA-2 case. It could improve the ADEV resolution as outlined in the test result shown by bnz and in addition compensate for the internal oscillators strong sensitivity to changes in the gravitational field. But it will unfortunately not be able to correct for the small, but notable counting offset when using the FA-2 to measure frequency.

[jpb]

The good results for the measurement of the internal oscillator against itself are interesting but, unfortunately they don't really measure a noise floor in the sense of how low a measurement of another oscillator can go as they are measuring a good correlation of a relatively noisy source. The measurements by TexasPyro show much higher ADEV when the internal source is measured relative to an external reference.
The noise floor measured with an external reference are more directly applicable as that is the setup you'd want to use with a good reference oscillator. The results are still pretty good given the price of the instrument.

According to this data sheet (CTS 196 series) :
https://www.ctscorp.com/wp-content/uploads/2015/12/OCXO196.pdf

The ADEV of the internal reference at 1s is ~ 10^-11 so though the internal correlated measurement seems to go down to 10^-12 the measurement of an uncorrelated source will be at least 10^-11 even if the source being measured is only 10^-12.

[Jarl]

Thanks for the heads up. jpb.

I have the phase lock of the internal OCXO in operation just now and the improvement in Noise Floor ADEV is as significant as could be expected, but only when using the internal OCXO to feed the CH1 input.

If I feed the CH1 input from the disciplining house standard - which has very low phase noise - the improvement in Noise Floor is much less impressive.

An interesting lesson learned....

[Jarl]

As a follow-up to jpb's comment and in order to illustrate the difference between ADEV measurements in loopback-mode and ADEV measurements on an uncorrelated source I attach a set of measurements made on a spare LPRO 101 rubidium oscillator which I have had in continuous operation for several years in my workshop. This oscillator has its own power supply and is in every way completely separated from my 10 MHz house standard.

The blue curve show the ADEV measured with the FA-2 using the 10 MHz house standard (+12 dBm, sinusoidal) connected to the external ref input.

The red curve show the same measurement but with the FA-2 OCXO being phase-locked to the house standard.

As can be seen, no visible advantage is obtained by phase-locking the OCXO to the house standard since virtually the same result is obtained by using the external ref. input directly - at least with the sort of real life signal obtained from a LPRO 101 rubidium.

Just for comparison the green curve show the noise floor of the FA-2 when looping back the undisciplined OCXO output to the CH1 input - that is just the unrealistic, correlated measurement described by jpb.

It would be of interest, if someone with access to two independent 10 MHz low phase noise sources could repeat the measurement as shown by the blue curve in order to establish the "real life limit" of ADEV measurements with the FA-2. 

[hgl]

This is a test with 4 free running OCXOs of the same type but not at the same time

[Jarl]

Many thanks to hgl for sharing your test results. They seem to be very close to the previous results of noise tests using the external reference input and the loopback-method discussed earlier.

Please tell us how your FA-2 during the present tests was controlled via it's external reference input - by what sort of reference oscillator ?. And please also tell us the type of, or technology behind the four test oscillators.

[hgl]

I have 50 Ocxo from a German company KVG produced around 2005. Parts are not labeled but the data sheet in the Attachment fits. Test oscillator was connected to external input and ref out to CH1.
I also have 9 Swiss made Rubidium Temex MCFRS01 with 4x10e-11 / month drift and can repeat the measurement but that takes time because they need a heat sink and a special connector and I can't just replace them.

[jpb]

I have 50 Ocxo from a German company KVG produced around 2005. Parts are not labeled but the data sheet in the Attachment fits. Test oscillator was connected to external input and ref out to CH1.
I also have 9 Swiss made Rubidium Temex MCFRS01 with 4x10e-11 / month drift and can repeat the measurement but that takes time because they need a heat sink and a special connector and I can't just replace them.

It would be interesting to know what the results are with the signal from the OCXO split and passed to the external input and the measurement (DUT) input.

The loop back tests give very low numbers but I'm still not sure if this translates to an equally low noise floor if, for example, two HP10811s with ADEV of less than 10^-12 were measured against each other.

[hgl]

I also tried a BNC T-coupler and a 75 Ohm distributor with my Rb and remember no relevant differences. Also no relevant differences between sine and square at the external ref input.

[edigi]

While I don't know how exactly the FA2 works or what is the exact reason for the worse ADEV value when external reference is used (I can also confirm that, checked with T-adapter), but my guess for the root cause is that the external reference is connected to a different pin of the Max II CPLD. Either the routing to that pin is noisier or the routing within the CPLD is noisier.

When measuring own reference most of the phase noise effect of the reference is taken out from the measurement. This is because when the interpolating part is measured (the time difference between the signal edge and the next reference edge) if the signal = reference (maybe with some shift) even if there is phase error, the phases for signal (= reference) and reference are moving together.

Thus when measuring own reference most of the phase noise of the reference is eliminated. That is what we're measuring is the noise of the measurement setup itself.
So while I maybe wrong with this whole theory, this means that the measurement setup for the external reference is worse.

This is unlucky, since in case of the external reference input a better precision reference can be used. However in the light of this result it seems that with external reference only the accuracy can be improved on the price of loosing some stability.
The good thing is that FA2 is still pretty good even when using external reference.

One way to cross check this theory is temporally removing the OCXO from its place (or maybe its signal output only to provide for the rest of the circuits the same load) and directly connecting an external reference there.

Some comments to older posts.

The 190823 board has a 6 pin IC and the other board did not have it.

True, but even the IC is in place, some DC blocking capacitors are not (missing also from my PCB) so net result is the same: That block probably planned to do some input signal shaping is not doing anything even in case of my variant of the board.

The delays probably calibrated and stored in the EEPROM. The question is: are the devices individually calibrated during production or was a single characteristic recorded once for all?

Probably not. In the FPGA solutions typically much better precision is aimed with a single measurement. Thus in FPGA case it matters a lot that the thermometer code is linear. In case of FA2, in statistical mode probably multiple measurements are done during the measurement period and the expectation is that even the thermometer code is not linear it averages out.

[Jarl]

Just to clarify: When using the FA-2 without external reference the signal from the internal OCXO is available on the reference output BNC, +6dBm, and sinosoidal - in spite of the oscillator itself delivering a square wave signal.

When you apply an external reference to the FA-2 - sinosoidal or square wave and with an amplitude within the limits specified in the user manual -  the signal from the external reference is available on the reference output as a sinusoidal + 7 dBm signal - instead of the signal from the internal OCXO.

This means that you unfortunately are not able to make an uncorrelated measurement of the instruments noise limit by using an external standard connected to the external ref. input and at the same time looping the signal from the external reference output back into the CH1 input because the two signals are essentially from the same source, just with som internal buffering and filtering added.

If one want to make an uncorrelated measurement of the noise level which can be obtained with the FA-2 in a "real life" measurement there is a need for two stable and independent low phase noise oscillators, one applied to the ref. input and the other applied to the CH1 input.

[edigi]

If one want to make an uncorrelated measurement of the noise level which can be obtained with the FA-2 in a "real life" measurement there is a need for two stable and independent low phase noise oscillators, one applied to the ref. input and the other applied to the CH1 input.

It's very unlikely that there is anyone on this forum who could do that measurement. Let me elaborate a bit.

Like I've explained above when correlated measurement is done (either looping back to CH1 the REF output or splitting external reference to EXT REF in and CH1) the reference phase noise impact is mostly eliminated.
That's why you see so outstanding ADEV figures. Mostly equipment measurement noise is measured (OK correlated), reference noise is excluded.
However if you'd like to measure uncorrelated noise level, the phase noise of the 2 references better to be at least one magnitude of order below of the phase noise of a typical OCXO since in that measurement you can measure only a combined noise level of FA2 and the 2 low phase noise oscillators.
That means something like BVA OCXO (maybe there is also something better/newer available but that's so much beyond my reach that I did not search that at all).
So while maybe there is somebody on this forum with a single copy of BVA OCXO (and willing to play with FA2) for that measurement you'd need 2 of them. Good luck finding someone with 2 copies of that OCXO and willing to play with FA2...

While I did not think too much of this but maybe using DDS that can output phase shifted and frequency shifted signal could be better. Not totally uncorrelated but checking parameter by parameter the impact of loosening (relaxing) the correlation a bit.
At least from equipment perspective it's more realistic, although more complex as well. Not something that I'd be willing to waste time on. After all we're talking about a 100USD frequency meter designed for and used by hobbyists.

[Jarl]

I am a bit more optimistic with respect to the resources available to members of this forum, edigi. After all we have already seen a report on a test of the FA-2 done with a Caesium-standard.

I look at the problem from a practical angle. Although the FA-2 is a cheap and not very well documented instrument, the correlated measurements show an interesting, theoretical limit of the instruments stability and resolution when doing ADEV tests with external references in loopback mode.

Now to the good question: What will be the noise limit of the FA-2 when used for some practical testing of ADEV ?

Using my 10 MHz house standard as external reference I have established the ADEV of the uncorrelated output from a spare LPRO 101 - which is well known for not-so-stellar performance with respect to phase noise. But it is an indication of the ballpark-figure to expect.  What I now would like to know is whether it would be worthwhile for me to invest in better instrumentation by comparing similar test results from other list members who has access to better references and test objects.

I respect your point of view - that it is a waste of time, but please allow me to find the problem interesting and worth further investigation.

[edigi]

An atomic frequency reference/standard is probably easier and cheaper to obtain than a really low phase noise oscillator despite that some organizations that can afford new technology (like cryogenic sapphire oscillator or alike) replace old equipments.

I at least don't have anything that could even come close to the phase noise requirements of the measurement that you've described (could try second FA2, SA reference output or some double oven OCXOs though).

[FriedLogic]

There are plenty of surplus oscillators with an ADEV well below 1E-12 at around 1s. Some of the HP10811s, Morion MV89s and many others perform that well, although most do not. Old oscillators often do not perform a well as they once did anyway, and the disposal and reclamation process that they go through can damage many more.

The problem of course is that there's no way to know for sure that they work - let alone work that well - before getting them. Anyone with a bunch of good oscillators like that which appear to work fine may well find that they have a couple of very good ones. The only way to know would be to try them.

There are some very expensive oscillators available as surplus, but unless the supplier can give recent test data for that particular oscillator, or accepts returns, it becomes a very expensive lottery.

[hgl]

https://www.ebay.com/itm/OSA-OSCILLOQUARTZ-BVA-8600-quartz-crystal-oscillator-FREQUENCY-STANDARD/174124227960?hash=item288a9cdd78:g:JnEAAOSwGzBdy6gM

[jpb]

https://www.ebay.com/itm/OSA-OSCILLOQUARTZ-BVA-8600-quartz-crystal-oscillator-FREQUENCY-STANDARD/174124227960?hash=item288a9cdd78:g:JnEAAOSwGzBdy6gM

One sold fairly recently (a few months ago) for around a third of that price (£1500 I think) - I was tempted but it was just too pricey and you have no guarantee that its performance is still at the 10^-13 level, especially as you'd need two of them to even get a measurement to find out!

[hgl]

if you have 2 and find a difference you need a third to find out which is bad 

[jpb]

Looking at the noise floor plots (for loop back) in this thread I notice something a bit different from my counter and from plots for other ADEV systems (and this is true of measurements from 3 different people), the slope of the curves go as -1/2 instead of -1.

Attached is a plot of a couple of my setups for my counter the FCA3100 which has a one shot resolution of 50 psecs.

The top curve (straight line) is for the counter just measured with a split buffer output - it starts at a bit over 10^-10 (2 x 50psecs = 10^-10 at 1 sec) and drops as a straight line with a slope of -1.
The lower curves are for two different measurements where I've used a mixer and an AWG to try and get heterodyne amplification. The noise floor is lowered but below around 10 seconds the slope is much less (nearer -1/2) and I find that real measurements in this area are a bit odd, the DUT and the Ref being swapped over give very different results.

The measurements for the FA2 have this same slope of -1/2.

According to W.J.Riley's Handbook of Frequency Stability Analysis, a slope of -1/2 corresponds to White FM while -1 corresponds to White PM. That is with a slope of -1 it indicates a fairly constant phase noise (measurement noise of phase) whilst -1/2 corresponds to noisy frequency. I guess this makes sense in that the FA2 measures frequency with a gap between (dead band) while my counter measures back-to-back phase even when measuring frequency.

[FransW]

if you have 2 and find a difference you need a third to find out which is bad 

Whay happens when you use 5 OCXO's (different make or same make), 3 CS's, 3 RB's and 2 hydrogen masers?

Selecting the most accurate one seems nearly impossible.
Just average and further rely on NIST/BIPM?

Or just extrapolate to the Planck time?

[jpb]

I think for standard time they use a collection of Caesium standards distributed about the country - but then you have the issue of transmitting time between them.

An additional problem is that if you have a collection of oscillators in the same lab then they are all subject to the same temperature/humidity and if they react in similar ways there must be a degree of correlation (in simplistic terms, if you had a bunch of OCXOs or GPSDOs and they all have similar F/T slopes) how do you get independent references? All your clocks might be in agreement but they might all be wrong together! 

[FransW]

The question was rhetoric.
The point is that absolute measurements do not exist.
You might want to have a look at:   https://de.wikipedia.org/wiki/Frequenzkamm,
a "frequency comb generator" from the Max Planck Institute for Quantum Optics.
The inventor ended up with a Nobel prize in Physics (2005).

[EV]

Here are pictures when same 10 MHz signal is connected to EXT-REF and CH1. With 10s gate time Avg is about 0.02 mHz high and with 1s gate time 0.2 mHz high.

[EV]

Here is the same test with Agilent 53131A counter.
- Number of samples 100
- Gate time 1 s
- Avg is 0.02 mHz low

Test with FA-2
- Number of samples 137
- Gate time 1s
- Avg is 0.23 mHz high

FA-2 needs longer gate time (10 s) to get same accuracy!

Has somebody else tested this?

[jpb]

Here is the same test with Agilent 53131A counter.
- Number of samples 100
- Gate time 1 s
- Avg is 0.02 mHz low

Test with FA-2
- Number of samples 137
- Gate time 1s
- Avg is 0.23 mHz high

FA-2 needs longer gate time (10 s) to get same accuracy!

Has somebody else tested this?

This is an interesting result.
The Agilent 53131A counter is 10 digits/sec according to its spec which I guess is around 10^7/10^10 = mHz but that is for a one off - you measured an average over 100 samples so it should improve to 0.1mHz or better (I think the scaling goes as sqrt(N) where N is the number of samples).
The FA-2's spec is 11 digits/sec for 10MHz - 99.999999MHz or 10 digits/sec for 1MHz-9.999999MHz so should be good for around 0.1mHz so for one off and down to 0.01mHz over 100+ samples.
But the FA-2's spec gives the "resolution" as 0.1mHz at 10MHz for a 1s gate so if this is the least significant digit then it is coming out a couple of counts high.

So on the face of it the Agilent is much better than spec (which is not surprising as I guess that they are conservative in their specs), while the FA-2 is around 2 to 3 times or more worse than spec or perhaps the spec may be accurate (the resolution and speed are as stated) but there isn't an accuracy spec.

What puzzles me is how this translates to ADEV measurements. The Agilent 53131A has a one-shot resolution of 500psecs, my counter - the FCA3100 has a one-shot resolution of 50psecs and this translates pretty closely to an ADEV noise floor of 5 x 10-11 at 1sec (which all makes sense).
The results of loop back tests on the FA-2 give a noise floor (apparently) of around 2.6 x 10-12 at 1 sec which implies an accuracy or one-shot resolution of less than 3 psecs which is ten times better than the top of the range Agilent counter (20psecs).
This doesn't make much sense to me.

I don't (yet) have an FA-2, my present ADEV setup seems to be good for around 10secs upwards and I'd like to get to around 1 sec upwards. The loop-back tests on the FA-2 implies this may be possible (what I want to measure is typically around 2 to 3 x 10^-12 at 1sec) but though the FA-2 is very cheap for what it is, I still don't want to spend £100 if it isn't going to do the job I want to do.
Similarly the much more expensive FSA3011 (around £400) promises less than 1 x 10^-12 at 1 sec but the results on time-nuts implies that this is more like 3 x 10^-12 which is much more borderline.