Hi

Your Rb should be good to about 10 ppt at  1 second. Thats 10 ps. Your OCXO might be good to 1 to 10 ppt at 1 second that’s 1 to 10 ps. Your 5334 has a measurement resolution of 2 ns single shot (which is what matters in this case), that’s 2000 ps. You would like to have a resolution 5X better than the thing you are trying to measure.

Your counter gate time error in ppt scales directly as tau. It’s always 2 ns, but at 10s tau it’s going to be 200 ppt not 2,000 ppt. Your Rb ADEV likely scales as square root of tau. At 100 seconds it’s 1 ppt, which would be 100 ps. You still aren’t there yet.  Who knows what the OCXO is doing, let’s say it’s <= 10 ppt. Still not there.

At 10,000 seconds the Rb might get to 0.1 ppt. That would be 1,000 ps.  Your counter still isn’t there. Who knows what the OCXO is doing at 10,000 seconds. The counter might follow the OCXO.

———————

Simple answer - doing this directly with a counter is not the best approach.

Bob

On Feb 6, 2014, at 5:23 PM, Bob Stewart bob@evoria.net wrote:

Most of us work with what we have. I think it's pretty cool what Bob is doing with a 5335A. Someday he may buy or build something better. Your best may not be his best, or my best. What is best at one tau is not at another. Time-nuts has never been about being best (otherwise we all loose to USNO); it's about learning and exploring this interesting field of time & frequency with what we have at home, whether it's the ADEV of 60 Hz or ADEV of a 5335A/GPSDO, etc.

/tvb

On 2014-02-06 15:13, Tom Van Baak wrote:

I never knew about these different versions of ADEV.
Can you point me to any reference?

Rick

Hi Tom,

If I understand you right, the only thing I need from the adev program is the OADEV values and those are the new ADEV?  Here's the plot for the past 24 hours using "0" for bins.  It takes a few seconds to run, doesn't it? 

Thanks for the ideas on what to measure.  I'm going to replace the EFC divider with low temp coeff resistors and do a software bug fix, and hopefully that will be the end of fiddling with this thing and I'll be able to get some really long runs in.

http://www.evoria.net/AE6RV/GPSstd_PLL/Plots/ADEV.2.6.png

thanks for all the help,

Bob

Hi

My intent was certainly not to stop anybody from doing anything with what they have. My concern is that this is a lot of work for modest return. A simple single mixer setup for  $20 would dramatically change things….

$3 mini circuits double balanced mixer
$3 op amps x 2
+/- 15 V supply you already have (hopefully).
$5 for a piece of perf board
$3 for 3 BNC connectors.
$3 left over for resistors and capacitors.

Offset the local reference one hertz or so and let the 5334 do it’s thing. Your resolution is now plenty good enough to see what’s happening.

Circuit:

Mixer driven by your two sources
L/C filter between the mixer output and a positive gain op amp (OP-37 or similar)
Op amp set up with enough gain to give you 28V p-p when the mixer is saturated (OP-27 or similar)
Next op amp run as an inverter / limiter and driving the counter, use diodes in the feedback path to do the limiting.

Not the most elegant circuit. Not the highest resolution possible.  Cheap / easy to build / simple to troubleshoot. Used a lot of times by a lot of people.

If you want to get fancy terminate the mixer in 500 ohms at audio and 50 ohms at RF. Fancier still is to do some single pole R-C high pass / low pass in front of the first op-amp. Neither one adds much cost. They do make it slightly harder to build.

To totally blow the budget go with an RPD-1 mixer rather than one of the simple ones. Termination would then be 500 ohms at RF and 5K at audio.

Bob

Yes, I’m assuming you get the mixer at the hamfest / 20 piece price rather than just ordering one at a time from Mine Circuits.

On Feb 6, 2014, at 6:52 PM, Tom Van Baak tvb@LeapSecond.com wrote:

I think I see the problem.  I was wondering about using the 1PPS output from my Rb in a test.  Cobbling all that together would be a quick bit of work for you, but I spent my life in IT.  I'm good with a soldering iron, but I readily admit my shortcomings at hardware tinkering.

Bob

On 07/02/14 01:09, Richard Karlquist wrote:

Hopefully this is a good start:
https://en.wikipedia.org/wiki/Allan_variance

In the references I also put NIST SP 1065:
http://tf.nist.gov/timefreq/general/pdf/2220.pdf

Cheers,
Magnus

Hi Rick,

There are a couple of separate issues regarding ADEV.

In old literature ADEV was computed using adjacent segments of data. This is about all you could do with a HP 5360A computing counter. Once real computers got into the game, it was possible to use the "overlapping" version of ADEV, which "milks" more information from the data set. You can see the two different formulas for computing it at: http://www.wriley.com/paper2ht.htm

See calc_adev() source code at: http://leapsecond.com/tools/adev_lib.c

Really, the only thing the overlapping version does is use a "stride" of 1 instead of tau. This is possible when you have the whole data set in memory. The more primitive back-to-back ADEV could be computed as a summing sum, requiring no data storage at all (typical of 60's and 70's instruments).

Regardless of back-to-back or overlap, there's also the question of how many points to plot. Again, in the early days, because both computation and plotting was very time consuming, people tended to plot only a few points per decade. Maybe tau 1,10,100,1000 or 1,2,5,10,20,50, or 1,2,4,8,16,32, etc. To make it look more like a graph they would connect the dots with lines (and guessing). These days, calculating ADEV is so fast there's no need to even draw the lines; just compute ADEV for every tau you can imagine and the dots connect themselves due to their density.

Stable32 has an "all tau" option in which case ADEV is computed for every possible tau. E.g., 1 to 100,000. However, it turns out this is overkill. Not so much for small tau (say 1 to 100), but once you get up to the thousands or tens of thousands there's usually no significant difference between using tau N and N+1. And it can actually take a lot of time to compute ADEV hundreds of thousands of times. So we are now in the era of "many tau" which computes lots of tau per decade. Think of it as a logarithmic sweep of tau instead of a linear sweep. For large data sets this is orders of magnitude faster than "all tau", yet it still fills in all the gaps in the plot with real points, not extrapolated lines. Note that Timelab does "many tau" by default.

And the third issue is, of course, what child in the ADEV family to use: ADEV, MDEV, TDEV, HDEV, etc.

Does all this make sense? I can post graphic examples of all these issues if you're interested.

Thanks,
/tvb

Yes, just grab the code for calc_adev() in http://leapsecond.com/tools/adev_lib.c and always set ovlp = 1. Either way, it's ADEV; it's just that the overlap method gives better results (smaller error bars).

Suggestion: do not label the x-axis "time in seconds". This gives the impression that the axis is some sort of elapsed time. Better wording might be "integration time", or "sampling interval", or "averaging time", or "tau", or "sampling interval, tau".

/tvb

Thanks for the correction.  "Averaging Time, τ, tau" it is.

Bob

On 07/02/14 04:22, Tom Van Baak wrote:

If you read up, you discover that the overlapping trick only came in as
an inspiration for ADEV with a pair of articles from J.J. Snyder which
took inspiration from ways to get laser frequency estimates quickly. The
overlapping technique came to be introduced alongside the tau
pre-filtering for the modified Allan Deviation, as can be found in the
original MDEV article.

Actually, it's not that simple. You actually have a wider palette of
selections to choose from:

Frequency or time stability - ADEV vs TDEV - a scaling issue

Pre-filtering - ADEV vs MDEV - a tau-averaging filter allows better
noise separation

Derivate processing - ADEV vs HDEV - 2nd vs 3rd phase derivate, higher
derivates will surpress more systematic frequency drift components

Degrees of Freedom processing - non-overlapping, overlapping, total, theo

In principle you can choose algorithm from the full combinatorial
matrix, but all the slots isn't filled in, as there is no point in doing
non-overlapping MDEV and TDEVs, since overlapping is so simple and gives
so good performance. There is no point in doing non-filtered TDEV, as
MDEV gives better analysis than ADEV. The scaling is trivial for both.
This is to show that the progress of development in these various fields
have been on-going. TOTAL and Theo is better in general, but might be
more processing than it is worth.

I've tried to convey this on the ADEV wikipedia page, but I'm sure it
can be improved.

Dr. Allan makes the point that one should be using MDEV rather than
ADEV, as MDEV was what he wanted ADEV to do, but could not originally,
so he was so happy to fix it. I agree with his motivation and analysis.

Anyway, what is the problem you are trying to solve? The "correct"
answer depends from case to case.

Cheers,
Magnus

On 07/02/14 04:31, Tom Van Baak wrote:

Averaging time is unfortunate in several aspects, part for it not being
averaging being done and part for the fact that averaging is a separate
processing step that can be done, and you can make a plot orthogonal to
the ADEV plot with various degrees of averaging-pre-filter.

The MDEV uses a tau averaging for a tau observation time.

The tau is really the observation time, and ADEV is the frequency
stability for that observation time, that is the RMS relative frequency
noise when observing it tau seconds later. Similarly the TDEV is the
time stability in RMS s for the observation time tau.

ADEV also rises for long taus on oscillators, where you expect even
better averaging to keep it going further down.

The field is confused enough, so I want to avoid this confusion.

Cheers,
Magnus

I agree.

The phrase "observation time" is probably no better than "averaging time" for the x-axis label; they both give the impression of elapsed time to newcomers. Both "observation" and "averaging" connote run time or elapsed time or experiment duration. I would stay away from the word "time" on the ADEV x-axis completely. Using the word "interval" (as in "sampling interval") is good; the least likely to be confused with elapsed time. Also, it's always nice to add "tau" to the axis label since that letter is universally associated with sampling in T&F metrology.

Stable32 uses "Averaging Time, tau, Seconds". TimeLab avoids the issue by not having labels.

/tvb

On 07/02/14 05:27, Tom Van Baak wrote:

Observation time is better, but it needs to be used properly not to be
confused with measurement time. You can never make a single term become
completely unambiguous from all interpretations, rather you pick one and
define what it means and stick with it.

Observation period is what I've used in the ADEV wiki. Avoids "time" as
it is often confused for elapsed measurement time.

Stable32 and SP 1065 (both by W. Riley) uses the loose term Averaging
time, as that is what it was called initially, before the analysis
became more complex and they needed to separate terms.

Cheers,
Magnus

Somewhat like Bob (Stewart) I've spent my life in IT with a little background in Ham Radio. Now, infected with the Time Bug, I'm slogging through tf.nist.gov, wenzel, and leapsecond to try and level up.

I'd like to ask Bob (Camp) if he has a readily available link for a suitably quite homebrew mixer, similar to the one whose parts list he was discussing?

Thanks to Magnus for his links (later in this thread), to which I'd like to add the following for those on the list, like me, who are neophytes.

Fundamentals of Time and Frequency: http://tf.nist.gov/general/pdf/1498.pdf

Using a Time Interval Counter to Measure Frequency Stability: http://ipnpr.jpl.nasa.gov/progress_report/42-90/90S.PDF

The later link, while a little dated, still does a good job of covering concepts.

Thanks again,

Jimmy...
N5SPE

Hi

I pretty much wire them up on a piece of perf board when I need them. I’ve never bothered to do a real pc board. The OP-27 and OP-37 op amps are getting a bit old / expensive. If I was going to do one, I’d probably do some digging on the op amps first. There may be new(er) parts that are cheaper.

Like anything else, done one up there’s a lot more pain involved doing one than doing a batch. PC boards make things easier.

———

The real benefit comes from a couple things:

1. Mixing down to a lower frequency. The counter gives you 9 digits a second no matter what the (> 1Hz) input. Lower frequencies give you more digits. Any reasonably quiet mixer will do this for you. That includes the DBM’s with connectors on them.

2. Amplifying the beat note as much as you can ahead of the limiter. Slew rate matters. Any quiet audio amp will do this for you, provided it’s got the frequency response and is very quiet at the low end.

3. Limiting with a circuit that has good noise performance at low frequencies. The poor guy who did the counter could not just focus on the low end. We can.

There’s lots of ways to do each of those things. You can easily improve any of the three “chunks” over the very simple circuit I outlined. You can easily get caught up in the optimization process and turn this into a very complex project. For one second ADEV at the 1.0 x 10^-12 level (1x10^-13 at 10, 1x10^-14 at 100, 1x10^-15 at 1000) a very simple circuit will do the job.

—————

The other alternative is to get something brand new with warranty and support like a TimePod. Symmetricom will happily sell you one. They work amazingly well and there’s no muss no fuss comparing a pair of devices. They are just a bit over the $10 to $15 budget (like by three zeros). Accuracy wise it will crush my little circuit, and mine doesn’t come with a cool GUI that reports power levels and the like.

Am I trying to sell you one of these - no of course not. The point is that there’s a massive step cost wise going to new gear. Even at ten cents on the dollar there’s still a massive step. (and yes in this case you can take that as a “anybody want to sell a TimePod for 10 cents on the dollar?” question, I’m in a shopping mood … contact off list of course). I’m not holding my berth for the inbox filling up with offers. Finding this kind of gear used in good condition at a good price is not easy.

—————

One other approach would be group builds. There are a number of these going on off list and aimed at selling the result. Because of the off list nature of this, it’s often tough to know what people are doing and how well they are doing. They tend to turn into limited production run builds and then vanish.

Short runs make it tough for newcomers to stock up on stuff. There simply isn’t enough need for these things to keep a steady supply readily available. That’s not at all unique to this area. A lot of hobby electronics (and mechanics and ..) has exactly the same issue. I could fill several pages with examples.

—————————

Long rambling reply to a simple question - sorry about that ….

Bob

On Feb 8, 2014, at 6:02 AM, Jimmy Burrell jimmydburr@gmail.com wrote: