Author Topic: Help me understand allan deviation measurements.  (Read 43421 times)

0 Members and 1 Guest are viewing this topic.

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Help me understand allan deviation measurements.
« on: November 07, 2015, 06:14:03 am »
For the basics I do understand that allan deviation is a measurement of a oscillators stability. Since my prologix gpib came today I figured that I should try it out. For this test I compared my 5335A's 10811 oscillator against one of my C-Mac CFP-04A oscillators. rather than getting the usual curved allan deviation plot, I got a rather straight line(some lumps) with a negative slope.
I figured that this would assume that the oscillators were drifting the same amount, but to get a number into the E^-17's would be like measuring the noise floor of the counter.
* it has been a few months since I last looked into time-nutting, kind of rusty in that regard.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: Help me understand allan deviation measurements.
« Reply #1 on: November 07, 2015, 06:47:27 am »
Hello,

I guess that the resolution of your measurement is not high enough.
(containing no noise).
So perhaps all measurementvalues show 10.0 (MHz)

With best regards

Andreas


 

Offline VK5RC

  • Supporter
  • ****
  • Posts: 2672
  • Country: au
Re: Help me understand allan deviation measurements.
« Reply #2 on: November 07, 2015, 07:28:11 am »
I don't know enough about how the 5335 uses its ocxo to comment on your graph,    one interesting experiment I did was to use one 10MHz source as both the reference and the test signal,  as a way of looking at the phase noise and other errors of the primary instrument.
Whoah! Watch where that landed we might need it later.
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: Help me understand allan deviation measurements.
« Reply #3 on: November 07, 2015, 07:29:01 am »
12digit/s counters such as the 53230A or SR620 should have a noise floor of about 2...3 E -11 / tau(s).
So ADEV(1s) = 2..3e-11 for example.

You are doing something seriously wrong in your data collection or data processing if you get 1e-14 at 1s.
Without posting more details about your setup it's impossible to say anything more.

 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #4 on: November 07, 2015, 07:58:19 am »
I think I'm making progress, changed from a time interval measurement to frequency. Ran a short test, took quite a while(over an hour).
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: Help me understand allan deviation measurements.
« Reply #5 on: November 07, 2015, 08:06:38 am »
Hello,

I guess you used the wrong sample interval this time for allan deviation calculation.

With best regards

Andreas
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: Help me understand allan deviation measurements.
« Reply #6 on: November 07, 2015, 08:16:51 am »
I think I'm making progress, changed from a time interval measurement to frequency. Ran a short test, took quite a while(over an hour).

That looks more reasonable. You need to have the same gate time (1s) in both your counter and the software so that the statistics go right..

For fun, also do the same measurement with the counter measuring its own reference. i.e. just a cable from the back REF-OUT to the CH1 input. That should give a rough lower bound on the noise floor.
Extrapolating from the 12digit/s counter value of 2e-11 at 1s with your 9-10 digit counter you should see maybe 2e-10 to 2e-9 at 1s
 

Offline edpalmer42

  • Super Contributor
  • ***
  • Posts: 2260
  • Country: ca
Re: Help me understand allan deviation measurements.
« Reply #7 on: November 07, 2015, 05:43:14 pm »
An AlDev measurement is actually a three-way measurement.  At any time value, the number on the graph is the combination of the counter, the reference, and the device under test.  Unless they're all similar, the result is mainly due to the worst of the three.

As awallin and VK5RC mentioned, measuring the counter's internal oscillator is a good way to check the counter's noise floor.  The 5335A has a minimum resolution of 1 ns.  Since counters typically do better than their spec, you should expect to see a 1 sec. AlDev value that's somewhat less than 1e-9.  For larger time values, the graph will extend down and to the right on the graph i.e. 1e-10@10sec, 1e-11@100sec, etc.  Eventually it will become a horizontal line, but you might have to wait a crazy amount of time for that.  This test gives the best numbers you'll ever see with that counter and it coult be argued that it isn't a 'real world' number since you're measuring a source that's synchronized with the counter's timebase.

A more 'real world' noise floor measurement is to take an external oscillator (e.g. your 10811) and connect it to Channel A via a T connector.  Then go through a cable to Channel B.  This guarantees that there will be a delay between Channel A and B.  Now measure the delay and calculate the AlDev.  Make sure that the cable doesn't move during the measurement and keep the temperature stable.  The results might be similar to or maybe somewhat worse than the above measurements.

If the results of either of the above tests are signicantly better than expected, you've got some setting wrong and you need to sort that out before you start making real measurements.

Wherever possible, you should use time interval measurements rather than frequency.  When you use frequency, you're averaging the measurment over whatever gate interval you select.  This can obscure the information you're trying to measure.  Time interval measurements measure a single event.  Some counters can average time interval measurements - make sure you don't do that.  Also, some counters measure time interval better than they measure frequency due to their internal architecture.  My HP 5370B and Wavecrest DTS-2077 are like that.

It's usually a good idea to divide the signals down to a lower frequency before making the measurements.  This prevents the phase offset between the signals from wrapping around.  These wraps must be removed before analysis.  Timelab does it automatically, but it's still better to start with clean data.  If it's not convenient to divide them down, just keep it in mind and try to adjust the frequencies to minimize phase wraps.

Once you've figured out how to make reasonable measurements you can start to understand what they mean.

Ed
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #8 on: November 07, 2015, 11:31:30 pm »
So I ran it again, with some rather interesting results. The ending graph resulted in a negative slope as before, but when the measurement was about 76%(last time I checked) the resulting graph looked like the usual allan deviation graph.
first pic is the resulting graph, second is the frequency difference graph.
In the third image I have zoomed in into the first large peak, I noticed that at random the frequency would go from being really stable, to suffereing a rather large frequency shift, then going back to being stable. hmm
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline edpalmer42

  • Super Contributor
  • ***
  • Posts: 2260
  • Country: ca
Re: Help me understand allan deviation measurements.
« Reply #9 on: November 07, 2015, 11:55:43 pm »
Glitches are a fact of life.  There may be a problem with your setup or it could just be courtesy of Mr. Murphy.  Zoom in on the glitches and use F4 to delete them.  Answer 'Yes' to the question about maintaining phase continuity.  The AlDev graph isn't valid unless the phase and frequency graphs are clear of glitches, phase wraps, jumps, or any other anomaly.

Ed
 

Offline DimitriP

  • Super Contributor
  • ***
  • Posts: 1288
  • Country: us
  • "Best practices" are best not practiced.© Dimitri
Re: Help me understand allan deviation measurements.
« Reply #10 on: November 07, 2015, 11:58:46 pm »
In the frequency difference graph the flattenes of the line (except for the glitch ) looks suspiciously .....flat
   If three 100  Ohm resistors are connected in parallel, and in series with a 200 Ohm resistor, how many resistors do you have? 
 

Offline edpalmer42

  • Super Contributor
  • ***
  • Posts: 2260
  • Country: ca
Re: Help me understand allan deviation measurements.
« Reply #11 on: November 08, 2015, 12:07:05 am »
In the frequency difference graph the flattenes of the line (except for the glitch ) looks suspiciously .....flat

That's okay.  The glitch causes the graph to autoscale and the wiggles are too small to see.  Once the glitch is removed, the wiggles will reappear.

Ed
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #12 on: November 08, 2015, 12:15:43 am »
here is a zoomed in pic of the flat section of the frequency difference graph.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline DimitriP

  • Super Contributor
  • ***
  • Posts: 1288
  • Country: us
  • "Best practices" are best not practiced.© Dimitri
Re: Help me understand allan deviation measurements.
« Reply #13 on: November 08, 2015, 12:30:10 am »
Are you still measuring the C-Mac CFP-04A ? I don't know anything about it so maybe being "unstable" only towards an  "up" direction is normal.

   If three 100  Ohm resistors are connected in parallel, and in series with a 200 Ohm resistor, how many resistors do you have? 
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #14 on: November 08, 2015, 01:13:05 am »
As an interesting note i found out that if i push slightly down on the desk the frequency shifts several 10's of Hz. 
As an aside the tuning voltage(provided by a precision psu) is stable. Shouldnt affect the readings much(less than 40uv p-p).
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #15 on: November 08, 2015, 06:33:12 am »
I re ran it again, I will really have to learn to use this correctly, because I can't. The following screenshots were taken within several minutes of each other.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: Help me understand allan deviation measurements.
« Reply #16 on: November 08, 2015, 08:16:59 am »
I re ran it again, I will really have to learn to use this correctly, because I can't. The following screenshots were taken within several minutes of each other.

ADEV is only for quantifying noise processes.
If you have systematic effects such as glitches or voltage/temperature steps etc. then the ADEV result will be more or less meaningless.
You would need to look at the time-series of frequency measurements and either discard runs with glitches/steps, or perhaps manually delete the glitches.
(ofcourse it's better to track down what is causing the glitches and fix that!)

Playing around with 53230A and/or 53132 counters and paying reasonable  attention to cables, grounding, powersupplies, etc. I have not seen glitches like that.
 

Offline DimitriP

  • Super Contributor
  • ***
  • Posts: 1288
  • Country: us
  • "Best practices" are best not practiced.© Dimitri
Re: Help me understand allan deviation measurements.
« Reply #17 on: November 08, 2015, 09:31:01 am »
Since you mentioned flexing the desk causes a 10s Hz change, you should probably eliminate that first.
If you have a wooden floor, chances are even walking around can make a difference
Kinda hard to make suggestions sight unseen but either try different cables, or shorter cables, keep your cell phone tablet/computer monitor/laptop away from the measurment setup and turn off the ceiling fan etc etc.



   If three 100  Ohm resistors are connected in parallel, and in series with a 200 Ohm resistor, how many resistors do you have? 
 

Offline HighVoltage

  • Super Contributor
  • ***
  • Posts: 5453
  • Country: de
Re: Help me understand allan deviation measurements.
« Reply #18 on: November 08, 2015, 09:56:20 am »
Can you post a picture of your setup?
There are 3 kinds of people in this world, those who can count and those who can not.
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #19 on: November 08, 2015, 04:36:36 pm »
Mini update, fixed the desk issue. Another power cable was pushing up against the 5335A's cable. At least that removes the frequency shifts that were visible on the display. I then re ran it in period mode, and it was a clone of the first graph(the one that goes into the E^-17), untl i shut down everything.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline CatalinaWOW

  • Super Contributor
  • ***
  • Posts: 5170
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #20 on: November 08, 2015, 05:19:16 pm »
Something to think about.  If you are trying to evaluate the ultimate capability of an instrument, deleting glitches etc. may be a valid procedure.  If you are trying to evaluate the performance of your installation as it is being used you can't just edit them out.  They are reality.

Same question goes for "fixing" glitches by using short cables, turning off ceiling fans etc.  If the fix can't be used in normal operation of the equipment, any measurements you make with the fix in place are not relevant to the performance of your system.  This may make for more difficult fixes, or lower real performance of the system, but that is the way it goes.  So I encourage you to fix your glitches, but make sure that your fixes are applicable to your standard use of the system.  Maybe it IS practical to turn the ceiling fan off whenever you are making a measurement.

One of the best ways to make this point seem important is to think how you would respond to a manufacturer of one of your toys that you thought was underperforming who said that of course those specs only applied in a shielded room that was temperature and humidity controlled, operating from a battery supply.
 

Offline DimitriP

  • Super Contributor
  • ***
  • Posts: 1288
  • Country: us
  • "Best practices" are best not practiced.© Dimitri
Re: Help me understand allan deviation measurements.
« Reply #21 on: November 08, 2015, 07:47:54 pm »
Mini update, fixed the desk issue. Another power cable was pushing up against the 5335A's cable. At least that removes the frequency shifts that were visible on the display. I then re ran it in period mode, and it was a clone of the first graph(the one that goes into the E^-17), untl i shut down everything.

Did you save any of the measurement files? I'd be curious to run them here to see what I see.
If they are too large to attach here, I'll PM you my email if you don't mind sharing.

   If three 100  Ohm resistors are connected in parallel, and in series with a 200 Ohm resistor, how many resistors do you have? 
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #22 on: November 09, 2015, 08:13:28 pm »
here are the frequency/phase difference measurements, (reading---measurement)
I can still upload the images if you need them.
I was thinking, should the counter be in talk, or listen mode. currently in talk.
files to large,
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline VgkidTopic starter

  • Super Contributor
  • ***
  • Posts: 2710
  • Country: us
Re: Help me understand allan deviation measurements.
« Reply #23 on: November 09, 2015, 08:21:29 pm »
^ I was only able to upload the frequency excel reading list, here is the current acquire window(gpib not connected)
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline DimitriP

  • Super Contributor
  • ***
  • Posts: 1288
  • Country: us
  • "Best practices" are best not practiced.© Dimitri
Re: Help me understand allan deviation measurements.
« Reply #24 on: November 10, 2015, 12:07:51 am »
I removed the big spurs and it looks "better"
At least you can see there is
A) some kind of disturbance in the force during the measuring period  and
B)  some "settling" going on.

If you start another session, leave them on for 24 hours before trying to collect data


 

   If three 100  Ohm resistors are connected in parallel, and in series with a 200 Ohm resistor, how many resistors do you have? 
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf