Author Topic: Help me understand allan deviation measurements. (Read 43856 times)

edpalmer42 · « **Reply #100 on:** December 16, 2015, 04:26:33 am »

So, the three traces are all for the same oscillator? Green - 1 day ago, Pink - 4 weeks ago, Blue - 6 weeks ago? And that oscillator has been powered up almost continuously over that period?

The green trace has higher noise levels than the others. That's why it's been pushed up. Notice that at the 1 second point, it's at ~9e-10 whereas the other traces are ~6e-10. The lack of ripples in that graph may or may not be significant. It could be that the noise is masking the ripples.

The part of the green trace from 200 seconds up looks quite good. A flat section like that shows the noise floor of the measurement. There is no hint of aging like there is in the red and blue graphs. But again, it might be that the noise is masking the aging.

I don't know what else to tell you. If you want to buy some more oscillators to compare with what you've got, go for it.

Some good brands of quartz oscillators are Wenzel, HP, Symmetricom, Oscilloquartz, and MTI. There are others - that's just off the top of my head. Stick to 10 or 5 MHz unless you have a reason to get something different. Newer units have the benefit of better technology, but older ones have the benefit of time to stabilize. The longer an oscillator runs, the better it gets. An old, high quality oscillator can be a magical thing - or a paperweight.

Make sure that you get ones that have EFC - some don't.

Rubidium oscillators on ebay have gotten expensive in the last few years and it's hard to tell the difference in specs between things like the LPRO, FE-5650/5680, or X72 so there's not much point in stocking up on them.

Ed

Vgkid · « **Reply #101 on:** December 16, 2015, 04:50:46 am »

I ended up having to retype the last post. Green-about 1 day powered on, pink-4wk power on. Blue- 6wk powered on.
I did not do any editing to the blue traces file, no frequency spurs.
I will attach the timelab file in a few.

Vgkid · « **Reply #102 on:** December 16, 2015, 06:03:35 am »

Here is the TL file comparing the cfp04 to the lpro-101

Vgkid · « **Reply #103 on:** December 18, 2015, 01:40:36 am »

Ran the lpro(ch1,50-ohm) against the HP10811. No oscillation occur, but it is rather different. Unfortunately the at the bottom of the graph is flat, but lumpy.
Pics /files later.

Vgkid · « **Reply #104 on:** December 18, 2015, 07:20:30 am »

Abra,Kadabra,Alakazam.
The oscillation is still there. Instead of starting at the 20's=30's mark. It starts in the 400s mark.

DimitriP · « **Reply #105 on:** December 18, 2015, 06:53:21 pm »

It is interesting in a magical, cant-get-rid-of-the-wiggles kinda way.
I'm running out of assumptions

Vgkid · « **Reply #106 on:** December 19, 2015, 06:55:11 pm »

I'm enentually going to get the other 5335A online. need to fix my car first.
I assumed that the wiggles would be gone if I just ran the 10811 vs LPRO. Since the internal oscillator of the counter should be rather well shielded. Next I will put the stp/cfp in a AL box,, with coax running to the counter. Heck I my try your suggestion, of the twisted pair power connections(No coax unless I get desperate).
Else

DimitriP · « **Reply #107 on:** December 19, 2015, 07:35:20 pm »

Quote

Since the internal oscillator of the counter should be rather well shielded.

TrustNo1

Altough it's "shielded" by the tank that is the 5335, there are a couple of things I dont like.
For one, it's right there in the air intake stream.
Depending on the room temperature that must be causing the oven to come on "every so often".
Even some bubble wrap around it so the air doesn't hit it directly would be an improvement.

The second thing that I don't like very much is that whatever vibration the fan produces gets transfered to the oscillator since it's bolted directly on the PCB. The 10811 has "pretty good specs" concerning vibration but there is also a reason why HP produced the double oven , shock mounted version !

Vgkid · « **Reply #108 on:** December 27, 2015, 01:44:27 am »

While working on the 6114A i noticed that the led grounded/ungrounded light on my Phillips surge protector either stays on,off, or as currently alternating between those states.
Ordered some parts for my original 5335, so i will soon have another counter to try.

DimitriP · « **Reply #109 on:** December 27, 2015, 03:53:31 am »

Sounds like it's time to replace the surge protector too...

Vgkid · « **Reply #110 on:** December 27, 2015, 09:16:12 am »

It has gone back to showing grounded, and staying that way...
I think an exorcism will be needed as well.

Vgkid · « **Reply #111 on:** December 31, 2015, 04:22:01 pm »

Mini update:
Waiting on parts from mouser so i can fix the counter. I forgot about the relay socket last order.
10.86 MHz issue, made a discovery.
So the STP2145A and the CFP04 are wired in parallel to the same psu which is set to provide a luttle more current then is needed, under steady state conditions.
On Tuesday we lost power after both ocxo's warmed up the psu switched from cc to cv mode. I checked the 2145A last night, no major issues. Checked the CFP04 that 10.86 MHz appeared again.
powered off the psu, set the trip current higher(enough so that cc mode wouldnt engage) and powered on the psu.
All is well.

Vgkid · « **Reply #112 on:** January 22, 2016, 07:28:52 am »

Mini update V2:
Rather than trying to spend time and hunt down the wiggles in the graph. A plot I ran last night(just posting Al Dev,FreqDev) showed some massive jumps in the graph. So these jumps have seemed to almost be present in every plot I have made once it has started getting cold,even noticed some starting off. In a previous post I mentioned that the surge protectors ground light would often flicker on/off, or just stay off. So I also noticed that(posted about this earlier) walking across the room/ sometimes moving the gear would majorly throw off the readings. So I swapped the power cords with others(more solid fit, at least with the counter), of both the counter, and the psu(powering the oscillators). Same effect.
Next: I separated the 2 surge protectors, placing the counter/psu on the newer/better one...Same problem.
Next: ran each surge protector from a different outlet...same.
Next: ran the surge protector without a ground(used one of those 3prong to 2-prong adapters) ...same.
Next: hooked up the surge protector up to the wall(no adapter), wiggled the plug, no movement. pushed on the plug, no weird changes.
Next: I tried to replicate the actions of the anomalies, pushed on the desk. Sometimes this would get a change, but not always. Moved to the psu/ counter, same thing. walked across the room, sometimes a issue. Noticed that it only did anything near the center of the room. singled it out to a few square feet in the middle of the room.
The only thing I can think is happening is that there is a bad connection in the chandelier in the room below mine.

DimitriP · « **Reply #113 on:** January 22, 2016, 09:11:50 am »

Do you have a shortwave radio to take listen up, down and around 10 MHz to get an idea of what sort of "loose RF' you have there ?

edpalmer42 · « **Reply #114 on:** January 22, 2016, 04:05:55 pm »

Do you have a UPS? Try disconnecting everything from the wall and run it off of the UPS. Noisy power or ground could inject noise into your setup.

Isn't it fun to play in the parts-per-trillion range?

DimitriP · « **Reply #115 on:** January 22, 2016, 07:38:16 pm »

Quote

Isn't it fun to play in the parts-per-trillion range?

A long time ago, as long as the FM receiver's AFC could track it , an oscillator was deemed "stable enough".
So it is a little bit of a change !

Vgkid · « **Reply #116 on:** January 22, 2016, 08:56:08 pm »

An interesting aside to the large jumps is that i was using my 3456a to measure the output of my precision Lambda psu and couldnt notice any jumps. I was even using 1nplc and no filter.
@ DimitriP: i believe I still have a SW radio. I wws probing a upset psu(for ripple issues, giving unstable readings) an could notice the 10mhz signal(not freq measureable) on the ripple.
@ edpalmer: Last week I was browsing CL for one.

Vgkid · « **Reply #117 on:** February 08, 2016, 07:15:20 pm »

I further went through and removed any switching psu's except the laptops(running it iff a battery for 2hr didn't help, didn't save that one). Also removed any other sources of 10mhz oscillators. Removed the one from the 5335A(both the ocxo, and the xo*) As evident in the graph, the oscillations are still there. Even used a twisted pair of power cables, and a different psu as well.
Then ran the ocxo into frequency input, and into chA(effectively measuring itself) No oscillations in the graph.

DimitriP · « **Reply #118 on:** February 09, 2016, 08:49:13 am »

One or more of our assumptions is completely wrong. Now we need to figure out which...

Vgkid · « **Reply #119 on:** February 10, 2016, 04:25:44 am »

For those who are curious, I'm uploading the results of the test with only the ocxo's/RB powered up. * On the lpro box, a bunch of the outputs are terminated. The laptop is running on battery, hence the short run time.

edpalmer42 · « **Reply #120 on:** February 10, 2016, 04:47:09 am »

And when I zoom in on your data, here's what I see on the phase data screen. A low frequency beat signal that takes about 40 seconds to complete a cycle i.e. approx 0.025 Hz. It didn't show up in your noise floor tests, but every time you try to compare two oscillators, it's there. Until you can find the source and beat it to death, those oscillations will appear in all your data.

Have you tried using a different counter?

Vgkid · « **Reply #121 on:** February 10, 2016, 05:02:25 am »

I can use another 5335 in about a week. It might get interesting in a few months, I will move the whole setup to the garage.
* I will try running the lpro against itself in a few days.
**Reminding myself to pick up some sma terminators.**

edpalmer42 · « **Reply #122 on:** February 10, 2016, 05:37:11 am »

Can anyone else duplicate Vgkid's tests with a 5335? Is there something weird about that model that's causing this?

Vgkid, maybe check your 5335 over carefully. Look for cracked ground traces, bad solder joints, loose screws, etc. One of these sort of things might not show up in a noise floor test because everything is local or internal but could mess up a real measurement where you're looking at external devices.

Vgkid · « **Reply #123 on:** March 14, 2016, 05:48:27 am »

Remember that power issue I noticed when it came to the grounded light on the surge protector.
Well I was checking my ocxo's. Touched the counter with 1 hand, and the metal bench with the other and noticed a tingling feeling.
Grabbed the fluke 87, and measured 60v, did that with my other pieces of gear, the same thing. Though a few reported about 11v.
I will troubleshoot further tomorrow.

DimitriP · « **Reply #124 on:** March 15, 2016, 03:02:53 am »

Quote

a tingling feeling

Better than strange voices coming from the TV !


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Help me understand allan deviation measurements. (Read 43856 times)

Share me