VLFN Characterization of Reference Devices / Topologies

[zhtoor]

-

[Andreas]

Hello,

you should look at the "LISA" papers.
There are some examinations for using voltage references for Laser Interferometers.
As far as I can remember they use a pair of equal references and measure the difference to characterize the low frequency noise.

with best regards

Andreas

[EmmanuelFaure]

(q3) "can popcorn noise be filtered?"

Yes
http://www.wat.edu.pl/review/optor/11(1)45.pdf

[Andreas]

Hmm,

it means that popcorn noise can be distinguished from a more or less well known signal with same amplitude.
But for a reference the question is different.

E.g. on my AD587LW#04 I measured up to 6 uVp popcorn noise.
On the last popcorn screening measurement I did not get one single jump.

So the price question is: is the reference now at the 6uV higher or at the 6uV lower level.
Or has it even different levels? And might be somewhere in between?
And in which direction will it jump next?
Which of both levels is the one with the higher probability.
Or will the propability of the level even change with environment conditions?

with best regards

Andreas

[Andreas]

Hello,

i do not really think it is bi-stable.
There are different jump amplitudes even on the same device.
One time I caught a positive jump followed by 2 successive negative jumps.
(note there is a high pass 0.1 Hz which makes a e-function following a jump)

with best regards

Andreas

[branadic]

it does look like that the popcorn noise is some kind of a quantum phenomena.
(some kind of a "bi-stable quantum tunneling" state?)

No, it's not a quantum phenomena.
Heavy metal iones, impurities and crystal imperfections are traps for electrons. Depending on the "size" of them, they have different time constants before they release a captured electron. Todays seminconductor are more pure so what is left are impurities and crystal imperfections, that cause the popcorn noise.

-branadic-

[3roomlab]

with the resolution set to 6 1/2 digits (fast mode) and averaging set
to 10 last samples (repeat mode vs. rolling mode) resulting in a sampling rate of 3.89Hz (257 millseconds/sample)
to be as close to 10NPLC as possible.

if i remember correctly my old keithley K2015, based on the look of the coarseness of the graphing (ie bit depth). the resolution of your meter i think is too coarse for 6.5D. you probably need at least 100NPLC to have the last digit become meaningful.

[branadic]

I came across this paper today A Methodological Approach for Detecting Burst Noise in the Time Domain and wonder if someone has successfully analysed burst noise in time domain using wavelets.
There is a package for GNU Octave called ltfat that supports the function dtwfb.m but until now I wasn't able to understand how to use it similar to what was done in the paper above. Can someone help or has another methode to extract burst noise?

-branadic-

[branadic]

have been trying to install LTFAT for octave for sometime now 

Simply extract the package into the folder ...\Octave-4.4.0\share\octave\packages if you have the latest version of Octave installed and load the package using the command pkg load ltfat within Octave. That's fairly easy, but I didn't understood yet how to use the function dtwfb and associated functions to seperate the burst noise from the zener voltage of my reference. The help files are not helpful either.
Next step would be to analyse the different levels, time intervals and time durations of the jumps.

-branadic-

[branadic]

There's no need to post data or waveforms, it's just a generell question as almost any reference seem to have burst noise, even though some of them show only rare  events and need long observation time.

-branadic-

[branadic]

I measured sixteen LMx99 with Pipelines low frequency noise amplifier (0.1 - 10Hz) yesterday.
My references are sitting in precision sockets on a proto boad. I found two LM399 from National Semiconductor with large changes in low frequency noise if I blow air on them. All other references showed absolutely no changes.
To verify that I made no mistakes I interchanged the references on the proto board without any difference in results. So it seems that their heater is unstable, thus not the best choise for building a stable reference.

Has someone made similar observations on low frequency noise as an indicator for temperature instabilitiies of heated references?

Attached are two shots of the same reference, first picture with the reference covered from air currents, second picture with slightly handmade air currents.

-branadic-

[Andreas]

Hello branadic,

Is the material of the legs of the two devices the same?
(Thermal EMFs ?)

Different "hat" material perhaps?

Photo of the setup?

with best regards

Andreas

[branadic]

Hello Andreas,

the two LM399 are from a batch of four LM399 with same date code made by National Semiconductor with same lead material, that I order some years ago from RS Components or Digikey. Can't remember that correctly. Two of them behave like described while the other two don't show such an effect nor does any of the Linear Technology specimen.
The board only has an individual 7.5k resistor, a 100nF capcitor and the precision socket for each LMx99 feed by 15V, that's it.

-branadic-

[GerryBags]

I'm wondering if some of the noise you guys are picking up is from secondary cosmic rays, cosmic noise in other words. Cosmic ray flux at ground level is higher than it has been due to the weakening geomagnetic field (https://www.sciencealert.com/new-study-shows-that-earth-s-magnetic-field-is-weakening-more-rapidly-than-we-thought). This noise affects any measurement conducted over time, but obviously it depends on the sensitivity of the measurement whether or not this influences the final result.

There are algorithms for removing the effects of cosmic rays on digital images, I'm not sure about other data sets. You can get hold of curves that show the short-term fluctuations in cosmic ray flux due to solar storms (and even terrestrial storms, I was surprised to learn) and perhaps this could be correlated with periods when the noise in your measurements is higher?

If you want a real-time look at what the CRF is like in your location when you are carrying out your measurements, you could knock up a simple cloud chamber. Many designs work with dry-ice to super-cool the isopropanol vapor, but you can do the same think with butane decanted into  a pyrex vessel with a covering to keep it liquid and have a copper rod as a heat exchanger connecting the liquid butane to a copper plate in the bottom of your alcohol chamber. If handling liquid butane makes you nervous I have seen designs for cloud chambers using peltier coolers, which should be ok for a small volume of vapor and as you only need to get an idea of the frequency of the rays in the same volume as your circuit's elements occupy this shouldn't be a problem (the noise affecting a whole instrument rather than just the voltage ref might be a bit trickier... especially with something like a Fluke 5440