Multichannel time series analysis of Vref data

[rhb]

Is anyone here doing this?  Autocorrelations, crosscorelations, correlated noise removal, etc.

I've seen lots of comments about seasonal variation, but have never seen it quantified.

Reg

[Conrad Hoffman]

The math is way beyond me and probably many others. There are some really sharp people here (you included!) but when you look at the overlap between people who can, and people doing references, you probably get a number in the single digits at best. 

[KE5FX]

You might try playing with Stable32 if you haven't already.  It's meant to be a time-nuts tool rather than a volt-nuts tool, but there's obviously a lot of overlap when it comes to the math. 

[rhb]

I can't write the software to do it if I don't have data.  If I had received data files 10 years ago everyone would have an OSSW program to do it.  No need to learn all the math though that does help.

Everyone is using FFTs without understanding them.  Not knowing the math will get you in trouble in fringe cases, but most of the time it's just "blow and go".

Consider 100 references in a single enclosure packaged such that one may reasonably assume small variations within the enclosure.  If the ambient temperature changes, all the references will shift by a certain amount relative to each other.  Even without temperature data, given enough data deriving the temperature change from the voltage changes is just an algebra problem.

FWIW there's a lot of overlap with the behavior of quartz crystal oscillators too.  From a mathematical physics perspective they are all related to well known and reasonably well described phenomena.

Reg