Time well spent?

[andrew s]

Just to prove I don't spend all my time in the SGL lounge here is some work I am doing on filtering spectra. Its using Python so I can batch process 10s of spectra.

Currently I am comparing Wavelet an Gaussian filtering. Still very much a work in progress. 

Regards Andrew

[vlaiv]

How do you know if particular feature is noise peak vs regular peak?

Could you apply both on reference spectrum polluted with noise to see which one fares better?

How about adaptive smoothing? I guess you extract spectrum from series of images that you probably stack or maybe you stack spectra extracted from multiple images. Could you use standard deviation of samples and intensity itself to estimate SNR at particular point and then apply smoothing based on SNR (lower SNR - more smoothing, better SNR less smoothing)?

 

[andrew s]

Yes  @vlaiv, I will try test spectra and I also know the instrumental resolution from high signal to noise spectra. I just happen to be using a real spectrum here as I had to program the various .fits related actions using astropy and get to grips with matplotlib.

I am still researching various methods of estimating the optimal wavelet coefficients.

I am primarily interested in changes in the continuum and the H alpha and H beta lines.

All the details in the TiO2 bands are way below the low resolution R ~ 90 of this spectrograph. 

Regards Andrew 

[JeremyS]

So it looks as if you ironed your spectrum flat @andrew s 

[andrew s]

Rain this morning so playing with "Algorithme a trous" still to implement a scientific approach to filtering the wavelet coefficients but making progress.

You imaging boys and girls have it so easy! 

Regards Andrew

Edited November 20, 2020 by andrew s

[JeremyS]

29 minutes ago, andrew s said:

Rain this morning so playing with "Algorithme a trous" still to implement a scientific approach to filtering the wavelet coefficients but making progress.

You imaging boys and girls have it so easy! 

Regards Andrew

Sounds exotic. Will you be having vin blanc or vin rouge?

[ollypenrice]

18 hours ago, JeremyS said:

So it looks as if you ironed your spectrum flat @andrew s 

There's a thought. I wonder if Andrew considers shirt wrinkles as a form of noise? Even worse in the case of sheets when you're trying to get to sleep.

Olly

[Seelive]

5 minutes ago, ollypenrice said:

There's a thought. I wonder if Andrew considers shirt wrinkles as a form of noise? Even worse in the case of sheets when you're trying to get to sleep.

Olly

Let's not go down the ironing route again please 😆

[JeremyS]

10 minutes ago, Seelive said:

Let's not go down the ironing route again please 😆

Already started to slide down that path. Must be the teflon coating on @andrew s’s Tefal

Edited November 20, 2020 by JeremyS

[Paul M]

@andrew s So this is from data aquired with your remote observatory?

No real idea what, if anything, it tells me but it looks impessive!

[Seelive]

Not exactly sure what you are trying to achieve but there's an old article on a simple method to the determine spectra noise published in the STECF Newsletter #42 back in 2007 that can be downloaded from https://www.spacetelescope.org/about/further_information/stecfnewsletters/hst_stecf_0042/

[andrew s]

10 minutes ago, Paul M said:

@andrew s So this is from data aquired with your remote observatory?

No real idea what, if anything, it tells me but it looks impessive!

Yes, I take low resolution, high cadence spectra of flare stars. Maybe 1000 to 2000 a night. Once in a while I catch a flare when the stars output can more than double in a few 10s of seconds. The data is necessarily noise hence the need to filter it.

Regards Andrew 

[andrew s]

10 minutes ago, Seelive said:

Not exactly sure what you are trying to achieve but there's an old article on a simple method to the determine spectra noise published in the STECF Newsletter #42 back in 2007 that can be downloaded from https://www.spacetelescope.org/about/further_information/stecfnewsletters/hst_stecf_0042/

Thanks I am trying to follow best modern practice from zanumber of references. Will follow up on yours.

Regards Andrew 

PS The paper is about measurement rather than filtering. 

Edited November 20, 2020 by andrew s

[andrew s]

Here is the result of the filtering. Still working on how best to display the data. On this 3D plot you can see the H alpha decaying slowly back after the flare.

Regards Andrew

[JeremyS]

16 minutes ago, andrew s said:

Here is the result of the filtering. Still working on how best to display the data. On this 3D plot you can see the H alpha decaying slowly back after the flare.

Regards Andrew

Interesting representation and result Andrew. I was just wondering if displaying as a single surface rather than stacked spectra might make the gross changes more obvious?

[andrew s]

4 minutes ago, JeremyS said:

Interesting representation and result Andrew. I was just wondering if displaying as a single surface rather than stacked spectra might make the gross changes more obvious?

Yes I am looking at surface plots now but  matplotlib is a challenging beast a d I have to fumble my way forward. Regards Andrew 

[JeremyS]

2 minutes ago, andrew s said:

Yes I am looking at surface plots now but  matplotlib is a challenging beast a d I have to fumble my way forward. Regards Andrew 

Fumbling forward is my normal stance 🙂

[andrew s]

@JeremyS here is a surface plot. No one view does it justice so it a fumble into animation next!

Regards Andrew

[JeremyS]

24 minutes ago, andrew s said:

@JeremyS here is a surface plot. No one view does it justice so it a fumble into animation next!

Regards Andrew

Looking good, Andrew. For me that’s much clearer.

Looking forward to your Mexican wave animation.

[andrew s]

@JeremyS this is it for now

Regards Andrew

[JeremyS]

2 hours ago, andrew s said:

@JeremyS this is it for now

Regards Andrew

Super, Andrew 👍🏻

[andrew s]

Now I have subtracted the star so only the flare shows (plus residual noise from the star

Regards Andrew

 

 

Edited November 28, 2020 by andrew s

[JeremyS]

32 minutes ago, andrew s said:

Now I have subtracted the star so only the flare shows (plus residual noise from the star

Regards Andrew

 

 

That’s even clearer Andrew 👍