OK Tonight (30/3/09) these are my results.
I have posted
etacar -from VSpec -red hydrogen, green helium, purple spec of eta
etacar-20sec-specfitraw.zip -raw spec image
etacar-20secfit-speciris.zip -image through iris y=7
etacar-20sec-speciris.zip -.spc file from VSpec
I hope, for you to fiddle.
**I think my previous spec may have been contaminated by a fluoro light**
Given the shape of the Ha area.
Have fun, I did, I spent the afternoon chaseing a 1m black snake away from the observatory, I still don't know where it went . Double checked every inch of the obs floor before entering tonight.
Last edited by theodog; 30-03-2009 at 10:19 PM.
Reason: Add info
Still caught up trying to process some Ha solar images; but a quick look shows that the emission has disappeared....
This one now looks more like the early spectra... how did you figure the "contamination"
This one now looks more like the early spectra... how did you figure the "contamination"
Last night while imaging I noticed the trees around light up very slightly. I asked my partner if she had turned a light on and yes she had. I restarted the run. She told me she did the same during the nights observing and I didn't complain, - I was inside using MaximDL sequence so I didn't notice the light.
I don't know if it did effect the spectra, but it has to be noted.
Yes we are still talking.
Woohoo! Daylight saving is gone... clear sky and in the obs tonight!
This is the first spectrum (half) processed with my SA in my new filter wheel.
Shot this spectrum of Eta Carinae so I can measure the dispersion of the new setup - 8.98 A/pixel with the camera directly behind the filter wheel. It means maybe a bit more processing with the DMK21 since it's nearly always 2 frames to join to get the full spectrum, but that's OK at least for now. Maybe a DMK41 might appear on the scene.
Seeing was pretty ordinary.
I shot some more spectra of Sirius, betelgeuse, Hadar, Alpha Cen, Gamma Crucis, etc yet to be processed.
Nice spectra Al, I do like the Sirius spread. The hydrogen lines seem to line up a cracker.
I couldn't greet the end of daylight saving with starlight as I have my gear striped down for RYSTARS and a school observing night. As soon as I have my gear together I will have a go at Sirius myself.
Attached is a spectrum of Betelgeuse captured last night. It has been corrected for camera response based on the response curve generated from my Sirius Spectrum of the same night. Now if the camera response curve is accurate, shouldn't the corrected spectrum agree with the library spectrum?
The blue spectrum is my Betelgeuse corrected with the camera response from my Sirius spectrum. The green is the library spectrum for M2i (which is the nearest to Betelgeuse in the VSpec library).
Notice the library spectrum is low at low wavelengths, and high at high wavelengths. This suggests to me that the camera response curve is not correct.
I've been thinking for a while that to get a true camera response curve would require averaging the response curves from the full range of star classes (O to M) so that the low signal to noise portions of the spectrum can be handled properly.
Who's been a busy boy then...
I've been using the few minutes of sunshine to image with the SM60 ( and get practise in processing Ha images) The nights have still been too cloudy to be able to do ANY imaging!!!
Al, I think the fall off of the camera responce below 4000 and above 7000 is so rapid/ steep that the smallest error in compensation with throw things out dramatically. I know I have that problem with the Canon 350 ( the 300D has the Baader UV-IR filter fitted, and that gives a very clean cut-off at 3950 and 7000)
Certainly try different reference stars, you may be able to find a compromise.
What if you calibrate on the Betelgeuse and then "correct" the Sirius spectra????
Who's been a busy boy then...
I've been using the few minutes of sunshine to image with the SM60 ( and get practise in processing Ha images) The nights have still been too cloudy to be able to do ANY imaging!!!
Al, I think the fall off of the camera responce below 4000 and above 7000 is so rapid/ steep that the smallest error in compensation with throw things out dramatically. I know I have that problem with the Canon 350 ( the 300D has the Baader UV-IR filter fitted, and that gives a very clean cut-off at 3950 and 7000)
Certainly try different reference stars, you may be able to find a compromise.
What if you calibrate on the Betelgeuse and then "correct" the Sirius spectra????
G"Day Ken.
I've been playing about with similar arcane ideas .
Another thing that could influence this is the black point on the image. This lot of spectra that I took has the black point at about 11000 counts on the intensity axis. Probably the result of either not having the brightness control set up zero, or maybe from using so much gain in order to keep exposure time down (due to seeing). 11000 divided by next to nothing gives a fantastic lie for camera response...
That zero error or black point noise I think could be contributing to the spurious camera response curves. What do you think?
I started having a play with subtracting a constant from the spectrum and I think that's part of the way to go. I'm about the call it quits for the night though.
So, I'm starting to think I need to put a step into my work flow, that I subtract the black point offset from the raw spectrum before doing any other operations.
I think that if that doesn't fix the problem, then that combined with averaging the response curves (across the star classes) should.
Hi Al,
Do you do all the processing frames; flats, dark, bias?
I have the same issues that my instrument responce seems untransferable but I figure it's because I only do darks.
I had the same thoughts about instrument responce for all classes.
Perhaps we could use others final spectra as library and compare?
It will be the weekend before I'm up and running again.
G'Day Jeff
I haven't been doing darks or flats at all. So far most of my spectra have been captured as AVIs and stacked in registax, so that smooths out the noise. So far it's only Eta Carinae that I've captured as a still (as it's not very bright). But you are right, subtracting a dark should remove a lot of the black point noise if it is due to brightness/gain settings on the camera.
Subtracting the black point offset in VSpec should have a similar effect though.
Here's the fruits of tonight's foray to the fringes of sanity.
I reworked Sunday night's Sirius and subtracted the Black Point offset before doing the division to generate the camera response. The results of the reworked spectrum are in the first image. I found the spectrum was crap below about 3400A.
I then used the new camera response curve to process Betelgeuse captured on the same night. The result was rubbish. It slightly shifted the hump in the raw spectrum but did not get close to the library spectrum above about 7000A.
So I generated another camera response curve from Betelgeuse. Again I subtracted the black point offset first, taking care not to make any values negative. It was a bit of suck-it-and-see to get the subtraction right, if it isn't right the division curve will look like rubbish where it approaches zero, but it's a simple matter to correct the offset adjustment and try again.
The final camera response I got from Betelgeuse is very different to what I got from Sirius. What I like about it, or what I find encouraging at this stage, is that it is similar in shape to the chip manufacturer's published response curve.
I have put all 3 camera response curves in the last image for comparison.
The next step will be to test the Betelgeuse response curve, and see what it does to the Sirius Spectrum...
The Imaging Source curve for the DMK21 chip (ICX098BL) shows about 50% at 4000 and 7000.... this seems to match the Pink curve??????
Very interesting stuff!
A dark and sky correction may make some difference , but looking good...
Al
I think the camera response curves are going to be dodgy for a few reasons. I'm not sure how we can avoid this. The A stars (like sirius) arn't very bright in the IR so the levels are low compared to the green etc. When you try to divide another spectrum by this curve it enhances the IR more than it should. I tried this with a red star and got a vastly enhanced IR end to the spectrum that wasn't real.
The same problem will occur if you use a red M star as the response curve and try to calibrate a hotter star but with the opposite effect of over enhancing the blue end.
I suppose a more realistic result would be achieved by making a curve from a "B G and M' star and averaging them. The difficulty here would be how to normalise them first so that the weighting for each spectrum was even. Maybe normalise all of them at a specific green frequency and go from there.
In the end I don't think it matters terribly as we arn't measuring the flux at a particular frequency although this is possible with reference stars etc.
The Imaging Source curve for the DMK21 chip (ICX098BL) shows about 50% at 4000 and 7000.... this seems to match the Pink curve??????
Very interesting stuff!
A dark and sky correction may make some difference , but looking good...
I'll have to examine the curve again ... from my memory of it - based on shape only, I thought the green curve was better.
Quote:
Originally Posted by Terry B
Al
I think the camera response curves are going to be dodgy for a few reasons. I'm not sure how we can avoid this. The A stars (like sirius) arn't very bright in the IR so the levels are low compared to the green etc. When you try to divide another spectrum by this curve it enhances the IR more than it should. I tried this with a red star and got a vastly enhanced IR end to the spectrum that wasn't real.
The same problem will occur if you use a red M star as the response curve and try to calibrate a hotter star but with the opposite effect of over enhancing the blue end.
I suppose a more realistic result would be achieved by making a curve from a "B G and M' star and averaging them. The difficulty here would be how to normalise them first so that the weighting for each spectrum was even. Maybe normalise all of them at a specific green frequency and go from there.
In the end I don't think it matters terribly as we arn't measuring the flux at a particular frequency although this is possible with reference stars etc.
I agree. In theory, if we had a uniform light source across all frequencies, we could very simply get a perfect camera response.
Because we are using non-uniform light sources, any area where the signal is low becomes very inaccurate.
I think I can see 2 possible ways to establish a wide spectrum camera response curve:
Average the normalised camera response curves from B, G and M stars as you suggested (if necessary expand that to include the other classes too, but I don't think that would be necessary);
Join the low wavelength section from the B spectrum, to the middle from the G, to the long wavelength section of the M class. Might be a bit fiddly scaling the connections, but this should address the magnification of errors/noise that a low signal causes.
I don't want to sound defeatist but, are we asking for an accuracy that is unattainable? Is it too hard to map the CCDs?
Would we need to ensure that the spec falls onto the same pixels each time? Maybe each pixel has a different responce to each wavelength. So if the spec image is moved between two exposures it would give a different response.
A single dimention pixel array should be easier to calibrate.
One way to check this maybe to allow the spec to drift up across the camera array and measure the levels along the extended Ha line for example. Do they change along the Ha line? (Illustrated by Terry's image)
ie I think we need to get the image to land on the same pixels each time.
I see what you're getting at, Jeff. The only problem with drifting the spectrum that I see is when it comes to fainter stars - more background stars to affect the spectrum. After thinking about it a bit, I don't think the pixel to pixel variation will matter too much, unless you're using a slit or can focus the zero order image into 1 pixel (or near enough to). Even a well focussed spectrum is several pixels wide so any pixel variation is just a small part of the noise in the image.
After some more playing about I'm just about at the point of shelving it for a while. Basically, for my camera anyway most of the response curves I make are pretty close in the range from 4000 - 7700 A which is visible light (OK it's a bit short on the violet end).
Outside of that range is really where the trouble starts for me. At the UV end the response is useless IMO, but there's a tantalising signal amongst the noise in the IR, and this is what I've been trying to sort out.
My main reason for pushing all this is for stars like Eta Carinae for which we can't access a library spectrum like the others. Unless the camera response spectrum we use is repeatable and transferable, then what do we use? The camera response from the hottest star we have access to?
So I'm just about at the point of giving the camera response thing a rest for a while, concentrate on 3900-7700 A and take it from there.
BTW I played about making a hybrid camera response curve. I averaged the black point adjusted Sirius and betelgeuse curves, and then cropped it at 7700A and joined on the IR end of the Betelgeuse curve. Like all things, it was a compromise.
Al.
Last edited by sheeny; 09-04-2009 at 07:15 AM.
Reason: afterthought
away from the observatory, I still don't know where it went . Double checked every inch of the obs floor before entering tonight.
It means maybe a bit more processing with the DMK21 since it's nearly always 2 frames to join to get the full spectrum, but that's OK at least for now. Maybe a DMK41 might appear on the scene
.
. Time for bed!
.
