More DSLR Ha questions

[ollypenrice]

Hi All,

I'm still trying to help a guest with images taken in a modded DSLR through an Ha filter. SInce I don't use a DSLR I'm not at the height of my powers!

Here's an image in RGB TIFF simply opened in Ps. No calibration has been applied and no processing. It was opened in AstroArt in RAW and converted to TIFF for export into Ps.

As you can see, it's massively black clipped but I have no idea why, or what the histogram of a typîcal Ha linear raw looks like from a modded DSLR. How normal does this look?

And why is it so clipped. I never touched it, honest Gov!

Thanks in advance,

Olly

[Photosbykev]

the obvious answer is exposure too short or iso too low depending on aperture but I don't think you would be looking for something that simple?

Was the histogram on the camera or capture software very similar?

[Beardy Bob]

I can't help much, but I can definitely tell you that this is not normal for a Ha shot through a DSLR so something has definitely gone wrong somewhere down the line!

Maybe:

Settings when converting in AstroArt? (I always convert in the Canon software.)

Exposure time?

[ollypenrice]

Exposure was 20 minutes and 800 ISO... I wouldn't call that short for a DSLR? F ratio was 7.

Olly

Edit; I just looked at the histo of the RAW in Astro Art, something called a CR2 file, and it was also clipped in this way.

[Photosbykev]

Exposure was 20 minutes and 800 ISO... I wouldn't call that short for a DSLR? F ratio was 7.

Olly

the exposure sounds plenty Olly. Conversion problem then? if you don't have DPP try loading it directly in PS using the Adobe raw convertor

[Robny]

Exposure was 20 minutes and 800 ISO... I wouldn't call that short for a DSLR? F ratio was 7.

Olly

Edit; I just looked at the histo of the RAW in Astro Art, something called a CR2 file, and it was also clipped in this way.

A 'CR2' file is a Canon Raw file

[Photosbykev]

I'm wondering if AA5 is transparently processing the CR2 file during import. Personally I'd try to load the CR2 direct into PS via Adobe raw convertor, also look at the histogram on the back of the camera which should be pretty accurate. A quick alternative is to download "Faststone image viewer" which can display the CR2 file directly

Failing that, post the CR2 file onto dropbox or mediafire so we can take a look at it.

[Gina]

As I recall there's a setting for noise reduction which works by black clipping the data but I don't use a DSLR for astro imaging any more.

[Photosbykev]

As I recall there's a setting for noise reduction which works by black clipping the data but I don't use a DSLR for astro imaging any more.

yes, long exposure and also High iso NR, hopefully they were turned off

[JacobvonChorus]

When I did Ha through with my DSLR I found a large amount of exposure was required. On the Pacman, I required 10.5 hours of exposure to get the noise down. Long exposures are required as well. With the DSLR, dark calibration is a requirement I found. Darks around the same temperature (the temperature is critical) are required, and lots of them. After calibration, get rid of all the green and blue pixels. In Nebulosity, there is an "Extract red" option. This will leave you with a monochrome image at 1/4 the original pixel count. The green and blue data are useless. Then register/stack as normal. The linear image is somewhat difficult to process because of the noise but as long as there was enough exposure to begin with, it should be just fine. Getting the required exposure and the darks is a real pain though. I spent hours trying to regulate the temperature of my DSLR using a fan (reduce temperature) and a blanket (raise temperature) to get the temperature close enough to get useful darks. If the darks aren't good enough, then the image will be very noisy. Those reasons lead me to get a CCD, CCD is awesome!

Jacob

[GlassWalker]

Exposure was 20 minutes and 800 ISO... I wouldn't call that short for a DSLR? F ratio was 7.

I'd say it was under-exposed. It's not a short exposure, but the optics are slow. Workaround would be increase ISO so that the image data gets above the zero point.

For indication, for most Ha subjects other than the brightest, I'd aim for at least a couple minutes at f/2.8 or faster, ISO1600, on a modified camera.

[ollypenrice]

I'd say it was under-exposed. It's not a short exposure, but the optics are slow. Workaround would be increase ISO so that the image data gets above the zero point.

For indication, for most Ha subjects other than the brightest, I'd aim for at least a couple minutes at f/2.8 or faster, ISO1600, on a modified camera.

That's twenty minutes per exposure, of which this is a single frame. It seems to me that if the data is pre-clipped like this there isn't much you can do with it in processing. I just don't understand where the clipping is happening.

Olly

[GlassWalker]

To relate it to what I use comparing relative brightness only, and disregarding noise:

20 minute exposure at f/7 is equivalent to 3.2 minutes at f/2.8.

But you're one ISO step lower, so that's equivalent to 1.6 minutes.

I think in a previous thread you said it was an unmodified DSLR, that's at least another stop loss. You brightness levels would be equivalent to well under a minute exposures in my configuration.

Generally for checking framing I might use 30 second exposures, and Ha is hardly detectable. For brighter subjects I go to 2 minutes which is still a fight against the noise floor. 4 minutes help a bit more although I don't go longer since my local light pollution levels dominate there.

You do need to set a DSLR so that the signal is in the useful output range. If you can't get more light on the sensor, that means jacking up the ISO. If you're going to try again, at the least turn up the ISO to 1600. If you have faster optics that'll help a lot too.

It looks like the rosette lurking in the top image, if it might help I can see what I got previously for comparison?

[ollypenrice]

It's a modded camera. The stacked image is not bad at all but what is bugging me is simply why it's black clipped. Where did the left hand halsf of the histogram go???

Give me a CCD!!!

Olly

[opticalpath]

Olly, when I examine the image you posted, pixel values in the Red channel are just above zero - not clipped. Blue and Green channels ARE clipped, so when you view the RGB histogram it will look clipped. What does the histogram look like for the Red channel only?

Also, in the PS Levels options, what values have you set for clipping the histogram view? Is it possible it's just the histogram *view* in Levels that's clipped, not the data?

Adrian

[GlassWalker]

There's numerous research papers out there where various people have tried to figure out how a DSLR processes images, particularly for use in astro. If you want to dig through all that it might have the answer. But fundamentally, next time you try turn up the ISO, it'll shift everything to the right and help get things out of the output noise.

[opticalpath]

I just looked at the image in PS using the Histogram/ All Channels View. There is (low level) data in the red channel; only the green and blue channels look clipped ...... so the composite RGB view also looks clipped.

Adrian

[RikM]

I haven't looked at the histogram but it is taken through an Ha filter, so blue and green will be significantly lower signal than red. It could be that blue and green are not 'clipped' but just not above the floor yet.

If the red is not clipped, and you are intending to add the Ha to a data set taken without the Ha filter, could you just pull the red channel, process as for mono and add it to the red from the non-filtered stack (high contrast / blend mode lighten etc.)?

[opticalpath]

I haven't looked at the histogram but it is taken through an Ha filter, so blue and green will be significantly lower signal than red. It could be that blue and green are not 'clipped' but just not above the floor yet.

If the red is not clipped, and you are intending to add the Ha to a data set taken without the Ha filter, could you just pull the red channel, process as for mono and add it to the red from the non-filtered stack (high contrast / blend mode lighten etc.)?

Yes, blue and green channels are effectively absent (not 'clipped' as I said), so this image is in effect monochrome. If you view the histogram in composite RGB view mode, it will always look clipped because of the zero pixel values in blue and green that mask the red data that IS there. The PS histogram in 'Luminosity' view (or red channel only) shows the actual Ha data.

Adrian

[ollypenrice]

Thanks for the pointers, all. This is weird, though, because for me the red channel when stripped out or examined in Levels in the RGB image most certainly is clipped - according to Ps.

I've tried separating the red, looking at the red within RGB and converting the RGB to greyscale and in all cases the histo is clipped.

I think I'll just give up on this one! I'm flummoxed.

Olly

[opticalpath]

Olly, what I don't get is that when I run the cursor over the image in PS, it shows red pixel values that are low .... but not zero - see the screen grab above - definitely not clipped in red.

Disregarding the histogram for a moment, what pixel values do you actually see in the red channel of the image when you run the mouse over it? You probably already checked this, and sorry if it's stating the blinking obvious, but is it possible you have the black clipping level set above zero in the histogram view options in Levels? Even if it's set at a low % value, the histogram would *appear* clipped if the (non-zero) pixel values were below that threshold. In the screen grab above, I set the Levels black-point clipping option to zero.

Adrian

[Gina]

I've done NB - both Ha and OIII with my modded and cooled 1100D by importing the stack into PS and removing the G & B channel data (in the case of Ha) using levels. The important thing was not to use the level shifting in DSS or that caused data clipping. Then convert to greyscale and carry on as with a mono image, I used ISO 3200 or even 6400 which was feasible when the camera sensor was cooled to -15C. Of course it was an extremely inefficient system and I needed very long exposures (I forget just how long now but hours in total as I recall). What an enormous difference I found when I got a mono CCD camera! Exposures are counted in minutes rather than hours and everything is just so much simpler! I was quite pleased with what I achieved with the DSLR but so much wasted time and effort in comparison - it's mind boggling!

[Shibby]

Hello,

If what you say is true, Olly, about the Red histogram (and I have no reason to doubt you!) then something is not right... See attached images, which is one of my 10min Ha exposures. RGB combined does have that "clipped" look (G+B have no signal, not much noise either) but the red channel certainly does not.

You have to watch out for RAW loaders. By default, at least with Photoshop's RAW loader, they "enhance" the blacks by truncating the histogram. I can't see why Astroart would do that, though... The only application I have that I'm 100% sure does not mess with the RAW data is DSS, so I often view images in this.

You might want to try loading the RAW without debayering (e.g. superpixel mode) if Astroart allows this, in case the debayering interpolation is dragging down your red channel.

[jackrussell0232]

Olly, guys, come on. The dead giveaway here is the lack of noise. Unless Olly has moved from Southern France to the South Pole, 20 mins at iso 800 would show significant noise and far more hot pixels, even with the lens cap on, forget the Ha filter. I'll try to get a 10 min dark from mine from my other laptop to upload.

Something in the conversion from CR2 must have gone wrong. I capture using nebulosity to laptop, or save to the card and convert to tiff using the canon software.

Attached is an Ha DSLR (modded 550d) image, shot with an Altair Wave 115 at F7. This is 4x10min at 1600, so a single sub would be close to what you'd expect from 20 mins at iso 800.

[Gina]

Nice image