Canon EOS 300D vs 350D

[MarkS]

I'm planning a direct imaging comparison between the (plain glass) modified 300D and the (plain glass) modified 350D.  My calculations of read noise and thermal noise indicate that the 350D ought to be noticeably superior.

I already have:
1) The Dumbbell taken with the modified 300D   
      http://gallery.orpington-astronomy.org.uk/displayimage-502.html

2) The Veil taken with the modified 350D 
      http://gallery.orpington-astronomy.org.uk/displayimage-533.html

So, for comparison,  I can either take the Dumbbell with the 350D or the Veil with the 300D.

Which target do you think would provide the better comparison?

[Ian]

my first thought would be the veil, coz it's harder and will stretch the camera more, but then again, it'll mean more processing which might skew the results.

How about your oak tree? No processing, same lens, exposure etc. Maybe stopping the lens down to increase the exposure time without having to mess aound with tracking...

[Ian]

of course, thinking a bit more...

What are you testing it for? You've made quantitative studies of dark noise, QE and so on. What would the "star test" comparison show?

If you know that, it might be easier to pick a test subject.

Or just do both

[MarkS]

It's purely subjective.  Does the final image demonstrate the difference in quality predicted by the maths?

I already have a 300D unmodified Dumbbell and a 300D modified Dumbbell so I might do a 350D modified Dumbbell.  If the theory is correct then it ought to be possible to show a lot more of the fainter detail.  I reckon the S/N ratio on very faint nebulosity ought to improve by a factor of almost 2.

[MarkS]

I managed to take a single 5 minute sub of M27 with the modified Canon 350D last night before the clouds rolled in.

So now I have a direct comparison - 5 minutes ISO 800 taken with the (plain glass) modified Canon EOS 350D and the (plain glass) modified Canon EOS 300D on the Celeston C11 with F6.3 focal reducer.  Both images are taken using the Astronomik CLS filter and they are presented below, side by side.  The images are raw - no darks have been applied.

Before reading further, decide which one you prefer ....



The one on the right isthe 350D (it is slightly larger because it has 8 megapixels instead of 6) and the one on the left is the 300D (it has more hot pixels).

My preference is for the 300D because the colours are brighter and more vibrant.  So what is going on?  I've subsequently done some other tests against uniform white backgrounds.  The main cause seems to be that the pixels on the 350D are 75% of the area of those on the 300D so, on average you would expect them to collect less light.  But surprisingly this affects the different colour channels differently.  The green channel collects 10% more photons than would be expected, whilst the red and blue channels collect 10% fewer.  The gain on the 350D is approx 1.1e/pixel vs 1.3e/pixel on the 300D.  The net effect is that, overall, in terms of ADUs, the green channel on the 350D and 300D is equally bright whilst the red and blue channels are 20% reduced.

So although the 350D is far superior to the 300D in terms of read noise and thermal noise (both are approx half of the 300D figures) this is partially offset by smaller pixel size and the lower red and blue sensitivity.

All in all, the 350D is a better camera in terms of image quality but we may have to wait for the cold ambient Winter temperatures for this to become fully apparent.

For reference, below is the 350D image which I've colour balanced to be as close as possible to the 300D.   If we ever get another clear night, I'll shoot 2 hours of data including darks and then we'll have a much better comparison.

[RobertM]

Hi Mark,

That's a very interesting comparison but I am a little surprised by your results.  The OIII region (blue on 300d and green/blue on 350d) is usually mapped to green so in fact the 350d image looks (more) correct to me.  I believe the OIII emission line is almost exactly in between the green and blue channels but it depends on filter cutoff as to where it actually appears.

I also did some analysis of the 450d and Green is by far the most sensitive photosite on the chip, Red trails a bad third and in fact is the most dominant channel for noise as well.

Have you white balanced the colour channels on both images using sunlight or equivalent (such as a G2V star) ?  It's not possible against domestic light sources as they haven't got the correct colour spectral signature.

Robert

[MarkS]

Robert,

Yes, it's an interesting question about the OIII emission - it sits almost exactly on the boundary between blue and green so, as you point out, it's final appearance is very sensitive to the shape of the blue & green filter cutoffs.

On the side by side comparison image, no colour balancing whatsoever has been done - the image faithfully represents the ADU counts in the pixels. 

It would certainly be interesting to RGB balance both images against a known reference.  I'll give this a try (by either using an unsaturated G2V type star in the image or by taking advantage of sunlight to determine what the relevant scaling factors should be). 

[RobertM]

I think you'll find the Sun a difficult one ... to do it will require either a neutral filter or a calibrated white surface - both tricky in practice.

This is a good start for G2V stars :http://www.astropix.com/HTML/I_ASTROP/CUSTOMWB.HTM

[MarkS]

What colour does an OIII filter look like to the human eye?

That would tell us, at least, how the eye maps the colour.

[Ian]

Not really. Mine looks bluey-green, but the brain performs pretty flexible white balancing too.

How about setting up a spectrograph and image a spectrum with both cameras both with the same settings, then compare the histograms? A sunlight spectrum is probably the easiest (assuming it's not cloudy), or failing that a halogen lightbulb.

[RobertM]

The generally accepted colour for OIII is 'teal' which is more greenish than blueish.  Check out http://en.wikipedia.org/wiki/Teal_(color) - describes Teal as 'Teal Green' so you can make up your own mind on that.