Canon 600D versus the rest

[MarkS]

I got a chance to try out the Canon 600D this weekend.  So I thought I would take some test images to determine it's parameters for comparison against other Canons for which I've done similar tests.

The read noise at ISO 6400 is an astonishingly low 2.1 electrons which gives it excellent performance in low light but it has a relatively high dark current (even with the rear LCD switched off).

I loved the flip-out, articulating rear LCD and the 3-10x digital zoom when making HD movies.

Anyway, back to the parameters that determine it's usefulness for astro-work.  Here is a graph of how the dark current builds up over a 2 hour period of successive 5 minute exposures:


Note that the above test was done with the rear LCD switched off because with the LCD switched on, heat builds up rapidly.

The dark current (per pixel) of the 600D is roughly the same as the old 300D and 350D

Here is some data at ISO 800:



Note the read noise for the 600D is the lowest of the bunch.
The gain for the 450D and 600D are much lower than the rest, mainly because they are 14 bit cameras (instead of 12 bit).  But the gain has no real impact on performance - it's the read noise and dark current that matter.

In the next table I produce estimated figures for total noise (read noise plus dark current noise) at different ambient temperatures.  (Note that the 450D is not included because I don't have dark current data):



Note that the 600D has the lowest noise per pixel under typical shooting conditions.

However, I'm not sure this is the measurement we should be looking at.  Suppose you take the same image with all cameras and then blow it up to A4 size and then examine the noise.  It is the noise per unit area that becomes important.  The second part of the table shows noise per unit area of sensor.  The 350D (still my chosen workhorse for astro-work) comes out on top.  The 600D comes out significantly worse with getting on for twice the total noise per unit area of the 350D.  This is mainly because of the dark current - the dark current per unit area of the sensor is 3x worse than the 300D which means the noise is 1.7x  (square root of 3) as bad.

P.S. If anyone is up for doing tests on the 450D, 500D and 550D then I can send you instructions for what shots I need.

Mark

[JonH]

That's some pretty intresting stuff!
Which reminds me, I have taken all the dark frames for the 450D, just need to get round to the rest of it....

[Fay]

Did not know you had a 600D Mark

[MarkS]

Did not know you had a 600D Mark

I do now!  Bought it on Friday to replace my ageing and increasingly unreliable Nikon D70.

Mark

[Fay]

Is it for normal photography?

[MarkS]

Yes - the 600D is for normal photography. 
I'm not going to mod it for astro-stuff.

[Fay]

glad i made the right decision then, when i bought mine

[MarkS]

glad i made the right decision then, when i bought mine

Maybe yes, maybe no.  Gary Honis's tests give an indication that the 550D has more dark noise than the 600D both at the start and the end of a 2 hour continuous test run.  But I would prefer to see hard numbers in units of electrons - like the tests I've done.
http://ghonis2.ho8.com/T3iReview/T3iReview.html

It's one of the factors that persuaded me to opt for the 600D.  In practical terms there's not going to be a noticeable difference though.

Mark

[Mac]

Interesting.

Have you got the maths behind it, I's like to take a few readings of the D3,
It would be interesting to compare.
Mac.

[MarkS]

Have you got the maths behind it, I's like to take a few readings of the D3,
It would be interesting to compare.
Mac.

Yes, I'll post the maths and also a spreadsheet that you can enter your data into - it can then perform the more difficult parts of the calculations.

If you want a head start, look here:
http://forum.orpington-astronomy.org.uk/index.php?topic=3473

But I 've vastly improved the dark current calculations now - back then I had no idea that the sensor warmed up during use and that dark current increased as it warmed up.

Mark

[Mac]

Ta.

[RobertM]

Interesting.  How did you calculated noise per unit area ?

Robert

[MarkS]

Interesting.  How did you calculated noise per unit area ?

I used the pixel size of the 300D as a base.  Then for the other cameras I did the following calculation:

Noise_per_unit_area = Noise_per_pixel * (300d_pixel_size/pixel_size)^2

[RobertM]

Ah ok you're using simple scaling.  It's a shame manufactures don't consider image quality more important than the dash for resolution.

Robert

[MarkS]

Thanks to the hard work of JonH I now have complete data for the 450D:


Notice how the dark current graph for the 450D sits almost on top of the others (except the 400D).  I'm now suspicious of the anomalous 400D data - I'll have to go back and check it.

Here is the updated table of total noise:  [Later Edit - ignore the noise per unit area - these calculations are incorrect)


I forgot to mention that it is the overall signal to noise ratio that is the most important criterion.  Quantum efficiency has been increasing slowly - see this table for instance:
http://forums.dpreview.com/forums/post/34536743

Ah ok you're using simple scaling.  

You're right - I may be calculating this wongly.  I need to start with the SNR per pixel and then scale it up in the correct manner (whatever that is) per unit area.  I'll think about this.

[Later edit - I woke up in the middle of the night and worked out how to do this properly.  Assuming the figures for QE in that link are correct and then calculating the signal-to-noise ratio per unit area instead of noise pre unit area, we now get very mixed results.  The 600D should give the cleanest pictures for everyday photographic use and for Winter long exposure astro-imaging.  But it will be the worst for long exposure Summer astro-imaging.  I'll publish the figures later]

Mark