Canon 450D's noise level is worse than 350D

[dogbert]

Well it seems semi official now and confirms my view, having upgraded from a 350D to a 450D, that the 450D has

worse noise at high ISOs than a three year old model.

 

http://www.photozone.de/dslr_reviews/395_canon_eos_450d?start=2

 

This also goes against the generally accepted consensus internet view that Canon has been able to increase the

megapixel count in successive generations, while holding noise levels fairly constant.

[johnson_d.]

Bummer.

[bryantan]

How does it compare when you downsize the 450D's files to that of the 350D?

[dogbert]

I haven't done my own specific testing but in viewing 450D pics on my 17 inch laptop display I thought I noticed increased noise at ISO 400 and even slightly increased noise at ISO 200 in blue skies for example. With the 350D I had little concern shooting at ISO 400 and none at ISO 200, and that is viewing at the same size on the same screen.

 

Photozone has confirmed my suspicions.

 

On the 350D I seldom went above ISO 800, as I felt 1600 was best avoided. Withth e 450D it looks like ISO 800 might be best avoided too, at least for my purposes which lean towards landscape and travel.

[PuppyDigs]

However the 450D is 12MP whereas the XT is 8MP. So even if the 450D has a little more noise you're not likely to notice it

in same sized prints. Why? 50% more pixels means less rezzing up for large prints, so artifacts will be less apparent. For this reason I

suspect a 12x18 print from the 450D will be cleaner than one from the the 350D. Of course if you only view prints at pixel level this is a

moot point...

[dogbert]

I view on images screen at the same size, ie full screen, so the less rezzing up of the 450D is already accounted for. My impression is that the 450D has worse noise. I am not saying the noise is bad, but I do think it is a little worse.

 

As for prints I can't really tell as anything I print at 13x19 will typically be shot at ISO 100 or 200.

[sarah_fox]

Interestingly, my 5D has slightly more noise than my 10D, even though the pixel density is about the same.

 

Puppy's point is probably valid in general, though: When pixel density is increased at the cost of quantal noise, it's probably a wash in the final image. Taken to a wild extreme, imagine if we had a sensor with such high resolution that individual pixels *might* get hit with a photon but probably wouldn't. You'd end up with an image file of mostly zeros, with a large number of ones thrown in and perhaps a rare 0010 (2) or 0011 (3). The printed image would be very pointalistic when viewed microscopically. The noise would depend mostly on exposure, which would of course determine the photon/point density in the image.

 

Working towards the other end (low pixel density), one only groups quanta into tiny square areas. The cruder the grouping, the lower the resolution. However, the noise should remain roughly the same.

 

Then why the differences between my 5D and 10D? I'm wondering whether my 5D has slightly smaller photodiodes to accommodate the greater number of traces. The same factor might be at play when comparing the 350D and the 450D: More pixels mean more traces mean less room for photodiodes. I don't really know enough about chip architecture to know whether this is true. I've asked before, and nobody has had an answer, as to what the sizes of the photodiodes are throughout Canon's sensor evolution.

[sarah_fox]

I wrote, "However, the noise should remain roughly the same."

 

... er, I meant in the sense of viewing the whole photograph.

[Brad_]

>>> Well it seems semi official... <P>

 

Oh no... Drats...<P>

 

<center>

<img src= "http://pages.sbcglobal.net/b-evans/Images34/StocktonTunnel750.jpg">

</center>

[marc_bergman1]

For noise comparisons I prefer to see actual samples rather than just test charts.

<p>

If you look at the DPReviews test of the 450D you can see the camera retains fairly low noise and high detail up to 1600 ISO. If you look at the next page you can see that Canon's DPP does a great job at reducing noise.

<p>

<a href="http://www.dpreview.com/reviews/canoneos450d/page19.asp">DPReviews Test of the Canon 450D.</a>

<p>

Just using test charts to determine noise characteristics can be deceiving. They don't show the effects of heavy-handed noise reduction.

<p>

[derek_hofmann]

"I view on images screen at the same size, ie full screen, so the less rezzing up of the 450D is already accounted for."

 

Your screen can't display every pixel of either sensor, so the method it uses to resize the images for display is important. If it simply removes entire rows and columns of pixels until the picture fits (the faster method), you will still see more noise from a 12MP image than an 8MP image from the same size sensor because you're still comparing photosites of different sizes. You would need to resize the 12MP image to 8MP (or both to a common size) or make prints of each to do a fair comparison.

[KenPapai]

How does it compare when you print an 8x10 inch the 450D's image to that of the 350D?

[des adams]

Brad ... I like the image.

[dogbert]

Ken

 

I might do some test prints shot at ISO 400 and 800 to check and report back. I no longer have the 350D but I have plenty noiseless of prints from it shot at ISO 400 and under. With the 350D noise was visible in prints but not too bad at ISO 800, while I considered ISO 1600 best avoided if aiming to produce a fairly clean, detailed prints.

[Brad_]

>>> Brad ... I like the image.

 

Thanks Al - was from my XSi at ISO 1600...