5D MKIII ISO Test (Controlled Lighting - Studio)

[stageshoot]

OK Below are a set of test images in controlled studio lighting across the ISO Range from the 5D MKIII (Production Body) (UK) Firmware 1.0.7 No Noise Reduction or Sharpening, Direct transfer from RAW to JPG in Photoshop

FIRST - 100 ISO STUDIO LIGHTS BASE TEST



LINK TO FULL SIZE JPG

http://www.ebaystorage.webspace.virginmedia.com/100 (F).jpg

1600 ISO



LINK TO FULL SIZE JPG

http://www.ebaystorage.webspace.virginmedia.com/1600 (86).jpg

3200 ISO



LINK TO FULL SIZE JPG

http://www.ebaystorage.webspace.virginmedia.com/3200 (87).jpg

6400 ISO



LINK TO FULL SIZE JPG

http://www.ebaystorage.webspace.virginmedia.com/6400 (88).jpg

12800 ISO



LINK TO FULL SIZE JPG

http://www.ebaystorage.webspace.virginmedia.com/12800 (89).jpg

25600 ISO



LINK TO FULL SIZE JPG

http://www.ebaystorage.webspace.virginmedia.com/25600 (90).jpg

[neuroanatomist]

Thanks!

Question: did you decrease light levels by stops as you increased ISO (optimal), or did you just increase shutter speed (bias in favor of less noise at high ISO, relevant for shooting sports but less relevant for low light shooting)?

[mpistonephoto]

Not my post to answer, really... but it was Shutter speed.

All the Exif data is there.   Manual Mode.   I only checked the last 2, but went from 1/400 to 1/800.

Mike

[stageshoot]

Thanks!

Question: did you decrease light levels by stops as you increased ISO (optimal), or did you just increase shutter speed (bias in favor of less noise at high ISO, relevant for shooting sports but less relevant for low light shooting)?

This was just quick and dirty - Increase shutter speed to force ISO increase

[GL]

Looks good - about a stop difference from what I'd expect to get on the MkII. Interesting the jump from 6400 to 12800 is not as big as 3200 to 6400, so 12800 becomes more than usable.

I think, thigh, that Canon has hit the limits of its current sensor technology for usable ISOs up to 6400, and 12800 at a push. We'll need to see something very different - preferably in hardware, not software - to see a similar gain in high-ISO noise from the next generation of camera.

If the 1DX is indeed a RAW stop better than the MkIII, that would be very impressive indeed. Somehow I don't think it will be, but let's see.

[pedro]

Well with my current 30D in mind, this looks really nice. If the raws are 1 stop better compared to the 5Dii, how many stops is that to my current body...about 4 I guess. My H1 ISO 3200 don't even look like 25k without pp! This is the dream cam for me, for a long time ;-) About one year more and yes...here I come!

[funkboy]

First of all, thanks very much for posting this, stageshoot.  It looks like you've got a long way towards providing a controlled environment for your testing.  A few suggestions, if I may:

• Clearly you've set out to provide examples of noise at various ISO settings, but it's difficult to judge noise performancealone without being able to also see detail retention.  Your images only seem to have about an inch of depth of field (focused on the rolls of thread) and thus don't have as many exposure zones for judgement of detail retention as they could.  Next time you may wish to stop down enough to make sure that the whole scene is in focus (I assume you're shooting this on a tripod with mirror lock-up to minimize vibration).  Arranging the props in such a way as to ensure that they're all in the same focal plane will make this easier.  Just don't overdo the stopping-down as e.g. f/22 will lower potential maximum resolution due to diffraction.
  
• You might check the three-color histogram; it seems like you're blowing out the red channel on the magenta thread as there's significant loss of detail there.  This may be an issue of being almost blown in 14-bit RAW/prophoto color space & getting clipped when converting to jpg/srgb; using perceptual rendering intent when converting colorspaces may help if this is the case as color faithfulness to the subject is not of concern in this particular example.
  
• It would be nice to have objects in the scene that provide the full range of exposure zones (all the way from solid black to solid white) in order to get an idea of performance across the full dynamic range capability of the camera.  The DPReview standard test scene is an extreme example of this, but it doesn't have to be anywhere near that complicated in order to be useful.
  

Thanks again for posting the test images, they're certainly interesting as shot already.

[Martin]

NICE. If thats pure raw, wo any tweaks with jpg conversion I am impressed.

[unkbob]

Looks like the ISO 100 pic has a much greater DOF than the others.

[jrista]

Also want to thank you for the sample shots. It'll be really nice to explore how improved ISO and noise are on a properly normalized basis.

I would like to second @funkboy's requests though. Particularly the DOF, which is indeed pretty thin, and it seems to shift a bit amongst the first few shots, making comparisons difficult. The 5D III has a diffraction-limited aperture (f/#) of f/10.1, so you could probably shoot at f/9 to gain maximum quality, and increase DOF appreciably.

Also to @funkboy's points, you might want to shoot with the Camera Neutral picture style, and use the same thing when importing with ACR. That is the least attenuated camera profile, and should leave colors as natural as possible without causing any one channel to blow out before the others.

Finally, if you have a Gretag Mcbeth/X-Rite color checker card, including that would be awesome, as it would give us reflectivity from full black to full white, and color fidelity to boot.

Anyway, THANKS for the samples!

[marekjoz]

Thanks for samples.
25600 and 12800 look somehow what I expected. What I'm a bit dissapointed of is 1600 actually. I expected it to be a bit better.

[lida]

Thank you for your sharing!






-----------------------------------------
studio lighting camera flashes softboxes