Highlight Tone Priority ISO Noise Tests

Disclosure: These were done only to satisfy my own curiosity regarding how ISO behaves with the Highlight Tone Priority setting enabled. They are synthetic tests with the lens covered, intended to reveal noise in shadows in a way that is easily reproduced for each setting.

Probably should take this with a grain of salt. I did everything I could to avoid any mistakes, but it's always possible I missed something. I suggest you do your own testing... Especially if you don't like the results or disapprove of a synthetic benchmark. Post your results!

A still was taken first to record camera settings in EXIF and a 5 second video recorded
The MOV files were labeled by corresponding still EXIF data
The MOV files were converted to Cineform
Cineform files adjusted to bring out as much noise as possible so it's easy to observe
Scaled down, labeled, posted here

Nice job! Very interesting results. It seems 160 and its multiples are better then most others. I wonder why?

Nice job! Very interesting results. It seems 160 and its multiples are better then most others. I wonder why?

yes...very nice ex of 160 and its multiples..as well as the negatives of HTP

Here are a few more examples I put together to set a frame of context

Wow, this is enlightening. Esp. since I just shot for three days with HTP turned on...

Thanks for the tests shoots INFURNO!

Wow, this is enlightening. Esp. since I just shot for three days with HTP turned on...

Thanks for the tests shoots INFURNO!

How did it turn out?

Some guys swear by it, I'm still evaluating it's usefulness.

Because I underexposed like a *, everything is pretty usable. The "video look" in the highlight areas feels like it is reduced (lots of shinny, specular high-lighted blood in frame), but there is a little more screen-door red noise pattern to the footage. Again once reduced to 720p, like everything coming out of the 5d, it cleans up nice.

I'd like to post a few of the dailies but I'm hesitant to jump the gun...

Because I underexposed like a *, everything is pretty usable.

If you're not clipping your highlights then there's no need for HTP. Specular highlights will clip no matter what. Or am I missing something here?

BTW great test Infurno. Does anyone have an explanation of why the upper intermediate ISOs provide good results?

Thank you very much! I'm also wondering about those 160 ISO multiples as well... I wonder if there's something that makes it situation-dependent.

ISO 100 = ISO 100
ISO 125 = ISO 100 + digital push --> more noise
ISO 160 = ISO 200 - digital pull --> less noise, less dynamic range
ISO 200 = ISO 200
ISO 250 = ISO 200 + digital push --> more noise
ISO 320 = ISO 400 - digital pull --> less noise, less dynamic range
ISO 400 = ISO 400
and so on

The 5D II supports only ISO (100?),200,400,800,1600,3200 native. All other sensitivities are created in post-processing.