13 replies to this topic

[Steve Williams]

Watched a test showing the noise on the 7D video capture here; turn off your room lights and watch closely.
http://www.vimeo.com/10473734

This would show that there are some specific ISO sweet spots to shoot at, like 160, 320, 640. etc.

If this is real and I have no reason to believe it isn't, is it a leap to assume the same ISO settings would be an advantage shooting stills? I had noticed that 160 was very clean and had started shooting there but thought it was probably my imagination.

Is there an easy way to test this for stills?

Thanks,
Steve

[embedded video - mods]

[vimeohd][10473734[/vimeohd]

[TheRealDrew]

Watched a test showing the noise on the 7D video capture here; turn off your room lights and watch closely.
http://www.vimeo.com/10473734

This would show that there are some specific ISO sweet spots to shoot at, like 160, 320, 640. etc.

If this is real and I have no reason to believe it isn't, is it a leap to assume the same ISO settings would be an advantage shooting stills? I had noticed that 160 was very clean and had started shooting there but thought it was probably my imagination.

Is there an easy way to test this for stills?

Thanks,
Steve

Interesting video. Take a look here at this Article that one is on the 30D, but it can be in play here also. That example of the noise in the video is wild though...

My guess is the video is behaving like JPEGs in the article.

[wagsy]

Yes I read that on a forum too, the video proves it though.
Strange though how 160 is better than 100.

[betti154]

I've watched a few other 7D reviews/tutorials that suggested ISO 320 and 640 were sweet spots for video.

ds

[echeng]

I had read that 160 was the native ISO for Canon CMOS sensors. I wonder if that might have something to do with it?

[Otara]

Found this:

http://forums.canonp...hread.php?t=958

Claiming similar for 40D for stills.

Be great news if so, but the words 'too good to be likely to be true' come to mind.


Otara

Edited by Otara, 11 April 2010 - 11:24 PM.

[SimonSpear]

Strange though how 160 is better than 100.

Not just that Wagsy take another look - 1250 looks better than 125!

Wow. I'm going to be shooting for the next two weeks topside with a 7D so I'll have plenty of chance to take a close look.

Cheers, Simon

[TheRealDrew]

I had read that 160 was the native ISO for Canon CMOS sensors. I wonder if that might have something to do with it?

The article that I posted talked about that a bit (though I recall reading something else about native ISOs at one point. Could be I was thinking of this, it was awhile ago and took a bit of time digging this one up.)

The primary ISO settings (200, 400, 800, 1600) result from amplifier gain while the in-between ISO settings (125, 160, 250, 320, 500, 640, 1000, 1250) are calculated on the DIGIC chip from the nearest primary setting........

And you thought that ISO 125 is the next lowest in noise ISO setting after 100? Nope! ISO 125 has as much shadow noise as ISO 640.....

According to John, “When the camera is set to 160 and EC 0, it is just like the camera is set to 200 and EC +0.3, except that the 160 damages the RAW data by posterizing it, and completely discards the top 1/3 stop of highlights. For the JPEG user, the latter doesn’t matter because JPEGs always throw away those highlights, anyway”.

[yahsemtough]

Very interesting information if holds true.

Todd

[Drew]

Well that has been tested from the 1Ds MkII (from when I started reading it years ago), trying to find the "native" ISO of each camera.
For video, I believe the native ISOs are 160,320,640,1250 for CMOS sensors in general for optimal noise performance. This is the general consensus by all HDSLR shooters testing, including for filmout peeps. These guys have tested on scopes for optimal performance.
For stills it's more complicated. Noise is only one component in RAW. Dynamic range is also affected by ISO range. So one may get better noise performance by going to 160ISO but the dynamic range is also hampered, where there's less headroom for highlights etc.
Basically, for video, the 160 increments are ideal for noise performance, so long as one exposes dime on to protect highlights. For stills, one has to choose for optimum dynamic range and/or noise performance. Video noise is usually a bit more of a hassle to remove but it's not so for stills.

I'm moving this to video as this is much more relevant to video than stills.

[betti154]

thanks for the info. I've been shooting video at ISO800 on deep wreck dives, but will happily move to 1250 if it will give me comparative or less noice and more light! alternatively I'll move to 640 where possible.

I'm also on a GBR liveaboard trip next week so should have time to try some noise tests in well lit overwater conditions, vs dark wrecks off Sydney. If I come to any conclusions I will post something.

[HDVdiver]

These results are the same as what emerged for the Canon 5d2 a year ago.

Also have a look at:

http://marvelsfilm.w...e-and-iso-test/

In the digital world the relationship between ISO and "grain" isn't the linear correlation we are used to for film.

This is, partly, because the camera's hardware does not support the full true range of ISOs; some are "created" by information from adjacent "native" ISO values. Hence 160 ISO (native) is cleaner than 100 ISO (interpolation).

[Drew]

The hardware just does not support fractional stops but only full stops with the smaller cameras, and uses EV to compensate. That's why 160ISO is cleaner than 200ISO, but has less headroom, because the exposure is +0.3EV stops up from 200ISO.
With 8bit video, highlights are clipped anyhow. So the noise advantage stays. With 14bit stills, the highlight loss is more apparent, the 1/3 stop noise advantage results in a -1/3 loss in highlights hence clipping.
The native ISO is still 100, 200,400 etc, not the fraction stops.

[loftus]

Anyone know what Nikon native ISO is? 200 is the number I see on forums.