GH2 ISOs to avoid...Test results!

D

dvbrother

Well-known member
You can judge for yourself, but it's pretty clear that ISO 250, 500, 1000 and 2000 have the most noise and should be avoided if you want a clean image. What's interesting is that ISO 400 looks noisier than ISO 640.

Knowing that some ISOs use digital gain to increase sensitivity while others don't, I wanted to find out which ISOs were the culprits on the GH2. I recorded pure black by shooting with the lens cap on. I recorded each ISO from 160 to 3200, then in FCP, I brightened the clips by adjusting the midtones and saturation all the way to the right with the 3-Way Color Corrector. What you can see is how much noise there is in the blacks at each ISO.
All shots were on "Smooth" film mode with everything at minus 2.
And stick around for a little bonus experiment I did with the footage I shot this morning.

 
Last edited:
From this it looked to me like 160, 320 and 640 were all fine and I can see that, as two are multiples of the base value of 160, allowing me to assume they would share common elements to their existence. But 200 too and then not 400 ???

I don't have mine yet, is there no ISO 480 offered or even exist, or was that just not selected for this test ???

Disclaimer: I am not an ISO or cam guy yet, still really newb at all of this.
 
ChipThome said:
From this it looked to me like 160, 320 and 640 were all fine and I can see that, as two are multiples of the base value of 160, allowing me to assume they would share common elements to their existence. But 200 too and then not 400 ???

I don't have mine yet, is there no ISO 480 offered or even exist, or was that just not selected for this test ???

Disclaimer: I am not an ISO or cam guy yet, still really newb at all of this.
Click to expand...

ISO 160 is the base ISO of the camera - 320 and 640 are multiples (2x and 3x). Nikon or Canon (I forget the exact details) had the same issue with in between ISO's. It was cleaner to shoot @ ISO 800 and push it a 1/3 or 1/2 stop in post work rather than shoot at ISO 1000 in camera (which is in between 800 and 1600).

Keep in mind cameras only have ONE real ISO - the base ISO. Every other ISO is basically "amplifying" the base ISO. There must be some phenomenon where exact multiples do this more effectively than in between multiples.
 
Last edited:
I am missing the menue "1/3 iso resolution" from the GH1 ...or am I missing something?

KEWl test dvbrother! :-) I lke the snowflakes going up :-)
 
Last edited:
Very handy! Please, don't let this thread fall by the wayside. I vote to sticky or combine with other tips into a sticky.
 
timbook2 said:
I am missing the menue "1/3 iso resolution" from the GH1 ...or am I missing something?

KEWl test dvbrother! :-) I lke the snowflakes going up :-)
Click to expand...

I think it is just on by default now....they added more increments which makes more sense because they changed the minimum ISO to 160 which is actually maybe a little darker than GH1's 100. GH1 ISOs were always kind of out of whack.
 
dvbrother said:
You can judge for yourself, but it's pretty clear that ISO 250, 500, 1000 and 2000 have the most noise and should be avoided if you want a clean image. What's interesting is that ISO 400 looks noisier than ISO 640.

Click to expand...


No question about the multiples of base being better - but from iso 1000 I see that strange horizontal band that many people complain about. Very pronounced, in the lower 1/4 of the screen.

Given that it has been proven that GH1 iso sensitivities are understated, and that the real iso is always higher than that indicated,

http://www.dxomark.com/index.php/en/Camera-Sensor/All-tested-sensors/Panasonic/Lumix-DMC-GH1

does this mean that GH2 performance is actually worse than the GH1? - perhaps due to the increased pixel count?
 
...Um...band vs banding/FPN....there's really no comparison.

This cam seems to behave like the Canons where ISO 400 is noisier than 640.
 
Every CMOS camera has "native" ISO settings. It's when you stray into the Non-native ISO's that you will have problems. Stick with multiples of the base ISO and you'll be fine.
Most Nikons have a base ISO of 200-even though some can be set to "Low-1", which is ISO 100, the results look terrible, compared to the base/native ISO.
I'm not sure why this is....just that it is.
So, if 160 is your base, any multiple of that number, up to a certain point, will yield better results. Try 640(a multiple of base), and compare it against 400(non-multiple), and report the differences you see.
 
Lucas Adamson said:
Has anybody done these tests for the GH1? I'd love to see them if so..
Click to expand...

I just did this test on my GH13, I'm please with the results. Unless I did something wrong, my results were far greater.

Vimeo upload currently @ 18%. 45min remaining. Will post link as soon as it is complete
 
There are articles out there that explain which ISOs are native and which use some form of digital gain. Here's the summary of those articles, and they explain the results of the test I posted in this thread:

Native ISOs in a camera use analog gain, while the others use digital gain or reduction. 200, 400, 800, and 1600 are the native ISOs of the GH2. The ISOs below these native ones (160, 320, 640, 1250) use 1/3 stop of digital gain reduction (kind of like some professional video cameras let you choose -3 gain for cleaner blacks). ISOs 250, 500, 1000, and 2000, which are the ones just above the native ISOs, use 1/3 stop digital gain to boost the sensitivity, and that's why they exhibit more noise than the others.



The conclusion I draw about ISOs 2500 and 3200 is that they just add digital gain to 1600. 2000, 2500, and 3200 all exhibit really bad noise and are barely distinguishable from each other on my test video. I could be wrong, however.
 
Last edited:
Jonathan12uiz said:
I just did this test on my GH13, I'm please with the results. Unless I did something wrong, my results were far greater.

Vimeo upload currently @ 18%. 45min remaining. Will post link as soon as it is complete
Click to expand...

It wouldn't take my .mov file so I had to export as a .mp4 @ 10000kbps downscaled to 720. Granted now, though, you can barely see the noise.

 
thanks,
but your GH13 test method is not the same as what he did on the GH2.. try duplicating the same test as he explained
 
dvbrother said:
The conclusion I draw about ISOs 2500 and 3200 is that they just add digital gain to 1600. 2000, 2500, and 3200 all exhibit really bad noise and are barely distinguishable from each other on my test video. I could be wrong, however.
Click to expand...
So, if 1600 is not enough, 3200 is the only one that makes sense?!
 
You must log in or register to reply here.
Back
Top