GH2 Mega GH2 ISO Noise test w/ Quantum beta V9b. Interesting results.

The test is slightly flawed in the mids and highs. The exposure was not completely even on my piece of paper, and getting each exposure to match exactly was all but impossible. It would be better, I think, to use a diffusion sheet on the lens, lit from a soft dimmable source. BUT, I was curious why the 160 showed more banding than the others. It takes an extreme filter correction to bring it out, so maybe it's not a big deal.
I was a bit concerned with the blocky little squares that showed up in some. Almost like we were seeing sensor pixels or something...? who knows.
 
I guess the trick with 640 is to always remember when turning on the camera to go into iso and switch to 800 and then back to 640
 
jhero said:
Seems like it, just did a quick test and I saw a difference
Click to expand...

Happy to say, we're not seeing it here. Can post some still later, but not seeing a difference after restart.

EDIT: only with 640... we are indeed seeing a difference with 320 after on/off. Sucks. 320 looks SO freaking good when it's right, but it spooks me how easy it is to mess that up.
 
Last edited:
Just looked at my Real World test of 640, 400, and 200 outside, med exposure on everything, and I would say that the noise is equal among them.
 
Tim Joy, if you do more tests please consider testing 320 again (with the high ISO to low trick), as, while we've avoided 320 in the past, it's in fact giving a very clean image here.

Could it be that different ISOs behave differently with different film modes? 320 good with standard but not smooth, etc? Must continue to investigate.


EDIT!!!! - Want to have your mind blown? Do these same tests in HBR mode with firmware 1.1. 320 (coming from a higher ISO) is cleaner than all other ISOs. I just went back to seQUake and indeed 320 sucks all the way around with firmware 1.0, but with the new firmware at least, it's a different beast. Now since 1.1 isn't hacked yet, it's not a big deal, but when it is, keep this observation in mind. Also, we have a shoot in LA next week. Web based thing, long takes, 29,97, and so will be using HBR (need audio too) and... maybe 320!

Also, just an aside, but has anybody noticed how much better the noise behaves in HBR vs 24p in 1.1?
 
Last edited:
There have been a few videos showing the image improvements of HBR mode, not just in regards to noise but other areas like sensitivity and dynamic range, not sure why that would be
 
Devon's now infamous thread summary!

Smooth -2,-2,-2,-2 seems to be a good middle ground, while ISO 640 could potentially be the cleanest. Over-exposing looks to be the best route.
 
Cosimo Bullo said:
...
Could it be that different ISOs behave differently with different film modes? 320 good with standard but not smooth, etc?.........
Click to expand...

Could be the case. In the last couple of months I have always shot my stuff in standard mode and I got not the impression that iso 640 is that good. I have often made the experience that you´ll get different test-results shooting in real life than with a lens cap on. I´ve just made a very quick test and shot some tiles at iso 320, 640 and 800 - take a look:


http://vimeo.com/35418312

Password: noise_test

Is iso 640 really that much better in these shots? Does the different lens or the different patch make a difference?
 
Bob-
I always have my histogram and Highlight alert on. So when i'm shooting something I will expose for the highlights, and in many cases there are areas that are still blinking and showing over exposure, but at least I know. Those two in combination with the little exposure meter that reads +/- 3 stops (or is it EV?) will tell you exactly where you are.
I remember from days with the GH1 that I found it better to switch to a higher ISO and get a really hot image, than use a lower one and just barely get exposure. This test shows how most of the noise difference is in the lows, it evens out in the mids, and the highs are quite clean till you get to the illegal iso's.
 
Uwe-
I don't see a huge difference through vimeo at least, even on HD. I couldn't have told you which is which. I'll have to try 1.1
 
Again, they all look very close. When you enhance the contrast it seems to get progressively more noisy as the iso goes up. I just installed 1.1 so I'll give it a try. Can't believe we're looking at walls and lens caps... :)

EDIT- I uploaded a photoshop file that has all the iso's 100% grabs with adjustment layers to play around and see what kind of noise is being captured. Download HERE.
 
Getting the same results here as Uwe. 320, in Standard and using the technique, may be cleanest of all. Certainly better than 160 re dithering (or whatever you want to call it re gradations). Getting this result with fireware 1.0 and seQUake and and 1.1 stock.

(an aside, take a look at how clean cinema is at 1600 with Standard - amazing)
 
You must log in or register to reply here.
Back
Top