GH2 Noise Tests

[filmstox]

I've been enjoying my new GH2s for a week now and have started running a battery of tests to get to know the little guys better. Here's the first results.

These are Macbeth charts for each of the color profiles designed to display the noise profile of each. Each clip shows all of the available ISOs with shutter angle adjustments to maintain a constant exposure. In these tests, the exposure was set based on the camera's internal meter, even though it results in underexposure by varying degrees. I snapped a photo of the LCD for each clip which is visible at the bottom of the clip. There is no post processing other than transcoding and adding this overlay.

The clips were transccoded to ProRes in FCP and output as H264 in Compressor.

















I have the original card files as zips if anyone would like them, the entire batch runs just over a gig for the card files with stills, so I'm hoping someone can help with hosting the data if the full files are useful.

 

[Ian-T]

It's cool that you did this. Without looking through all of them can you give us your thoughts on the results?

 

[mpgxsvcd]

One thing to note is that the camera will choose different ISO values for each color profile to get what it says is accurate exposure. Here are the values the GH2 will select for each color profile.

Cinema - Zero gain
Standard - Zero gain
Dynamic - Zero gain
Smooth - Minus 1/3rd of a stop gain in ISO
Nature - Zero gain
Nostalgic - Zero gain
Vibrant - Plus 1/3rd of a stop gain in ISO

Now we also know that what it says is accurate exposure for each color profile varies as well. I think that Cinema is 1/3rds to 2/3rds of a stop underexposed, Smooth is 1/3rd of a stop underexposed, and Nostalgic is 2/3rds of a stop overexposed.

Basically I don't use Smooth because I think it is too far underexposed even though it has very good dynamic range. I use Nature because it tends to get the exposure right and still produce reasonably accurate colors.

 

[dvbrother]

One thing to note is that the camera will choose different ISO values for each color profile to get what it says is accurate exposure. Here are the values the GH2 will select for each color profile.

Cinema - Zero gain
Standard - Zero gain
Dynamic - Zero gain
Smooth - Minus 1/3rd of a stop gain in ISO
Nature - Zero gain
Nostalgic - Zero gain
Vibrant - Plus 1/3rd of a stop gain in ISO

Now we also know that what it says is accurate exposure for each color profile varies as well. I think that Cinema is 1/3rds to 2/3rds of a stop underexposed, Smooth is 1/3rd of a stop underexposed, and Nostalgic is 2/3rds of a stop overexposed.

Basically I don't use Smooth because I think it is too far underexposed even though it has very good dynamic range. I use Nature because it tends to get the exposure right and still produce reasonably accurate colors.

Those conclusions are a bit different than mine. I shot a DSC Labs chart in a studio and also just a plain gray card in different lighting scenarios. With the gray card I checked the cameras spot meter reading against my Sekonic light meters incident reading. Cinema is the most accurate according to my light meter. Smooth is about 1/3 stop hotter than Cinema. It's possible my light meter is inaccurate, as it's an analog one.
When I used the DSC labs chart, I also found Cinema to be 2/3 stop underexposed and Smooth 1/3 stop underexposed when I compared the camera's meter reading against a waveform monitor. Though, in reality, shooting actual footage looks just fine with both of these modes. What I mean is, it doesn't look underexposed even though, according to test charts, it is.

 

[filmstox]

Conclusions

Conclusions

I'm still processing the tests, doing split screens and whatnot... but first impressions would agree with mpgxsvcd's observations on exposure variance between profiles.

The camera, and lighting for these charts is in a studio situation with no variation. The light measured an even 80fc across the chart as measured with my recently calibrated Sekonic 608c.

I set the exposure based on the camera's internal meter, interestingly the histogram accurately shows the underexposure that the meter called for. You can see it in the still at the beginning of the clips. So if I am interested in aggressively protecting my highlights I will pay attention to the meter, otherwise the Histo is giving a much more accurate representation of final exposure.

Since these tests were mostly intended to compare noise across various settings, I will comment on that.

I was reading some other threads that talked about ISOs to avoid, those threads actually inspired me to run this little batch of charts. I don't see a radical departure from a fairly linear increase in noise across the range. I was expecting one or two ISOs to pop out as either better or worse then expected. I didn't come away with those conclusions.

There does seem to be a slightly different quality to the noise at ISO 1250 and above. More aggressive.

I will have more thoughts after I have a chance to dig in a little deeper in the next day or two.

 

[dvbrother]

I was reading some other threads that talked about ISOs to avoid, those threads actually inspired me to run this little batch of charts. I don't see a radical departure from a fairly linear increase in noise across the range. I was expecting one or two ISOs to pop out as either better or worse then expected. I didn't come away with those conclusions.

There does seem to be a slightly different quality to the noise at ISO 1250 and above. More aggressive.

I'm the one who did the "ISOs to avoid" tests, though in some ways I wish I hadn't. Even though I feel the "lens capped" tests I did are accurate and in some ways valuable, when shooting real-world footage I find my "ISO's to avoid" (below 1250) to be fine as far as quality. Also, someone else did a test that seems to contradict my results. His shows excessive chroma noise at ISO 320 and 640. I performed my own test based on that, and while I found 320 ISO to be fine, sure enough, 640 ISO has an excess of chroma noise in the shadows compared to ISO 400 and 800. So after taking what I can from all these tests (yours included), so far I feel like the only ISOs to really avoid are 640 and anything above 1600, unless you're in a low-light emergency situation (or if you have Neat Video, which I don't).