Noise Test: Panasonic GH2 with Driftwood's 176mb GOP1

Gabel

With the hack we all know how noise has become more apparent (because of the sharpness). I figured that it was time to test out and see the levels at different ISO values. So here are my results (since I'm not a Vimeo-pro member, there is some bad compression, but have the originals available for upload).

I shot 10 seconds (which now might have been too long) at each ISO value, with the body cap on. Settings were:
- Driftwood's "GOPStoppa" GOP1 176mb hack.
- White balance at 5600K
- Shutter at 50

I've upped the mids in After Effects quite a bit.

The major surprise was 320, which looks awful... Shot a second one, but had the same results! Very surprising...

Here's the video:

And the link to the original clips:
http://www.megaupload.com/?d=5M4N08F1

driftwood

Devastating! Seriously though, I'm looking at improving a few things this week, and noise is one of the issues.

LPowell

I find the GH2's ISO160 minimum to be one of the most annoying little things about the camera. I understand that it resulted from Panasonic's decision to correct the inaccurate ISO calibration of the GH1, whose ISO settings are about 1/2-stop underrated across its range (e.g. GH1 ISO100 = GH2 ISO140). However, the absence of ISO100 on the GH2 makes it awkward to match minimum-ISO exposure with other cameras. Often I'll resort to using ISO200 across the board, if only to simplify things.

Something is definitely suspicious about the noise jump from ISO250 to ISO320 in the video above. I notice a similar increase in noise from ISO1000 to ISO1250. Consumer cameras often use digital scaling to simulate the 1/3-stop increments between the standard full-stop settings. From this video it appears that some of the 2/3-stop settings may be digitally downscaled from the next higher full-stop setting, e.g ISO320 = downscaled ISO400 exposure.

Gabel

@LPowell: Exactly! To me it seems that the 160 based ISOs have gone from the best to the worst settings. The odd part is that I see little difference between 160 and 200 and 640 and 800, but 320 and 1250 looks awful, having noticeably more noise.

v10tdi

FWIW - here is a link to fairly comprehensive GH2 noise test done earlier this year - but from still perspective rather than video. Interesting observations that might be relevant in video context. Interesting horizontal lines in some of the pictures. :-)

http://www.laesieworks.com/digicom/panasonic-lumix-gh2-noise.html

stefanos

@Gabel
do you have a PAL or NTSC camera?
it's an important question for me to determine if the awful horizontal band that occurs in the lower 2/3 of the image has anything to do with PAL or NTSC . I have a PAL gh2 and the horizontal band is a big problem for me sometimes. It is the same banding that is apparent in your test from iso 1000 and above, clearly visible from 1250 onwards, I guess y'all can see it. It is a very annoying problem to say the least since any underexposed areas in your footage to cross that lower 2/3 band and you may have a serious problem. This issue is much worse and evident if the underexposed area is white or sth other than black. The only way of fixing generally this issue is by crushing the shadows in post which you normally don't want. needless to say of course that when this banding appears, it also negates the possibility of lifting shadows.

driftwood

@Gabel Out if interest, what film settings did you use for the test?

Gabel

@Stefanos: It is a PAL camera, but I have set the video out to NTSC, so as to get 60p in slo-mo.

@driftwood: Nostalgic (I've been a Smooth user, but I'm gradually changing to Nostalgic) with everything set to -2 except Saturation which was at 0.

proaudio4

The good news is our noise testing is now atleast more accurate over the stock lower quality firmware settings.

I understand what LPowell and others here are saying, I just can't understand what we are seeing here>
how a higher ISO setting 400 produce considerably less visible noise than 320. WTF, over

Also, is there really anyway code wise to really lower the noise levels at a given ISO.
I certainly can not think of a way short of post processing which degrades the image.

Ian_T

This might be a dumb question...but....I wonder what the noise would look like at some of the other hacked settings? Could some of this "tweaking" we are doing to the codec have an adverse effect on the noise?

Gabel

@proaudio4: To me, it reminds me of when using settings based around 125 on Canon's, which had noise levels where one could go from 250 to 320 and get a huge difference in noise.

@Ian_T: I did notice 320 being rather noisy at 66mbit, but I'm having a theory that the hack could be the cause of "changing" native ISO.

Ian_T

@Gabel was this a Panasonic len used? I've always had a suspicion (even before the hack) that Pana lenses contibute to the noise we see. But it was just a thought.

driftwood

@Gabel Finally, did you shoot in M (Manual mode)?

Gabel

@Ian_T: No lens used, I had the body cap on.

@driftwood: Of course! Never use anything else.

driftwood

@Gabel Sorry if this is bothersome. Could you run your test at stock now - same settings - as a video - for reference?

Gabel

@driftwood: I could make a new one tomorrow, sure.

Again, feel free to download the original MTS files too!

driftwood

Got them, thanks for doing this, Im analysing them with Elecard.

Isaac_B

I definitely find the noise higher with the 176 GOP1. Actually, I think the 132 Driftwood hack visually looks more rich and better to my eyes than the GOP1, it looks thicker and the color more accurate. Maybe it's due to bandwidth, I'm not sure.

proaudio4

If Ralph_B or someone else capable of recording HDMI out compare to 176 GOP1, this would help.

@Isaac_B
I agree, there's nothing wrong with how the hack is dealing with noise. Higher bandwidth (less compression) is why we see it.
Now the odd changes in noise levels through ISO change, that's another story...

Ralph_B

I already tested 240M GOP1 compared to HDMI.

http://www.personal-view.com/talks/discussion/comment/18186#Comment_18186

That should give you a good idea about what's going on.

nomad

@v10tdi
It's only strange that the guy in that test wrote: "ISO 320 might be the "sweet setting". More sensitive, minimal loss of image quality."
Do we have that much variation in the cameras?

proaudio4

I hear you nomad, I'm wondering why he wrote that since the example here actually shows ISO400 producing less noise than ISO320....strange. I need to check this out.

v10tdi

Perhaps it might be the difference between stills and video processing within camera - or variance in sensors?

Rerbreg

http://colorbyjorg.wordpress.com/2011/03/09/panasonic-gh2-iso-dial-bug-found-noise/

kanintesova

@Rerbreg:

I can confirm this bug. I just tested with driftwoods 176 GOP1 quantmebaby. I had the same exessive noise at ISO 320 when i tested ISO's sequentially, going from 160 upwards (up to 800).

I then jumped back down to 400 and then did a second at 320 and the exessive noise was "GONE".

I have a PAL camera in PAL-mode. Peculiar indeed. :/

kanintesova

Ok, another quick test: 160 ---> 320 results in exessive noise. Then going to 640 and then back down to 320 noise diappears at 320 :/

I shut the camera off (whilst at iso 320 with no noise), then power it back on: noise reappears. :(

Seems you have to do "the trick" everytime you power on your camera and want less noise at ISO 320. :(

LESSON LEARNED: Never go from ISO below 320 to ISO 320, always choose higher ISO and then back down to ISO 320 (if you need to use it).

Howdy, Stranger!

Categories

Tags in Topic

Top Posters