Thanks for the test. I checked it out. I see no difference between your stock and hacked clips.

I got my GH2 a couple weeks ago and did similar static tests for myself in low light and concluded that the current hacked firmware out there is not worth the extra file size if I'm not going to notice any difference, so I stick to stock for now. I use the Leica DG 25mm F 1.8.

If anybody has a clear example of how hacked clearly outperforms stock, I'd love to see it. I've only seen one example that showed that high ISO noise was slightly less shitty on a hacked clip, but I wasn't very impressed. People say that you notice the difference in post when color grading, but I'd still like to see a video example of the color grading process being made easier.

Edit: After shooting a fight scene, I now see where the macro blocking occurs and gets ugly when pausing the playback

Bit depth and motion. Stock motion (GOP 12) looks odd and video-like the AF-100, IMO. Also, go shoot some scenes with allot of underexposed areas and you'll where the patches really work their magic. Plus, you can't even touch the colors in post with stock settings without the image falling apart. It was worse for grading than a 5D/7D codec. It was HDV territory.

@kavadni I'm viewing the MTS files on an iMac 27" 2560x1440, using VLC with a normal size window.

What are you viewing it on to see the difference?

@TGDude .. I view mostly on 2 x Samsung LCDs .. nothing special about them. I also look on my laptop, and my television.

I think you all are wasting your time explaining this one. But good luck. It's going to turn into another one of those threads quickly.

All I'm going to say is @questech...you need to spend a lot of quality time with your camera and a lot of time reading.

Hello All,

Thanks for all of the feedback!

@Ian_T, I'm a director, I don't have to point a camera for a living, so I'm pretty much done reading about this topic. I simply ran some tests, and they speak for themselves, that's all.

Again, these test were only for the scenes that you viewed (if you took the time to watch them) and if you read my posts and the ReadsMe file included in the zip, you know that I mentioned that I don't know if a high action scene would show the hacked firmware to be better or not.

But from my limited tests, I came to the conclusion that for low light dramatic scenes, like the ones in my tests; a) less noise with the stock firmware, and b) undetectable improvement with the hacked firmware in general. At least I couldn't see it and I looked at the footage at every magnification imaginable.

Now I will state that I did notice that in some of the test that others have conducted, in some cases the noise looks better, more like film grain at higher ISOs.

My tests were done at ISO 400 and I usually shoot under that because I light my scenes well and don't have to push the camera.

But my test mainly concerned noise in ETC mode and I believe the tests are absolutely conclusive... The unhacked firmware produces less noise than the hacked firmware.

I don't see what the big deal is, none of us invented the GH2, it's not like someone is calling your baby ugly :-)

The information shared should be unemotional and factual. We are seeking the truth, not brownie points... right?

My mind is open, but I guess I will have to run some action scenes with my camera eventually.

Thanks Again !!

@questech

Is this footage all supposed to be 24p in a 60i container or were your shooting 60i? Either way, the stepping you highlight in the png seems like it may be a workflow thing. 24p needs pulldown.

Hey Guys,

Thanks again for the replies!

First off, Ian_T, I apologize if you think that was meant for you, I was speaking in generalities, so other than my comment about your suggestion that I read more :) I was just trying to make a point.

Again My tests were limited to this static scene, and I am not disputing that there may be some benefits to hacking the firmware, I just haven't seen had the need to yet from everything that I have read vs everything that I have watched.

I use high end cameras in my business, so this is my first try at this type of camera. I really just purchased it to do YouTube videos using a green screen. I was just not expecting the noise to be this bad because the reviews made it sound really good. After seeing the noise (maybe I'm admittedly a little picky) I was prompted to start this quest.

I mistakenly thought that a hack might fix what Panasonic neglected, but I didn't know enough at the time i.e. that a hack wasn't going to change the what the sensor was doing. I originally thought that the hacks were changing the internal programing that affected the DSP and could modify the way the camera dealt with the information from the sensor.

I dig the camera and will eventually get the free time to move it around a bit, maybe at a tennis match so that I can experience the difference between the hacked and unhacked firmware.

So anyway these test results were strictly to see if any of the hacks would perform better than the stock firmware in ETC mode. I have a lens that I really like that requires ETC mode.

What frustrates me though is that I have watched a lot of video of hacked firmware, and upon close inspection, I see a lot of blotching that looks like little blurry swarms and other artifacts. I guess my expectations exceeded what this little camera could do, especially when many tout how good it is when hacked.

But don't get me wrong, I like the camera and will just have to use it within it's limitations.

Thanks Again !!!

@TGDude

The real answer is that it is easy to show clips where you can’t tell any difference at all and it is just as easy to show clips where the differences are very apparent. It all just depends on what you are shooting and what you hope to get out of it.

The hack definitely has its place. You just have to know what each different group of settings is intended for. For the most part just doubling the bit rate will produce results that the average user could not possibly distinguish from the highest bit rate hacks. That is why we have the No Adverse Affects settings.

If you have to ask which settings you should use without telling us anything about what you want to improve then the No adverse Affects settings will give you the quality you seek with all of the reliability of the stock firmware.

However, if you know that you are going to pull the footage to extreme levels or if you are going to be filming a scene where the entire frame is in motion(Like running water) then the settings that Llpowell and driftwood have produced will give you the results you desire.

The only thing that gets me is when someone uses an extreme setting and then complains that it had write errors for them and they can’t see any difference when they post it on youtube. If you really know that you need that level of bit rate then you need to be ready to Pony up for the recommended memory cards as well. And you also better be doing more with it than just sharing it on youtube or you are probably just wasting your money.

So many people just blindly go into this asking “What settings produce the best quality”. Instead you should be asking yourself what settings will suite my needs the best.

I Second That!

Different settings affect different aspects of image quality. If you only test static scenes you are effectively ignoring one of the most important aspects of video – that is, motion. Many settings – particularly low GOP ones – address the fidelity of motion and aren’t intended to improve the fidelity of static subjects. Other settings (i.e. long GOP ones) are intended to maximize the quality of static subjects. In no particular order, here are some of my findings concerning settings:

The factory settings for 24p1080 produce a maximum I frame size of about 900K Bytes. The most I’ve seen produced by any setting comes to about 1180K Bytes. That’s an increase of a little over 30% - a modest, but significant increase. I frames effectively set the maximum IQ of a clip.

When frame size is increased everything is rendered with better fidelity – including noise.

The “look” of video can be changed by modifying quantization table entries. Basically, one can trade off how much fine detail (little, busy parts of the image) is rendered for smoothness, and vice versa.

Motion is rendered perfectly with GOP 1 settings – the tradeoff is that the maximum frame size is necessarily smaller, which results in lower IQ for static subjects.

If motion rendering isn’t important to you (e.g. you are shooting relatively static or slow moving subjects) there is little to be gained by using a setting that is over 32Mb. If you take the ratio of the maximum frame size with factory settings (900KB) to the maximum possible frame size (1180KB) you get 1:1.31. If you apply that ratio to the factory bitrate (24Mb) you get 24 * 1.31 = 31.47. That’s pretty close to 32Mb.

It’s very hard to determine the fidelity of motion rendering by examining individual frames. Even if frames look fine, that doesn’t mean they are an accurate representation of the actual subject at that particular moment in time. Sometimes individual frames can look fine, but when the clip is examined in motion things just don’t look quite right in ways that are hard to quantify (I don’t mean the obvious stuff – like stuttering).

24Mb is a fairly respectable bitrate for 24p/1080 AVC video – achieving a large percentage of potential quality. All these settings are arguably addressing potential improvements within the realm of diminishing returns – you’ll not see dramatic improvement, rather you’ll see incremental improvements. Also, when final output is rendered for, say Vimeo, these improvements will probably be almost invisible. Now, if you are doing a lot of grading, doing slow motion, etc… - anything that demands maximum image quality, the improvements offered by good settings will become much more apparent.

Also - what Lee Powell said...

Chris

Quite possibly THE most informative page I've seen ANYWHERE about the GH2 Hack, cheers guys !

I think one thing some of us on this forum need to keep in mind is that many of us have had months to get to know exactly what the different hack settings will and will not do, even as new settings pop-up at a very rapid rate.

@questech's original belief that the hack would result in increased noise reduction isn't an unreasonable one. I think a lot of people may mistakenly come in with that preconception until they start reading up and testing it. His post in this forum is a continuation and he went back and re-tested based on our discussions on the former, so kudos to him for making the effort.

Let's be clear on this: most hacked settings are NOT designed to increase noise reduction, they are designed to preserve more detail. That means that if you do more noise reduction in post, you should have a much easier time with high noise footage than before. If you do not do noise reduction in post, you may find that you have both more detail and more noise in your footage, depending on the scene shot and the setting used.

As I mentioned in the other thread, the key to getting the most out of the hacks tends to be to find a deficiency in the stock settings and look for a setting to address that.

Low-light noise isn't rendering with the motion or detail you want? CM Night is one setting that can really help that. As of my testing at the time of its release, I found it unmatched by any other setting in that area.

The Group Of Pictures (GOP) setting is spaced too wide for you? Use a GOP1 setting to make sure that each frame is rendered individually.

Image quality drastically decreases when there is higher motion in the scene? Most of the hacked settings address this to varying degrees and it is a good place to see the difference.

Image quality seems to be getting lost to macroblocking in a scene with high detail? Use a setting optimized for greater detail, like Sedna, GOLGOP, Flowmotion, etc.

Image quality seems to break apart during grading? Try a hacked setting with a higher bitrate.

These are some of the sorts of situations where the difference is easiest to notice. If you aren't pushing the stock codec (for instance if the scene has low noise, a not very demanding dynamic range, low detail and no motion) you are unlikely to notice any difference, especially if you are not grading.

I would be suggesting specific settings more but 1) I have more experience with Driftwood and Ralph_B settings than the others and don't want to be too unfair and 2) there have been a LOT of new settings or updated settings released recently. I would like to stress that Driftwood's settings are not the only high quality settings available - I just really like the look of GOP1 and he makes more of those settings than the other authors.

Okay, so that's my 2 cents. Most of it is a repeat of what I said before, but hopefully it is still helpful to some people. :)