Jump to content

Close
Photo

Comparison between unhacked GH2 and Quantum V9b


  • Please log in to reply

#1
Axel

Posted 18 July 2012 - 04:36 PM

Axel

    Advanced Member

  • Members
  • PipPipPip
  • 1,013 posts
  • LocationGermany
As promised here:


Comments?
Either you care - or you don't

#2
kirk

Posted 18 July 2012 - 05:11 PM

kirk

    Advanced Member

  • Members
  • PipPipPip
  • 137 posts
I hacked my GH2 only for the PAL/NTSC switch, the ISO... but kept the original bitrate. Haven't regretted so far :) The images, at least if exposed carefully watching the histogram like a hawk, are of amazing quality as they are. From the various hack/no hack comparison videos I have downloaded the differences seem to be microscopic variations to an already awesome quality, not a quantum leap. So I chose not to worry about instability, expenses with faster and faster cards being filled at a worrying rate etc....

All in my humble, inexperienced opinion of course. Not critisizing or attacking anyone's approach, opinions and expertise ;)

#3
Kickass

Posted 18 July 2012 - 05:44 PM

Kickass

    Member

  • Members
  • PipPip
  • 23 posts
1. Looking at the original vimeo, not only is there compression in that file (which makes pixel peeping more difficult) but I can also see you have used an "easy" workflow, most likely through cs5, which - although fast does yield poor rendition of the channels, in particular if you check the red channel in high contrast areas you will see interlaced-like-artifacts..

Use either 5dtoRGB to convert to prores before rendering out or export your project to AE and render out from the AE engine (not completely sure you will get similar results as with 5dtoRGB but it should be better than the premiere engine). Also, it is better to render out to prores before converting to .mp4 - for retaining maximum information / rendition of the original that is. (Obviously, this workflow is not necessary for all kinds of work, but for something like this it makes sense to keep as much as possible of the original.

2. If someone don't find there is any use for high end short gop settings they might as well stick to factory, or some of the very good high end / intermediate long gop ones.

3. A long gop (group of pictures) encoding will find it easier to deal with moving "stills" rather than a moving camera, lots of motion e.t.c. and vice versa.

#4
Axel

Posted 18 July 2012 - 07:57 PM

Axel

    Advanced Member

  • Members
  • PipPipPip
  • 1,013 posts
  • LocationGermany


Not critisizing or attacking anyone's approach, opinions and expertise ;)

Beware!
To suspect any hype to be a myth of no substance and talk about it openly makes you an outcast, and who wants to be?

Use either 5dtoRGB to convert to prores before rendering out or export ...

This is not my video. It's from a Berlin guy whom I know through the german forum SlashCam. He is an amateur, as I am, and he works in a cinema, as I do. I know that he uses Vegas right now. 3 years ago he made Strutz with an HC9 and Magix, with rather elaborated compositing work and impressing camera work, check his vimeo channel. Little Big Berlin is his best known clip, a funny tiltshift short.

He certainly doesn't have first rate professional equipment or software, but he undoubtedly has "an eye". He loves the Lumix, but very early doubted that the hacks made a big difference.

He made the tests for his own peace of mind. There is no blessing in trying to convince others of an unpopular "personal view" or trying to "prove" something. Hear yourself: You should have used another method. Why? Because with Vegas or on vimeo the comparisons don't count? Why, it seems they count if they are demos of new patches ...

I use hacks myself, but I am happy with lower or medium bitrates like Vanilla, Unified or the like. Don't attack me for being unable to see any clear improvement with higher bitrates. Trust your own eyes.

Either you care - or you don't

#5
Germy1979

Posted 18 July 2012 - 09:01 PM

Germy1979

    Advanced Member

  • Members
  • PipPipPip
  • 577 posts
What's funny is, I don't even think i gave the original firmware a chance. I just plugged the card in, followed the instructions, and hacked away. I don't see any difference.

#6
Kickass

Posted 18 July 2012 - 11:24 PM

Kickass

    Member

  • Members
  • PipPip
  • 23 posts
I said; use whatever works for you, did I not?
However, for serious pixel peeping I think it makes sense to bring forth what is there to start with, and not heavily compressed h264?
The footage shows the tell-tale signs of an interlacing issue - which is good to know about - if working with h264 original footage in certain NLEs. At least if IQ is important to you as a user.

I am merely pointing towards what it is you might expect from certain settings and what you might expect from certain other settings, such as original firmware. Mind, it's very easy to cast a glance at the above, dismiss any adjustments to the original even if there are a variety of benefits that can be had if done so (f.i. there are settings which perform a bit better than factory settings that use LESS space and is even MORE reliable than factory settings..)
I encourage anyone who's interested in using the gh2 for serious work or just interested in getting as much as possible from their equipment to learn as much as they can about what people have found out, developing settings, shooting, coloring a.s.o.

Another important point, which I did not talk about in my previous post in this topic is what do you plan to use the footage for? Own personal stuff? Serious work? For the web? Possibly for a big screen? Already here it is very easy to say what might be worthwhile and what might not be..

P.S. I might be more critical than normally simply because of the "The Truth" title.. Such high horses are very rarely called for, on the contrary cause for alarm bells to ring.

#7
Axel

Posted 19 July 2012 - 04:22 AM

Axel

    Advanced Member

  • Members
  • PipPipPip
  • 1,013 posts
  • LocationGermany
I agree with you.

However, for serious pixel peeping I think it makes sense to bring forth what is there to start with, and not heavily compressed h264?


Agreed, however, serious pixel peeping is not what people do when they rave about the newest patch judging by a vimeo upload.

If I knew I had to deliver for cinema I surely had the highest bitrates, even if I didn't know for sure. Or skip compression altogether by using a Ninja (no audio then, but a pro who works for the cinema surely will record audio externally, no EX-tele also), because all this is about compression artifacts, not detail. Driftwood himself, when confronted with "The Truth", answered that every weaknesses of original FW come out in grading, especially in the shadows. True. But also the truth that it doesn't sound that humble on personal view.

P.S. I might be more critical than normally simply because of the "The Truth" title.. Such high horses are very rarely called for, on the contrary cause for alarm bells to ring.


I have to explain. It's high level irony. In the said german forum, these discussions went on forever. Tired of seeing yet another "proof" that the hacks worked, of course a highly compressed vimeo clip, someone started a thread named The Truth.

And the truth is, there is no proof. What you find out using your own judgement is what should finally count. 99% percent of what is said about the GH2 hacks is believed with eyes shut.
Either you care - or you don't

#8
kirk

Posted 19 July 2012 - 05:57 AM

kirk

    Advanced Member

  • Members
  • PipPipPip
  • 137 posts
Concerning intended use I might add that I own (and live in) a former small cinema. Projecting the Gh2 stuff I do (unhacked) in full HD onto my 16:9 screen (530 cms x300cms) the quality is stunning. Hack or no hack it is still a great camera! IMHO :rolleyes:

#9
Per Lichtman

Posted 19 July 2012 - 07:27 AM

Per Lichtman

    Advanced Member

  • Members
  • PipPipPip
  • 78 posts
Without having viewed the posted clip (it would be difficult to do so on this phone) I would like to offer the following.

1) Both the praise for and criticism of the hacked settings is often coming from a less than discerning evaluation, sometimes without a sense of how to do a controlled comparison. I do not feel it is fair to single either side out.

2) There are some circumstances where the hacked settings differentiate themselves from the factory settings, even without grading, and then there are others where they are almost visually identical.

3) To be able to see a difference between hacked and factory settings (without grading) it is generally necessary to shoot something that stresses the factory codec. If the factory codec settings are already able to handle the material well, then what is likely to be gained by hacking?

4) The situations in which the difference is most pronounced (especially between the newer high performance GOP1 settings and factory) are when the fine detail exceeds what the standard codec can keep up with (this generally means you will need a sharp lens, deep DOF and lots of detail in the scene) especially if the detail is ambiguous enough that it suffering from macroblocking in the original. Note that shallow DOF shots are less likely to show this difference. Shadows will often readily demonstrate the GOP differences and with long GOP encoding, a close examination can show results akin to watching them at a "lower frame rate", especially if you are shooting at higher ISOs. If the motion is sufficiently demanding in the scene to detoriorate into macroblocking in the factory settings, then the hacked settings can often improve results.

If you are not dealing with the cases described in 4) and are not grading your footage, then there is rarely a visual advantage to the hacked footage, at least when lighting and ISOs are kept in a flattering range. When lighting and ISOs get difficult, then some of the settings can really show their advantage. Check out the GH2 Hack Ref section at www.pasadenapulse.net and look at both Factory and Cm Night MTS clips to see how the difference is apparent without grading in such a worst case scenario.

On the other hand, there was very little difference before grading in another comparison I shot that day between the two settings with shallow DOF in broad daylight by grass in macro photography with low detail. It really is a case of whether the settings can keep up with what you are doing.

To most accurately judge differences without grading, my advice is to take two clips shot with identical camera settings under identical lighting conditions, from the same position, consecutively (or as close to that as you can get) and then use a 200 or 400 percent enlargement in your NLE to compare them. Somewhere on my Vimeo I did that between Factory and Sedna, unlabeled, and every experienced GH2 user that guessed got which one was which right (even though they flipped back and forth several times) right. Hopefully it was not one of the password protected ones but I cannot rember off hand. Also remember that high detail GOP1 settings tend to yield more detail in shadows and less defined areas, and that does not mean that they look less noisy (in fact often quite the opposite) if you do not plan on doing any denoisiing in post. But they can be cleaned up with denoising in post to look better and preserve more detail than factory settings.</br>One thing people often say (like Andrew) that is entirely accurate is that GOP1 improves the look of the grain in noisy images. I explained why above (to some extent) but I would like to emphasize that it has been shown to be rather consistently ture in my own testing.Note that outdoor testing is almost impossible as a way to compare different settings using 1 camera and natural light. The subtle changes in light due to the movement of the sun can easily change the results more than one might think and I have had to throw out many tests due to those variations. If you do want to do those tests, shoot setting A then setting B then Setting A then setting B again to try to discern the daylight variation vs the setting variation if possible.If you are using multiple cameras, you can test concurrently, but keep an eye out for any variables and match all settings and the lens on each camera. Also, ideally avoid any objects in the foreground close enough to cause a discrepenecy between the two cameras. So on top of a mountain, shooting into the distance might work, etc., but I have not been satisified by most such tests yet. Artificial controlled light makes it much easier to be sure what the difference is.So those are some of hard-won experiential notes on testing settings, learned during my time assisting settings developers since October.
Per Lichtman
Freelance Composer, Photographer, Director, Audio Engineer, Consultant, Instructor
http://www.pasadenapulse.net

#10
Kickass

Posted 19 July 2012 - 08:23 AM

Kickass

    Member

  • Members
  • PipPip
  • 23 posts
Thanks Per for a very pedagogic post.

Kirk, it seems like you misunderstood. Don't spread ignorance - it's not like you'd smear your expensive lens before a shoot; or add vimeo compression to your video before displaying it on your big screen.. (or maybe you do, what do I know).

Of course these settings pertain to compression. If anyone ever thought something else, then too bad.

#11
kirk

Posted 19 July 2012 - 12:51 PM

kirk

    Advanced Member

  • Members
  • PipPipPip
  • 137 posts
Kickass, I neither misunderstood nor care, and I certainly don't want to spread, convince or prove anything.... what would be the point? There can never be one truth in these matters....
But since the talk was about intended use I just wanted to say that although I don't aspire to produce anything for the cinema myself, the original firmware quality produces images that look very, very good on my cinema screen at least. Deep focus, wideangle, landscapey foliage stuff too. As I said, I'm sure the GH2 can be tweaked to produce even better results, but I haven't yet seen anything giving me a reason to do so. And as always, this is in my very humble opinion, and for my own use only.

#12
Axel

Posted 19 July 2012 - 03:06 PM

Axel

    Advanced Member

  • Members
  • PipPipPip
  • 1,013 posts
  • LocationGermany

3) To be able to see a difference between hacked and factory settings (without grading) it is generally necessary to shoot something that stresses the factory codec. If the factory codec settings are already able to handle the material well, then what is likely to be gained by hacking?


This is not congruent with what we experience in every day situations with the GH2. As many pointed out, there is a different treatment of noise with higher bitrates and/or smaller GOPs, in such a way as there is more noise, but more random, making the shadows look organic, making serious grading look better, but also ungraded lowkey-shots. The encoder has to make a frame without reference to the surrounding frames, and so noise will dither the blacks. Can we agree upon that? If so, then what has noise to do with stressing the codec?

4) The situations in which the difference is most pronounced (especially between the newer high performance GOP1 settings and factory) are when the fine detail exceeds what the standard codec can keep up with (this generally means you will need a sharp lens, deep DOF and lots of detail in the scene) especially if the detail is ambiguous enough that it suffering from macroblocking in the original.


Again: True as can be, but inconsistent with the phenomenons seen in real life. Though it is clear that - for example - the leaves of a cottonwood tree shivering in the sun can't be stored individually by 24 mbps @ 24p, and maybe a very high bitrate can show every inch of the individual leaves and this would become apparent by comparing the stills of the frames - what has this to do with having more details in motion? If what followed from your possibly true theory had any consequences for seeing or missing details when viewing and comparing the clips, then you should be able to link to a video (or thousands) where macroblocking takes away details in motion (sand, water, leaves, all the "ambiguous" detail you mentioned). It seems the encoder stresses the codec, but makes a good job.

To most accurately judge differences without grading, my advice is to take two clips shot with identical camera settings under identical lighting conditions, from the same position, consecutively (or as close to that as you can get) ...


There are many of those. I made quite a few myself. I couldn't manage to find unambigous differences. Because there are none in motion video. What I found true beyond any doubt is that the shadows look better with hacked firmware - very clear already at 44 mbps GOP 3, my first hack.

Different patches look different. If the newest one really looks better than the other is more a question of the accidental circumstances. I am not against trial & error, but what people praise as real progress is not based on any scientifically verifiable norm but rather fishing in troubled water. Literally.

EDIT: Not against hack. Exaggerating a bit for counterbalance.

Edited by Axel, 19 July 2012 - 03:23 PM.

Either you care - or you don't

#13
Per Lichtman

Posted 20 July 2012 - 12:06 AM

Per Lichtman

    Advanced Member

  • Members
  • PipPipPip
  • 78 posts
Axel, I respect your own testing and am simply sharing my own experiences from a greater volume of testing. Based on the posts Personal-View.com, no one that is not developing settings of their own has participated in more in the way of controlled comparisons than myself and the people I have enlisted to help with that. I say that as a matter of perception, not as a matter of fact: if people do not talk about what they are doing, then I do not know about it. :)

In regards to 3), as I alluded to later in the post, noise is an example of stressing the codec.The reason for this is that lossy codecs (whether MP3 in audio or h.264 in video, etc.) all have to differentiate between "real detail" and "noise". The combination of trying to allocate more of a low-bitrate to "the most important information" and a higher GOP, can often lead to noise being "simplified" to give more of the available bitrate to "the important stuff". I am over-simplifying greatly, but the general principle applies.

I am not saying that there is "no difference" at all, I am saying that the difference is most visually obvious in the circumstances I mentioned. Increase the motion, the ambiguous detail, the proportion of shadows, etc. in the scene and the difference becomes more obvious.

Noise is a form of detail, and the codec settings treat it differently. Generally, you lose other detail when you try to simplify it as well. I agree with you and nothing you said about that contradicted what I said earlier.

As regards 4), it is accurate based on my testing. Think back to shooting 60P on the GH1, for instance: if you go in telephoto on a high motion shot with no hack, you will generally get TONS of macro-blocking. I have some humming bird shots that illustrate this beautifully. You can see the same thing on the GH1 with unhacked FHD, etc. Some of these same types of shots that suffered from macro-blocking when tested with factory settings, do not show macro-blocking when tested with some of the hacked settings. For obvious reasons, this is somewhat difficult to showcase: you need to have a shot that is sufficiently demanding to stress the codec to the point of obvious macro-blocking AND is completely repeatable.

Most shots that have sufficient motion to cause the phenomenon are not sufficiently repeatable to demonstrate the phenomenon to a wider audience, so shooters are often left to do their best to "guesstimate" the difference based on their own testing.

If someone wants to demonstrate the phenomenon, try locking your GH2 down to a ceramic wheel (or similar) and spinning it at the same speed, twice (once with each setting) in a room with controlled lighting, without changing anything. Shoot factory settings first and then up the speed with them until the macro-blocking becomes obvious. Once again, deep DOF can be helpful and you may want to experiment with shutter angles narrower than 180 degrees.

You will have jello effect showing up pretty bad, but you will definitely notice a decrease in macro-blocking.

Regarding different hacked settings, here are a few from the same torture test. (Ugly, but it gets the point across).

http://perlichtman.c...hack-reference/

Here are a few that really show it, linked from that page directly.

Factory 1.11
http://www.perlichtm...ts/Stock1.1.MTS

EOSHD Vanilla 44
http://www.perlichtm..._Vanilla_44.MTS

EOSHD Unified 88
http://www.perlichtm...ts/EOSHD 88.MTS

Driftwood Sedna AQ1
http://www.perlichtm...Sedna_AQ1_A.MTS

Driftwood Canis Majoris Night
http://www.perlichtm...ts/CM_Night.MTS

Or if you do not want the video clips, here is JPEG (shot as a still) and two PNG stills from Driftwood Sedna AQ1 A and CM Night.
http://personal-view...9#Comment_59989

You can clearly see, if you reference the control JPEG first, that details like the hanging wire in the shadow in the left, etc. are much better rendered in CM Night than in Sedna, even though both are creating rather large files.
  • nahua likes this
Per Lichtman
Freelance Composer, Photographer, Director, Audio Engineer, Consultant, Instructor
http://www.pasadenapulse.net

#14
Per Lichtman

Posted 20 July 2012 - 12:08 AM

Per Lichtman

    Advanced Member

  • Members
  • PipPipPip
  • 78 posts
Also, here is the Sedna vs Factory comparison I mentioned a while back.



Password is Driftwood1

Like you said, the difference is pretty easy to see in noise and shadows.
Per Lichtman
Freelance Composer, Photographer, Director, Audio Engineer, Consultant, Instructor
http://www.pasadenapulse.net

#15
Axel

Posted 20 July 2012 - 04:30 AM

Axel

    Advanced Member

  • Members
  • PipPipPip
  • 1,013 posts
  • LocationGermany
Very convincing arguments. I will take good time to look at the clips. Thank you very much for answering so patiently.
Either you care - or you don't

#16
Axel

Posted 20 July 2012 - 07:42 PM

Axel

    Advanced Member

  • Members
  • PipPipPip
  • 1,013 posts
  • LocationGermany
Dear Per,
I downloaded and watched all the files you linked above. I would like to be able to talk so fast and yet so well comprehensible ;)

I compared the clips running simultaneously as loops, spread over three displays (of course overlapping), and really can't see clearly which one is better or not.

Much clearer is your "iso 10,000 i 1280 50p debate DivX.avi" which I downloaded from your site (which I find really good). Proves the whole shadow-hack-affair.

My own tests have a different approach. I like to see real life action, and luckily we have three cats who chase each other all night. I follow them through our poorly lit flat with the GH2 hand held (see grip in my avatar pic). This gets as close to stressing the codec as I would ever expect to need. What is important to me is motion rendering, if this is the right term. The same rules apply, but not all patches work equally well for motion.

I hope this thread encourages people to test more themselves. The GH2 deserves owners who care.
Either you care - or you don't




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users