24p Tests: CM Night, Cake 2.3, Sanity 5, FM 2.02, Stock GH2, ALLI, and SGOP4.

Ashun

While I was working on ALLI and Simple GOP4, I was constantly looking at other patches in order to better understand the various PTool parameters. In the course of that work, I thought I'd do some comparison testing of the 24H modes of Canis Majoris Night, Cake 2.3, Sanity 5, Flow Motion 2.02, the stock GH2 firmware, and my two incomplete settings. I would have tested one of bkmcwd's patches, but he releases too many :)

Each was tested with the 14-42 mm kit lens at 42mm, f/7.1, ISO 320, focused on a CRT screen with a rapidly changing color pattern. All were stable on a SanDisk 45MB/s card. I've uploaded a five second clip, Stream Parser Snapshot Report, I, P, and B screen grabs (if applicable) and a StreamEye quantizer frame.

I'll start with the highest bit rate first. As you'll see, bit rate is hugely important.

ALLI 1.0 (150,113,176 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/02_ALLI.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/02-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/02-DC%20I%20Q.png

Canis Majoris Night (148,454,022 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/03_CMNight.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/03-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/03-DC%20I%20Q.png

Simple GOP4 (132,283,819 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/01_SimpleGOP4.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/01-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/01-DC%20I%20Q.png
P frame: https://dl.dropbox.com/u/23922406/GH2Tests/01-DC%20P.png
P frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/01-DC%20P%20Q.png

Flow Motion 2.02 (120,587,686 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/05_FM2.02.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/05-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/05-DC%20I%20Q.png
B frame: https://dl.dropbox.com/u/23922406/GH2Tests/05-DC%20B.png
B frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/05-DC%20B%20Q.png

Cake 2.3 (60,094,762 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/07_Cake2.3.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/07-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/07-DC%20I%20Q.png
P frame: https://dl.dropbox.com/u/23922406/GH2Tests/07-DC%20P.png
P frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/07-DC%20P%20Q.png

Sanity 5 (36,991,314 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/04_Sanity5.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/04-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/04-DC%20I%20Q.png
P frame: https://dl.dropbox.com/u/23922406/GH2Tests/04-DC%20P.png
P frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/04-DC%20P%20Q.png
B frame: https://dl.dropbox.com/u/23922406/GH2Tests/04-DC%20B.png
B frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/04-DC%20B%20Q.png

Stock GH2 (23,096,427 bps)

https://dl.dropbox.com/u/23922406/GH2Tests/06_GH2.MTS
I frame: https://dl.dropbox.com/u/23922406/GH2Tests/06-DC%20I.png
I frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/06-DC%20I%20Q.png
P frame: https://dl.dropbox.com/u/23922406/GH2Tests/06-DC%20P.png
P frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/06-DC%20P%20Q.png
B frame: https://dl.dropbox.com/u/23922406/GH2Tests/06-DC%20B.png
B frame Q: https://dl.dropbox.com/u/23922406/GH2Tests/06-DC%20B%20Q.png

OK. Information overload? Keep in mind that you will never (or should never!) film anything like this. Some cursory analysis:

Canis Majoris Night is awesome, better than my ALLI. Flow Motion 2.02 is also really good, better than my Simple GOP4 while using less bandwidth. The rest have considerably lower bit rates, so expect a drop in quality. Cake's 60Mbps seems to be about as low as one can go without the image turning to crap. Sanity has nice I frames, but it's starved for bandwidth, so the P and B frames are bad. And the stock GH2 firmware completely falls apart.

Of course there are things other than 24H that you'll need to consider before choosing a patch, but in general, use the highest bit rate you can get away with given your shooting constraints. I've used Sanity to shoot at 45 minute speech, and it looked fantastic. If you're shooting a psychedelic nighttime rave, maybe use something else.

Take a look at the .MTS files. They're all under 100MB.

feha

@Ashun , thank you for useful tests.

LPowell

The Stock GH2 sample file shows what happens when the unhacked firmware collapses into Fallback Mode, surreptitiously degrading its image quality with coarsely encoded macroblocks. In the Flow Motion v2 thread, I documented this same Fallback Mode syndrome when shooting an ordinary forest stream. In my view, the unhacked GH2's tendency to silently resort to Fallback Mode makes it completely unreliable for anything more than consumer purposes.

driftwood

Fallback mode has been known since early days. Refer to cbrandin.

SuperSet

Many thanks for your extremely useful post.

LPowell

@driftwood Yes, I believe cbrandin was the first to conclusively identify Fallback Mode and detect it with his invaluable Stream Parser tool. However, this was in the context of GH2 patch development, where increasing the bitrate beyond the camera's unhacked limitations would often provoke it into Fallback Mode.

What I believe is not well-known is that the unhacked firmware is also prone to collapse into Fallback Mode, and can do so without warning when shooting ordinary scenery. As far as I'm aware, my post in the Flow Motion v2 thread was the first publicly documented evidence of this GH2 failure mode, and it's now been confirmed by Ashun's test footage in this thread.

driftwood

This was known in the early days, chris has mentioned with me over our own pms but quite possibly not brought it up in public. Chris is thee font of information. Stock settings are also prone to buffer errors. Stock settings are very erratic. Things get forgotten and very few people read the past threads.

A lot of stuff goes undocumented unfortunately as it sails past most peoples heads or testers simply move on without worrying about it. I like the way you document things but because you do, it is sometimes misconstrued that you are the only one who knows whats going on. :-)

LPowell

@driftwood Both Ashun and I have put in the work to prepare and upload documented test footage that provide hard evidence of a variety of things that people have claimed about hacked and unhacked GH2 image quality. I'd welcome any research of your own you may be able contribute, and I'd also ask you to show some respect for our work as well.

driftwood

Now youre being silly. Since when did 'Our' become one and the same? I have the sincerest respect for Ashun's results, however, what really pisses me off is the way that the way you write is I, I and I... you have learned a lot Lee, as well as give a lot. We all 'give' a lot, we all learn a lot. The earlier point was only refering to the fallback mode which Chris referenced ages ago it is not new discovery.

mozes

i agree with @driftwood, i did read it in the pro section off the forum, and its dated a real long time back.
But i am very happy with the explaining from @LPowell because off that i understand all a little better now.

Its the hardest thing on PV, finding the right info, and most when we trying to make good settings, a lot off stuff is already done before.

LPowell

@driftwood I see your personal attack on me as nothing more than an attempt to distract attention from the actual documented footage that both Ashun and have uploaded. Do you have anything tangible to contribute here, such as verifiable test reports, rather than hijacking the thread with how "pissed off" you are about my writing style?

driftwood

No its NOT a personal attack. Settle down, keep calm.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters