Driftwood ClusterX series 2:│moon│ЅріzZ│nebula│drewnet│Slipstream │Redshift

spacewig

@curtismack What I discovered was there are no noise-free blacks with the GH2. You can try this any time and see for yourself. Take an incident reading in a well lit room and shoot a part of it making sure you have some very dark-colored objects in the frame. When you play back the footage you will see the dark objects permeated with noise. I think the only way to minimize noise is to shoot with the lowest possible ISO, in this case 200 as I found the noise difference between 160/200 ISO negligible and with 160 you always risk getting the noise bug. Furthermore, when I calibrated my light meter I also discovered that ISO160 has a tad less ( I think it was .2 EV) dynamic range than 200.

4CardsMan

@Driftwood - Thank you for the comprehensive explanation.

@spacewig But if the light is soft with no intrinsic contrast, then you can add enough exposure to bring the blacks to grey, then grade them back down. Here is a frame grab with waveform shot under cloudy skies adding one full stop exposure to the camera's recommended setting according to the histogram. The first version is ungraded. Its whites just barely clip in the waveform. The second is the same frame after a quick grade in FCPX. The highlights don't blow out because the scene itself has no more than 3.2 stops of DR, and the lighting adds none. Buddha's black fur shows up as milky black in the ungraded version, and about right in the graded version.

WhiteRabbit

@CurtisMack, I aim to expose to the right (highlights), and not overexpose. While you may reduce the noise in the shadow areas in your example, you may have to live with blown out hightlights and lack of detail in other areas, such as the white fur near Buddha's collar, the leaf near Buddha's tail and the white part of the two tennis balls behind. That is a difficult shot, choices have to be made or compenstion with lighting control/diffusion/reflector. I don't know the answer, opinions may vary. Buddha looks like a good loyal dog. I miss living in rural parts.

4CardsMan

@WhiteRabbit To paraphrase Ansel, it's not overexposure if you do it on purpose when you know the outcome. I haven't done the actual exposure test to determine exactly how much extra exposure you can give in soft light and still keep all the highlights. In this shot, FCPX's waveform showed mild clipping. The GH2's histogram indicated correct exposure at f/2.5 (ISO 400), and I exposed the shot at f/1.8, one full stop more. So, it appears that in soft light 2/3 of a stop more exposure than indicated would bring in those highlights. However, in more severe light, it would be necessary to lower the lighting ratio, to a maximum of (I'm guessing at this) of 1:4 to bring Buddha's black fur into the grey area. When I acquire some daylight balanced soft lighting, I plan to test this and come up with a firm exposure/fill method which is simple to apply.

leothelion

Sorry to not contribute to the really great discussion going on but I did want to add another sample footage from a completed project.

Again, moon trial 4, 14-140 stock & fd 50mm 1.4 All iso under 800 No grading.

Thanks for the hack, settings and discussion.

WhiteRabbit

@CurtisMack, thanks for your reply and advice. It is an interesting balance to juggle.

Mirrorkisser

Overexposure is sometimes dangerous, too. you can loose a lot of DR and colorinformation in the overexposed areas. Slight noise in correctly exposed blacks is the smaller trade-off in my opinion, but everyone has to choose for himself.

@driftwood: one thing i am interested in: DR is probably rather a sensor or codec issue, but is it possible to gain DR with settings like Intravenus 444, that change the color information on the channels? Eventhough its more of a 330 than a 444 it could be different DR wise than 420?

thepalalias

@mirrorkisser @whiterabbit @spacewig @curtismack It is partly a matter of personal taste and what your primary goals are, of course.

For me, there is no "one-size fits all". When I am more concerned about noise, I go with the @curtismack approach (and yes, the approach still applies for the GH2 as much as it did with film but I would not claim expertise about the range). When I am more concerned about the highlights and light rendering, I am usually willing to except a bit of noise in exchange for it (more like what @mirrorkisser describes).

But in the end, it depends hugely on the subject. The number one priority in any shot that contains a human in the shot will generally be that human being's exposure. The rest kind of has to follow whatever I have to do in order to make their exposure work - maybe the shadows will get crushed in post to reduce noise, or maybe the highlights will get blown but I look at which one I have to do in order to get the skin texture I want and then I do it.

Now, if I am shooting something without a human subject, I can look at what is most interesting to me in the scene and set the exposure to emphasize that. Is there interesting geometry in the shadows that I want to bring out? Is there a softness to the lighting that I want to make sure and capture? Are their clouds or colors in the sky that I want to make sure and emphasize and preserve?

It is all about choices - often you can't get everything using the camera settings alone. You can use neutral density filters, contrast filters, etc. to make it easier or you can mess with the lighting but whether you are aiming left or right on your histogram, what @curtismack says about aiming for a particular dynamic range is true. If you can get all the information into a particular range, then the camera has a fighting chance of catching it all, otherwise you'll have to choose what you want most.

But getting back to @Driftwood settings, specifically. If you are going to preserve detail in the shadows, many of the @Driftwood intra settings can be counted on to ensure detail in the noise patterns. I like Moon 5 for high detail scenes and Sedna or one of the IntraVenus versions for lower detail scenes where the more aggressive use of bitrate makes sense. But I shoot both kinds of material with both settings - I don't normally switch around on-site while shooting unless I'm testing. :)

4CardsMan

I tried Nebula today on breaking surf with very favorable results. It ran 12 minutes on a Sandisk 16GB 45mb/s card, and required no trickery such as a power cycle or recording a short clip immediately after the long clip. It plays back in camera.

Please disregard the previous paragraph.. The results seemed so unbelievable that I reran PTool, created a new bin file and updated the firmware. Nebula does not span on the slower card and does not play back in camera. Additionally, FCPX doesn't see it without help. Apparently, the original bin file that I thought contained Nebula did not. My apologies for the error.

WhiteRabbit

@thepalalias, thanks.

omnidecay

Using moon 5, test:

(thanks @bkmcwd). I used SanDisk Extreme 30MB/s and I was able to record for 2 1min clips. Average bitrate around 148000kb/s (which is 18.6MB/s).

Why do people recommend getting the uber expensive SanDisk 95MB/s? The GH2 is not able produce enough bitrate to come even close to those speeds. Even my 30MB/s still had 1/3rd head room left.

I was playing around with moon5 which is awesome, thanks @driftwood!

vicharris

@omnidecay The larger hacks will not span unless you have the better cards.

maddog15

Looking for reliability as much as quality. Slipstream #1 Testing now. First impressions - excellent quality. Since the description mentions "slower sd cards" I'm testing with one of my first SD cards instead of my SDXC 64GB - 95MB/s card. While using SDHC 32GB Ultra 30MB/s I recorded for around 16 minutes straight. No crash or freeze or card write error. None. File is about 5.5 Gigs - recorded at 71.24 Mbit/s and playback in camera works fine! Incredible. I only get about an hour of record time with this card but still....I'm speechless it works so well. I'll post more feedback after tests tomorrow.

Thank you Nick. This is awesome. I really do appreciate your time, effort and patience.

Updated Using Slipstream #1 hack Let camera roll for 60 minutes using SDHC 32GB Ultra 30MB/s card. Playback didn't work (Forgot to do a quick recording before turning camera off) BUT all files were on the card. Clipwrap detected the multiple files and wrapped them back together.

Update 2 Just got back with around 30GB of footage. Used SDXC 64GB - 95MB/s card to be safe. Tested HBR, 24P high and 720 SH. Had no write problems the whole day. No glitches, lock ups or freezes. So far Slipstream #1 appears to be stable as a rock. SD card hungry ...but solid.

Here's a vid of the test shots I did today. Shot in 24p High. I'm very happy with the results. Still have a lot to learn but I'm really liking Slipstream. Any advise on getting better YouTube quality is warmly welcomed. And any other advise.... I have tough skin. Love learning.

omnidecay

@wicharris gotcha...who takes a 4min+ shot lol?

I did test again and no spanning on my card.

thepalalias

@omnidecay The situation is a lot more complicated than that. :) First of all, in the longer than 4 minutes category we have documentaries, event photography and several commercially released films ( for instance the tour-de-force single-take Steadicam shot in the Tony Jaa movie The Protector).

Second, you are not just looking at the average video bitrate but also the peak bitrate. In the years I have spent testing these settings with different cards (well more like a year and a half :) the recurring theme has been "settings that do not work on cards that should work on paper". So peak plays a role, in addition to spanning. You want the card to be sufficiently high performance and compatible that it does not run the risk of being a limiting factor.

Third, there are several modes that place additional stress on the card/camera combination. Maybe they handle 24H but can they handle VMM 80% mode in 24H, for example. People generally buy the 64GB SanDisk 95MB/s to get access to the best a given setting can offer, which sometimes means additional modes functioning properly.

Now, of course the demands of a given mode vary by the scene content as well, and sometimes less detailed scenes will run on slower cards, etc. But when you want the best stability, spanning, options, etc. there is a card available that has consistently outperformed others on the hacked GH2 in the situations where slower cards could not keep up. That is pretty much all that people are saying: not everyone needs to spend that kind of money. :)

Azo

@thepalalias

Thank you for sharing your wealth of knowledge with us. I really appreciate your in depth easy to understand explanations. Most of the technical things that get discussed on here go completely over my head :(. I don't know anything about I frames B frames P frames etc, yet whenever you explain things you do in a manner which is very easy to understand.

Best Regards

omnidecay

@thepalalias I actually have a 16gb 95MB/s just to run hack tests on. A friend of mine asked me to film their wedding and there is no way I could afford to do that with the 95MB/s cards. Naturally, since I am a good guy and not charging them these high end SandDisks are just too much. My 4min+ comment was supposed to be a joke :-). However, I do miss tape in that regard.

Do peak bitrates shoot up by 33%? I understand this would make the camera puke but I tested with the above video which is pretty busy. What I found interesting though is that there are no hacks out there that touch 30mb/s. So speed as it seems is not as important as some other unknown factor (to me). Sure, there are rogue bitrates but it seems silly to pay over $2 a gig for 95mb/s just to take into account the .05% of crazy frames. Is there no way to put a max upper limit?

There seems to be a misconception that slower class 10 cards can't keep up when in reality they can. If your card has write speeds faster then the applied hack then there should be no reason why that wouldn't work. There is something else causing the cards to not span or cause the camera to freeze. Again, it is probably known to someone but I cant find the answer. Just like in the video I used to test above. I saw an average of 18.5 MB/s which my SanDisk 30MB/s laughed at. I let that run using Moon5 for the full 4 min (a few times) and it was cake (minus the not spanning).

With that though, I am trying to find a hack that will currently span with my 30mb/s since I am going to be rolling strait for 12 hours that day. Probably working with footage mostly in 24p and 60p for slowmotion.

To reiterate, I love everyone who works on these hacks :-).

thepalalias

@omnnidecay When I need to shoot long events, I normally make image quality a secondary consideration as compared to stability when spanning. I normally use Flowmotion 2.02 if I am using slower cards (though there are several Driftwood ones that work well with the SanDisk 64GB 95MB/s). Somesettings really are designed to work well with slower cards and the descriptions often specify how to get stable spanning with them (which sometimes means shooting in 24L instead of 24H). If you cannot afford 24H spanning capable cards for an event, 24L really is fine. :)

In regards to spanning, I would suggest doing testing with less expensive 64GB SDXC cards. I have done several tests where a slower 64Gb card had better chances of spanning than a faster card of lower capacity.

Just remember, getting every last ounce of performance or bitrate out of a setting and getting reliable spanning performance at the same time are mutually exclusive. Pick the one that is most important for a given shoot and then do the best you can on the secondary priority - you will only frustrate yourself if you try to handle both equally. :)

@Azo Thanks! :)

@omnidecay By the way, I agree about the "x-factors" involved in why these do and do not work with a given card. That is part of the reason why the authors like @driftwood, @lpowell, @bkmwcd, @ralph_b, @mpgxsvcd (and so many others) have to spend so much time developing and testing the settings and why it makes more sense for most of us to pick the version of their setting tailored to our use than to try to modify it ourselves. But as mentioned before, the 64GB SdXC card mentioned earlier seems to use a different controller or something that makes a difference in spanning performance. I do not remember how great a gap their tends to be between peak and avg. bitrates in usual footage but there are ways to look at it using MediaInfo or StreamParser, for instance.

But like I was saying, I have had Flowmotion 2.02 with 24L as my fallback if I need the lowest chance of spanning errors I can get. I look forward to finding the Driftwood setting that is closest in spanning performance, soon. I just need to do more testing. :)

dado023

i have an idea to propose. Wouldn't it be great if there was a FAQ or wiki page with all active hacks cataloged, in a table or something, with data like "tested: YES/NO", or "Instantaneous Death Screen" test passed "ok" ...etc

thepalalias

@dado023 I've done testing in this area and posted Google Docs of my tests before but I don't like dealing with Wiki mark-up. It drives me nuts. So I'm happy to provide the data to someone else if they are willing to handle the Wiki mark-up.

dado023

@thepalalias why not share that particular google doc, and allow editing, so you would be free of maintaining it. That way other user would be able to input their feedback into it, right?

Zaven13

@omnidecay. I agree with @thepalalias. If you need longer recording times with lower cards like SDHC 30Gb/s, just go with 24L, FH and H modes of the top settings. Even at those levels you still get three times the bitrates of the original stock setting that comes with GH2.

driftwood

People should try Spizz, Nebula and DREWnet out - on L settings if they want long record times. I still havent had a span report from people on the L settings.

ALight

Shot on Moon Trial 5. Lens: Panasonic Leica 25mm F/1.4 @ F2.5, Hoya Variable ND Filter, ISO 160, Shutter 1/50. Audio: Rode NTG-1 + Tascam DR-100MKII.
Lightly graded in FCPX

I am also the main actor in both. I had a friend cam op for me after I set the exposure.

Thanks Vitaliy and driftwood.

luxis

Moon 5 spans on L setting on 64gb/95mb san disk.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters