GSpot, make your GH2 multi-orgasmic

apefos

I got two ideas to try to improve stability in 720p60 NTSC. These two ideas changes the goal of the GSpot configuration but they are an attempt to improve stability, see:

1 - change the 720p60 to GOP4, this way there will be 3 P frames in each GOP, and the amount of I frames within a second will reduce from 20 to 15. I do not know if this can improve stability but is and idea to try. I think the I frames size will increase a little, I do not know... and maybe more P frames means more encoding calculations... I do not know... this can be good or bad for stability... needs tests...

2 - in both 720p60 gop tables (opt1 and opt2) there is the number "43". an attempt to increase stability would be to change both "43" to "42" This will make the I frames a little smaller and the P frames a little bigger, but I also do not know if this can improve stability or not. what I know is that the number "44" in these places did not work ok togheter with the other numbers, the "44" works ok with the "13" in the other place and the "43" works ok with the "14". both combinations have similar results but the "14, ..., 43" and "43, ..., 14" gives a more constant size in the frames flux in my tests, but the tests was handheld... lowering for less than "42" in combination with the "14" will hurt the gspot noise/details benefits in the same patch.

As you can see both ideas can hurt the GSpot configuration where it is possible to get both good noise and good details, because in 720p60 the I frames and the P frames are both in the lower limit to allow both things in same patch. also gop4 will be worse than gop3 to show a filmic noise due to more P frames means more changes to mud the noise.

At this moment I cannot see other options to change for 720p60 stability, just lowering datarate or deal with short shoots with high datarate... maybe increase the frame buffer, but it is already very high and even a higher buffer would not be able to keep all frames if the flux to card is slower than what the codec needs. seems to be the limits of the camera hardware or memory card for 720p60.

apefos

@driftwood Here it is 4 test versions of the GSpot FlowMotion NTSC 96Mbps for compare the 720p60 stability:

G = GSpot original (gop tables 43 + gop 3), for comparison.

H = gop tables 42

I = gop 4

J = gop tables 42 + gop 4

feedbacks welcome

frullaccia

@apefos Dear Apefos, I couldn't test your gspot because I decided to use a better known version (the 13th september Workhorse Flowmotion 120). As soon as I can I will give you some report about the last gspot nebula 112, which at first sight look like one of the best.

driftwood

Did you edit and try my Cluster v4 'All 3 GOP' setting from 2012? See thread http://www.personal-view.com/talks/discussion/3454/driftwood-settings-series-5-cluster-v6-ultra-reliable-mysteron-crossfire-quantum-9b-sedna.../p1

download: http://www.personal-view.com/talks/uploads/FileUpload/87/9c0de6d88cd9af4895ed3725fefe17.zip

frullaccia

@apefos I have been testing gspot nebula pal 112, compared with workhorse flowmotion pal 120. The camera stopped in Fh mode gspot nb, using a Lexar80Mbps; it also stopped in h mode wk fl, using a Lexar90Mbps. As for the picture quality, differences are so little that I couldn't decide. Maybe workh fl has a stronger contrast, while gspot nebula has a best control in overexposed high lights. But this should be tested again in all modes. If I would choose, I'd rather say Workhorse in 24H and Gspot in Hbr (where maybe this patch has a better noise performance).

apefos

@driftwood already downloaded them, did an initial look, interesting gop table settings. I will experiment as soon as I have time to do a careful test. thanks.

@frullaccia thanks for reports, did you got stop recording or camera freeze? I think that the way to go is to use a lower datarate version, you are experimenting with the highest versions, so please, give a try on the 96Mbps PAL setG, it should work ok. or try loading each version to perceive the highest you can use.

As I said before, my card just allow me to test up to the setG versions 96Mpbs, higher than this I cannot do fully tests. but all the things in the ini are the same in all versions, just a change in datarate, frame limits and buffers, so it should work in high speed cards. in each version the frame limits and buffer settings are math calculations using frames size measurements from streamparser, including the H, I and J using the seconds that I could record, the same measurements that I did to the lower versions, so I was wondering it would work. The 24p can handle more high settings, but maybe these very high datarate versions touch the limits of camera hardware for FSH, FH, SH and H. My intuition and feelings are telling me that there is nothing to do in terms of stability anymore, it seems to be the limits of the camera. I will do tests with the patches pointed by driftwood to perceive if they have something which can help, and in other patches also, including the new ideas I got in last posts. but to be honest I think now is the moment to find which datarate you can use among the available datarates from setA to setJ.

I bought a 32GB 95Mbps sandisk today, hope it is not fake one... it will take some weeks to arrive here (45 days)... so I will can test better the higher datarate settings. I think it will be waste of time to try to test the set H, I and J with the 45Mbps card, so better wait this new card to arrive. It will also help me to develop my new matrices ideas.

frullaccia

@apefos The camera freezes. A datarate of 104 works fine on 80Mbps cards. Even a 112 and 120 are ok, but sometimes SH and H mode crash using AFon stabilizer and shaky movements.

apefos

@frullaccia thanks for feedback. AFon and stabilizer are difficult to work in very high datarates for 720p. try to find which datarate version will do the job. in the GSpot for PAL from setE and above are great quality, and in GSpot for NTSC from setG and above are great quality. I believe you can find one to give you more time recording with all these features enabled among all datarate versions. You can also try lower datarates versions below setE Pal and below setG NTSC for use with AF and OIS. With these features disabled maybe it will be more stable in the higher datarate versions. for me to do tests I need to wait the 95Mbps card to arrive.

apefos

@driftwood @frullaccia

I did new tests and found some good things:

did a try in a similar gop table from the "cluster v4 all gop3" , not a direct copy, but a similar design. it delivers a variable bitrate: when there is details it increases datarate, when there is just low light and flat surfaces it lowers the datarate. I perceived that this is not good because it hurts the noise quality in low light, so the previous gop table settings from GSpot are better to hold and keep the noise quality because it makes the gh2 almost constant bitrate no matter the image or light or iso.

did a try in my last uploaded versions using GOP4 for ntsc 720p60 and using the number 42 instead of 43 in the gop tables for 720p60: good results!

the GOP4 is very good for 720p60 because it allows the I frames to increase more the size when there are details in image delivering a better quality. Also, in the low light noise image, the gop4 does not hurt the fine random noise, it is good. comparing gop3 for 720p50 and gop4 for 720p60, the amount of time that the I frames will be on the screen are similar, and the amount of time that the PP (in gop3 pal) or PPP (in gop4 ntsc) will be in the screen are similar also, so the noise perceiving is pretty equal to the eye when comparing gop3 for pal and gop4 for ntsc in 720p.

the number 42 in the gop table instead of 43 did not hurt the image quality and seems to be more stable.

give a try in the GSpot FlowMotion version uploaded september 30 setJ version with the number 42 and gop4, it should be stable and great quality, give it a try on the death charts with AF and OIS desabled and also with AF and OIS handheld in outdoor shoots. If not stable enough compare it with the setI september 30 which has gop4 and the number 43 in the gop tables for 720p60. one of these may be the one for 720p60 ntsc.

apefos

BestICanDo

This seems to be the best I can do for 720p60 NTSC, I cannot find ways to improve the quality and stability anymore, only the 96Mbps version at this moment for stability tests, if it stop recording or freeze the only solution will be lowering the datarate.

@frullaccia @driftwood please do your tests, I recorded 2 minutes of the SDC4 death chart in NTSC 720p60 SH with continous AF and OIS enabled, handheld camera movements and steady, no stop recording, no freeze, 14-42mm kit lens.

HBR 30p and FHS 60i also working in the SDC4 death chart with AF and OIS enabled.

(I only tested the FlowMotion version)

setG = flowmotion matrix

setH = nebula tweaked matrix

apefos

Special Patch Version - GSpot FlowMotion

NTSC and PAL working in the same patch.

AF, AE and OIS working in the SDC4 death chart without freeze.

GOP TABLES tweaked carefuly for 720p50 and 720p60.

GSpot configuration for noise and details quality with improved quality/stability balance in 720p

Try both SH and H, try both FSH and FH, same datarate, but probably there will be one more stable than the other.

My tests shows that more stability than this only can be achieved lowering the datarate.

@frullaccia @TrackZillas @zsound @spreeni @JayUKB @driftwood reports and feedbacks are very welcome.

frullaccia

@apefos I will test it. By now I can't use your last ini file. Please zip it and redo the uploading. Thanks.

apefos

Special versions: NTSC and PAL working in same patch - FlowMotion Matrix

setC = 64Mbps

setE = 80Mbps

setG = 96Mbps

64Mbps, for situations where the electronic features of the lens and camera are a must have. lower datarate allows more stability and works better with the electronic lenses. try using some of these features enabled together, auto focus, auto exposure, idynamic, iresolution, ois image stabilization.

80Mbps, a mid point which allows some electronic funtions to work, but not all. when using electronic features, try use one at a time, many features enabled together can make stop recording to happens. these features must be use with care and not combined: auto focus, auto exposure, idynamic, irasolution, ois image stabilization.

96Mbps, higher datarate increases image quality but just works good with manual lenses without electronic contacts, manual exposure, idynamic off and iresolution off. Try it with electronic lenses, but with all features disabled.

tip: changes in exposure when recording a video can make the camera to stop recording in all versions. when the camera is pointed to an area with sunlight and then it is moved fast to a shadow area this strees the codec and can make the camera to stop recording. pushing the camera to the high limits will never be so stable as can be a conservative settings in the patch. the lower the datarate the less this issue will exist.

Careful tweaking in the GOP Tables after comparing with Cluster patch. Stable versions considering that there is no camera freeze with the Sandisk 45Mbps card. if the camera hardware are puched so far it can stop recording, but it saves the file and camera go back to stand by mode. so there is no problem in the ini file. it is the camera and/or card limits. Try a faster card or load a lower datarate version.

Format the card in camera for better performance. After changing the camera PAL/NTSC always turn camera off, remove battery, insert battery again, turn on again and format card.

when pushing the camera and card to the limits there are some tradeoffs, no free lunch... due to this there are different datarate versions, to chose best one for the camera+card+lens combo combined with the features to be used.

Tests done with the Sandisk 45MBps card

Enjoy

apefos

@frullaccia @driftwood

All settings in the patches are very well done now, seems to be nothing else to do. So now it is a matter of datarate.

before start experimenting with new matrices, I will need to find the correct datarates first, two of them, one for high quality noise and details at same time, and another for perfect stability with all camera and lens features enabled.

The post above already have three versions with this attempt to find these datarates, the higher for best quality and the lower for enable the electronic features. tested with the 45Mbps card.

I just would like to do a careful fine tuning in finding these datarates using the 95Mbps card to get the best possible from the camera.

first datarate will deliver the great quality for noise and details at same time, today the 96Mbps ntsc versions allows this, the pal versions can have this from 68Mbps to 80Mbps, I need to do new tests to find where it is the sweet spot for pal. probably this datarate to allow perfect noise and details at same time will not work with electronic features like AF, AE, OIS, iD and iR. It will be for native lenses with all features disabled and for manual third part lenses.

second datarate which allows to use all electronic features of lenses and body: AF, AE, OIS, iD, iR. This datarate will be lower and probably will not deliver the great noise/details quality simultaneously, needs tests. but will be very useful for using the features from the lens and camera.

I will try to put the first higher datarate in the SH slot and the second lower datarate in the H slot, I do not know if this will be possible, if not I will do two patches, first for high quality and all features disabled and second for average quality with all features enabled.

I think this is what people needs, reduce the number of patches and make it work for quality and stability, if possible in just one patch, if not possible it will be in two patches.

I will need to wait the 95Mbps card to arrive to do this (I ordered one), doing this with the 45Mbps card will be waste of time.

after this, with the two best datarates found, so I will can start experimenting with matrices because the patch base will be ready for it. I know that different matrices can need some tweaking in the patch but it will be small thing to do, because the main development will be already done.

Time to wait the 95Mbps card to arrive...

frullaccia

@apefos Today I'm using your last patch for our documentary film about a famous artist. I already used Wkhrs Flmtn 120 and Gspt Nbl 112 in HBR Pal. I have changed the datarate of this last Gspot to 104. Later I will tell you more about this patch.

apefos

@frullaccia you will get no benefits raising the datarate, just make it unstable. there are many settings tweaked in each version which works together for specific datarate, everything must be changed.

the last three versions are pretty good, stable and great quality. please follow instructions from my pevious post, it tells what version can do.

when the 95Mbps card arrives I will test the 104, 112 and 120 again.

I cannot work anymore without the new card, it is time to wait.

I will stop posting for a while until the 95Mbps card arrives, hope all the people understand...

frullaccia

@apefos Ok, I will use the 96Mbps. In any case today the patch has done a very good job. Frankly, the camera stopped two times this morning, but after a little time everything was ok. This patch is a damn good one! I think I will keep it for tomorrow.

apefos

GSpot: High, Mid and Low - developed considering quality and stability with electronic auto features enabled. Versions 07 october 2014

Use the updated versions, second uploaded zip in this post. (the older first zip is unstable in some modes)

I decided to finish the job for the 45Mbps card, so people who have this card or similar card can use these patches, with special tweaking for using the electronic auto features from lenses and camera body.

HIGH, MID, LOW: GSpot Mature Settings - updated ini files in the second zip

only three patches with correct datarates considering quality and stability.

24p datarate follows 720p datarate due to stability, but with a small step up in 24H.

versions are separated for 720p considering a choice between quality or to priorize the use of the electronic auto features from lenses and camera body.

try both SH and H, there is a small datarate difference and one can be a little more stable than the other. same thing for FSH and FH.

HIGH: setA, tweaked for quality, electronic auto features does not work, top quality for 720p60 NTSC, superb quality for 720p50 PAL.

MID: setB, electronic auto features works just for PAL, with luck only OIS works in NTSC, top quality for 720p50 PAL, average quality for 720p60 NTSC

LOW: setC, electronic auto features works for PAL and NTSC, average quality for 720p50 PAL, bottom quality for 720p60 NTSC

the choice of versions is more related about if you need electronic auto features or not. the reason is because the camera cannot handle high datarates with these features enabled, so it is a matter of priorize quality or features. for PAL this is not a big issue, using the MID setB solves both things, but NTSC is different, see below:

advice for PAL: use version MID setB because the image quality is great for 720p and the electronic auto features works.

advice for NTSC: use the version HIGH setA to get great image quality for 720p or use version LOW setC to get the electronic auto features working.

electronic auto features are: Auto Exposure, Auto Focus, OIS image stabilization, iDynamic, iResolution, ETC extended teleconverter, Noise Reduction. Try using just one electronic auto feature at a time, or try to perceive which features can be combined and used together without stop recording. 24H, 24L, HBR, FSH and FH are more stable with these features. SH and H are the most difficult to work with the features without stop recording, this is the reason there are the HIGH, MID and LOW versions.

these patches push the GH2 to the limits in 720p and 1080i so it can stop recording, but the ini files was tweaked to avoid camera freeze. if it stops, the file will be saved and camera returns to stand by mode, so no need battery pull, no shoot lost.

More conservative datarates because very high datarates just increases chances of stop recording with minimal benefits in terms of image quality.

Careful tweaking in GOP TABLES and DATARATES

Using FlowMotion MAtrix

tested and optimized for the Sandisk 45MBps 32GB SDHC card, and above.

Use the updated versions, second uploaded zip in this post. (the older first zip is unstable in some modes)

frullaccia

@apefos Your previous patch is excellent. Of course you are right: changing the datarates doesn't help at all. I could see it yesterday during my working test.

Going to try this last one.

apefos

@frullaccia please download the last updated versions, more stable, second zip file in my last post.

After tweaking the HML patches improved to make the electronic auto features to work, I perceived that probably the 95Mbps card will make no difference... It seems that the stop recording when using these features in high datarate patches is more related to the camera processor power instead of datarate flux to the memory card. After careful tweaking in all other settings in the ini file, the only way I could make these features to work was lowering the datarate, and the amount to lower the datarate is proportional to the fps (frames per second) being used. More fps means that the camera processor needs to work more to pack more frames. So my intuition says that a faster card will not make the electronic auto features to work in higher datarates for 720p. The updated HML patches (second zip file in my previous post) seems to be in the limits of datarate which allows the electronic auto features to work. I will do tests when the 95MBps card arrives.

After lots of work in the patches, I perceived that the stop recording is different from camera freeze. If camera freezes there are wrong settings in the ini file, the patch is not correct and breaks the codec. The stop recording and return to stanby mode is different, it can be related to two things: a slow memory card, or to a combination of high datarate + high fps + electronic auto features stressing the camera processor. Due to this conclusions I believe the 95MBps card will not make significant difference compared with the 45MBps card, because the card does not improve the camera processor. I will have the answers for this when the 95MBps card arrives.

frullaccia

@apefos I just started trying the last hml 45 zip (not the updated version) and I found it unstable (but with a wonderful 24H performance). I was using a 90Mbps card.

Probably you are right about sd cards speed. A faster card could not improve the camera stability.

apefos

@frullaccia the "not updated version" is unstable in some modes, this is the reason there is the "updated version".

apefos

H-switch

This is a new feature developed specialy for the LOW setC patch version of the HML bundle.

The LOW setC version from the HML bundle is useful for 720p60 NTSC because it allows the use of the electronic auto features of lenses and camera body, but to make this possible the datarate is low to avoid stop recording in 720p60... So I developed a switch to increase the I frames size when shooting a good light detailed image:

Use the SH slot for most situations, including low light, or good light without so many details, also for shoots with motion.

Use the H slot for very detailed images in good light, better for more steady shoots, it increases the I frames size which is useful in low datarate patches to get good details.

Probably some people are wondering why I do not put a high datarate in the SH and a low datarate in H instead of using different patches. The answer is simple, it just does not work. the datarate becomes lower in low light, but in good light detailed scenes the datarate keeps the same in both slots no matter the numbers. This behavior is also present in the Cluster GOP3, seems to be no way to solve, so the low gop patches with gspot configuration in gop tables cannot have different datarates in the SH and H slot. this is the reason I developed two patches for 720p60: HIGH and LOW and also the H-switch in the LOW. People in PAL lands are lucky because the PAL 720p50 has a sweet spot in quality and stability to use the electronic auto features, the MID version.

Things got very mature now, the HML bundle and now with this H-switch are the best patches I did until now.

apefos

@frullaccia @driftwood @Vitaliy_Kiselev @zsound @spreeni @TrackZillas @cbrandin @NikolayM @JayUKB @producer @joethepro @Aashay @caveport @LPowell @maxr @duartix @Azo @towi @yskunto @marciocons @kris @all other patch developers

Thanks for informations in P-V topics, for participation in the workhorse, boson and gspot topics, for all the help, it was essential for me to achieve the results.

It seems that I found the limits of the GH2 for quality/stability/features enabled and also the datarate and gop configurations for highest quality in 720p, and also great quality in the other modes. Now it is time for carefull tests using the GSpot HML bundle with H-switch, to perceive if it is completely reliable or if it needs adjusts.

After that, with the base already done, I will can start to experiment with matrices, taking into account the matrices design, quantization, and also it's interaction with the other settings in the patch, mainly the gop tables.

It is good to solve the mysteries, the mind keeps calm and there is no hurry, the job turn into pleasant to do, and now I can do it in parallel with the other tasks.

Gallo22

Great job, @apefos!

Howdy, Stranger!

Categories

Tags in Topic

Top Posters