kinoseed

5 hours ago, Geoff CB said:

On 4/29/2016 at 11:12 PM, outerbeat said:

Can't get rid of the quote box. So has anyone seen any further development of the 2.5K with the possibility of 24p  framerate?

for NX500:
Mod-menu > Custom Func. > Mod Resolutions > (select) DC as 2.5K
in camera-settings set and record DC video, it will be 2.5K at 24fps

NX500/NX1 have programmable hardware and in essence you set up the hardware which resized the input for the encoder. We found the registers, but working with them directly is problematic as soon as you update part of the registers stream gets out of sync and drops out (outerbeat gets the credit for that one). Modifying the encoder's resolution can't be effective at all by itself, as no matter how you change it's recording resolution the feeding video will be already downsized. This at least should be quite clear.
You can change any resolution to any other (with NX-KS it was available even for the last firmware) as they in fact only shuffle the available resolution, so you don't set anything "real".

When it comes to MJPEG, something similar will be much easier achievable by setting  IPeriod: 1 (or 0?), thus forcing HEVC with I-frames only. This is something that I shared with the vasile more than a month ago, when there was still some illusion for cooperative work. ; )  My linux skills are far less advanced and takes me more time to figure some of the things, and this was not on my list.

cheers

Before someone again asks for dna-proofs and starts spreading his sheets, and to avoid "measuring of calculations" contests...

here's a short clarification (ex 40mbps):

when you look at vasile's code you will see: 2362F2C0 2300F645
when you use vasile's mods you end up with those values set in memory: (45f6023 c0f26223)
where these values: (4ff4b443 c0f26223) are set by SAMSUNG for the same bitrate, before any mods, and for sake of sanity, let's take them as the "correct" ones.

( you can verify that yourself )

That closes the topic about the bitrate values as far as I am concerned.

btw, if you are wondering how "set $bitrate_040 = 0x2362F2C02300F645" actually "works out" in the end:
- it's written backward in pairs (to get what you set in memory, you get the last pair "45" from the end and make it first, "f6" second, and so on), which was rectified before setting in memory, so you in effect set this:
0x45f6023 c0f26223
which gets you half way to the correct value, which is:
0x4ff4b443 c0f26223

Thankfully the miscalculated first part is of non significance, so you get the bitrate, even though you set garbage in 50% of the memory patch.
again, be very sure what you patch and where

On 5/31/2016 at 8:26 PM, Otto K said:

@vasile is correct - if you are not sure what you are doing - don't do it. Added a large warning to github. 

And one should be very careful what values are used if you mod the bitrates.
for example lets take 40mbps, vasile's mode has: set $bitrate_040 = 0x2362F2C02300F645
However this is not the bitrate that should go in the address, but the real one is: 0x4ff4b443c0f26223
(and of course nx500-pro has a spacial case)

so keep that in mind too, if you plan to make changes you better use the real bitrate values

ps. if vasile's "had it" or not is his problem, but so far what I said were only facts

ps2: if you have questions, or look for the latest info/downloads for nx-ks, please use the facebook page, as I don't check the forums often. https://www.facebook.com/NXKS2/

I guess some clarifications are in order.
NX-KS works with the new fw versions (1.12 / 1.41) for both cameras, and sets the bitrates as before, with one addition of 400Mbps for time-lapses.

@MountneerMan, vasile was never really part of NX-KS, from the very start I did uses his mod, but very soon after Otto made "poker" I started using that, which allowed for faster loading, and higher stable bitrates. Vasile, even tough apparently he obfuscated the bitrates and had to be "hacked" to be used with "poker" was given full credit.

Lets say that "falling out" is put mildly as there was never a "functioning team", often resulting him demanding of proof for this or that, like when I posted a very high-bit video, he demanded that I show him jump-table for a hack or give him credit for the video that I made, as "apparently I modified the bitrates using his spreadsheet" ... and so forth. Lets say the constant slander was almost a harassment.

When the new FW showed up, I gave him the new addresses, but the values were suspect, and after waiting a week for him to produce the new bitrates (as I suspected were different) he just said "they are the same" while apparently he's been working on his "mod". Well something didn't add up as the values from Samsung did not match Vasiles.
So I redid his hack, and found he had miscalculated the values 4 bits out of 8 bytes, (which was not critical in the end, but gave different values) maybe it was not from his calculations, but error from his obfuscation - no idea, didn't really care to check his work that much.

I can't speak of his new "mod", as I have not checked it, and don't intend to.
But since he didn't contribute anything for the latest NX-KS mod, and only made sure his work can't be really used as opensource, was enough to let him go and do his own thing, although teaming with Otto was great.

That about sums it up. : )

But of course I had to say Otto has my full respect!
He is the one that made it all possible.

NX-KS has been available for the new firmwares for a few days now:
https://www.facebook.com/NXKS2/

(there is absolutely no need to stay with the old firmware)

23 hours ago, tugela said:

Debeyering uses information from adjacent pixels to reconstitute the color of the virtual pixel. If you use information weighted to a large circle of probability, the color will be more accurate but resolution lost. If you reduce the weighting of that circle of probability then resolution will be maintained but color will be inaccurate. That is what creates the halos. You would only need one setting to do that.

Once debeyered, the image would have permanently lost the information that was discarded, so either color accuracy or resolution at boundaries. Operations in post cannot recover either.

The process you describe will never produce contrast halos, only add blur. Debayer does take information from grb photosites to calculate the value of a pixel, but again, that happens in software, uses a weighted matrix, and there is only one that I saw in the kernel. (no probabilities are involved)
More on debayer: https://en.wikipedia.org/wiki/Bayer_filter

@Daniel Galli you can start from here:
https://github.com/ottokiksmaler/nx500_nx1_modding

52 minutes ago, tugela said:

Doubt it. What you are talking about is software sharpening. The "sharpening" that happens in camera is just a debeyering parameter. The camera either chooses luma weighted debeyering, or chroma weighted debeyering, depending on the settings chosen. Just remember though, once it is recorded, the information you have discarded by your choice can never be recovered. You cannot adjust it in post, because it is gone.

Doubt it.
Sharpening has nothing to do with the debayer, and there was only one such pattern in the kernel that I saw, and if you were right, there would be only 2 settings.
However the most common (and seems here is no difference) is the USM (hence the halos). And yes, even debayer is done in software. ; )
imho to avoid the USM use -10

cheers
 

Bulbramping (minus the need for a "bulb")
just keep in mind, that NX500 when in rolling-shutter mode will not do longer than 1/15 exposures
have fun

3 hours ago, vaga said:

If Kino Seed can get this out of his NX500 with bitrate hack, I think you have very little to worry about using this camera in lowlight. @kinoseed what are your in camera settings? (I guess there's no Gamma DR on the NX500, so this will make a very interesting answer, I think.)

about the exact setting, honestly - can't remember
it must have been with shot with the Rokinon, so assume F2
the original footage looks close the the second grading here, the first - curves correction was to check the details and info in the highlights, and the second for the shadows.

18 minutes ago, SMGJohn said:

I might be overreacting a bit, artefacts is something I rather not have in my videos but its something I can live with if its not as visible and the bonuses outweight its downsides.

After all, with no NR being applied and no digital sharpening is definitely enough for me to continue shooting 1440p25 which looks bloody gorgeous.

But the interlacing look alike artefacts appear more frequent for me for some reasons, I have no idea what causes them, maybe its the heat? Its around 20 - 30 degrees over here so I am giving the camera some workout, I use 140mbps I noticed with DC it uses the pro setting which was set to 150mbps for some reasons the NX500 wont do 160mbps on my 128GB Lexar card but it will do 160mbps on my NX1 just fine but 24fps is not in my taste definitely.

Also did the NX500 have NR applied in UHD and 4K mode before your hack is applied? Because as I stated earlier on page 64, UHD and 4K seem to lack the presence of NR and have similar noise patern like the 1440p mode, is this something you can confirm? Or was it already there before the mod was applied?

honestly, those cameras are a bit like a black-box, and you never quite know what and why they do it.
if your 2.5K turns out "interlaced" is absolutely useless. imho, it is not really interlaced, but looks like it, and you will not be able to fix it.
It is not the temperature, as in the span of 15 min, I shot 4-5 clips, and only one was "interlaced". Why - no idea. I don't think I changed the settings then.

My own camera can't be used as a reference as it has been messed with extensively, however I noticed that @outerbeat's video looks much more NRed, and at least here (on my nx500), the effects of NR are far less noticeable. In the mod there isn't anything that specifically changes NR. However especially at ISO 3200, I get quite a nice grain, which does not look NRed at all (at least at the very high frequency)

I though that it may be due to the different models, and the NX1's are the one suffering a lot from NR, but if you notice fine grain on NX1 footage, maybe it's something else.

edit:
if you have noticed it only on the NX500, and is different than before, the only difference from NX1is setting the VGA to 2.5K, which I'm not sure how may affect it, but there is also the "silent shutter" which you probably have enabled (but if it affects the NR - I don't know).

15 minutes ago, SMGJohn said:

Update on testing the 1440p mode on NX500 both VGA replace and DC replace 25/24fps.

Because I did not want to drag with me my 5kg 15.6 inch Clevo laptop to Greece I took my 2006 12.1 inch Dell laptop which shows my HEVC files like a slideshow thanks to its awesome nVidia GeForce Go 7400, anyway during my testing which I have done quite a lot of while I also tested various profile settings (I will share my result when I get home) I recommend to everyone to avoid the 1440p mode overall, there is too many artefacts here and there to be useful, it seem to band worse than 1080p which I did expect but my biggest issue it introduces these ugly lines all over the screen which looks like interlacing which varies too much from my test they are usually quite small but video players will think the video is interlaced and try to deinterlace it if you set it to automatic, most of the time they only appear in the sky, other times they appear pretty clearly in the video, I also had to turn down the bitrate to 140mbps even though my card does 160mbps just fine, the DC 24fps 1440p mode seems even more unstable, definitely avoid that one.

Just stick to shooting 1080p for now unless you can live with those lines appearing in ever 3 - 6 videos, funny enough I can shoot 160mbps 1080p and 200mbps with my fastest Lexar card so technically you are getting better quality overall out of the 1080p which you could just stretch to 1440p, remember that there are more bits per pixel for 1080p compared to lets say 2160p so you could drag more info out of shadows etc, remember that 1440p is NR free so there is definitely a bonus 2.5k but also a huge drawback, it definitely should be avoided if you do pro videos.

I would have uploaded some examples if the internet here was not slower than Pluto's year

I included the 2.5K as option to be set on the DC channel (for testing), and ran a few test shots today, and looks like there is no difference in quality from when being on VGA.
However with regard to the "interlacing" you are talking about, I get it only once in a while - maybe once in 50 recordings.
2.5K does exhibit more problematic artifacts indeed, and increasing the bitrate does not help here, however it's far from "useless", well... I guess it's a matter of personal preference.

here is an example from today's 2.5K / 180Mbps test

out of camera:

applied RGB curves only:

32 minutes ago, Garth Philpot said:

On a side note, would anybody be interested in working on a pdf with stills or an instruction video that we could embed in the post, so as to make this quite clear for new users. ie check firmware, reset camera to defaults, wipe card, place files in root etc, I can edit it in an afternoon, but would like someone to just check that the information is indeed correct. maybe some small points relating to what cards work best, and what speeds you will get out of them (maybe 5 most popular etc) I have 3 lens adapters arriving for EF, OM and M42 if there is any info that can come from those too...  

Checking-firmware is not essential, as if you don't have the proper one, the installation will let you know.
Wiping-card is not essential, as you can have other files and directories during installation, and it won't be a problem.
Resetting to defaults is not essential either, as it will work (i guess) with wherever you've mapped EV button to be.

recommended tool for SD-formatting is this one: https://www.sdcard.org/downloads/formatter_4/

there is almost no difference in the cards (almost), and the recommended bitrate depends a lot on whether you can with a few dropped frames, what fps you intend to shoot, what scene, at what ISO. (btw, my personal ISO choices are 100 and 3200, as they show least amount of high-frequency band stop effects). My personal settings are -180Mbps.

Got m42 graphite adapter (costs more than some of the lens), which apparently was created "upside-down", so when I mount the lens, the dials that have to be on top, are on bottom, (hope yours doesn't have that design issue) however it doesn't bother me, but the only really big concern is cold temperatures, and extreme condensation from the old lenses in winter days (which need to be removed immediately)

Since it's not a DSLR I guess the polarizer does not need to be circular, and you can go with linear (if they still sell them) and save yourself part that brings LP light to circular (which is sort of sandwiched in the circular polarizer filetr). You can stack two of them and make yourself a variable ND too. ; )

Other than that, if you have questions feel free to ask.

1 minute ago, Garth Philpot said:

I havnt remapped any buttons, but reset them to default just incase. EV = UP for me brings up the spirit level and histogram( i think). but I get the mod loaded etc, what I dont have is the BT icon top of screen, even though it says on in the menu

 

It's ok not to have the BT icon. You can turn on the bluetooth from the menu (which looks like you can't start)
Sounds like the keyscan is not loading, or maybe you are pressing EVF instead of EV?
sorry.. can't think of a reason why it wouldn't start.

14 minutes ago, Garth Philpot said:

thanks folks, let me put down what i can see, i appreciate your input.

body firmware is at 1.40.

it i look at your screen, i see a difference in that I have no BT icon at the top of the screen. reset all defaults in the menu, and re-installed the mod version you linked to...

here are the screen shots of how your screen is different, and how my menu says BT is ON

 

Yes, confirmed, press EV (hold) click ok, then release both.

 

the only other thing I can think that different is I dont have a lens to attach yet, I will have that tomorrow.

 

 

It would not install if you didn't have the correct firmware, so you don't have to worry about that.
If after boot, you see something like [Mod V1.xx], that means Mod is loaded.
Now check the key-scan. Press EV_UP, if it works you will get a message "Preset missing"
(I wonder if the keyscan didn't load for some reason, or your buttons are mapped differently)

48 minutes ago, Garth Philpot said:

ok, I have installed 1.67b, after a card format. it installs successfully, and it brings up the version number. still no difference for me (EV+ OK). what should this new menu look like ? perhaps I am just not seeing it and its a small icon or something ? do I have to have completed the dev menu unlock that i read somewhere back in this forum for this to work correctly ? thanks

also, it will not install 1.67b, it stays reporting 1.67m, even though it says [lets clean up'...] , now mod free...

Don't forget to turn-on Bluetooth, then press EV, hold it, and while pushed down, press OK (if you are an NX1 user), otherwise double-click EV.
When in doubt, follow the instructions to a tee.

11 hours ago, SMGJohn said:

The way HEVC works is definitely contributing to macroblocking and the NR specially which is present even at low 100 ISO.

I tested HEVC extensively by converting my huge 400GB Pokémon in original Japanese to 750kbps HEVC.

I did probably hundreds of tests and this was back when the X265 was quite young I also tested the H265 in Adobe Media Encoder CC 2015, same issues as always it seems to cause macroblocking in dark areas I also tested a 1080p version with fine grain option of the first episode of Pokémon and HEVC smears the details away, its gotten slightly better with more modern releases but NX1 has old HEVC from 2014 unless they have updated it via firmware which is possible but even then its still out of date from today's versions.

Interesting enough I found 8bit saves grain details more than 10bit does which is funny. Not saying recording is same as encoding but HEVC is designed for compression and not for details. When you go higher in bitrate the point of saving space becomes completely lost as H264 10bit seem to perform as good if not better, and this is witnessed in cameras like Nikon which has under 100mbps H264 yet produces beautiful images with very high dynamic range and good shadow details.

I can't agree.

Here is a 4K frame from 160Mbps mod, nx500, pushed around to check the latitude for grading.
(it was done in the rain, can't tell you the ISO, it's a bit soft as it was shot wide-open, and there are raindrops on the lens ... but it worked for me )
I did put the video on youtube, and it showed banding, macro-blocks and so forth, only after it was converted to h264, here are some frames direct from HVEC (the original is slightly underexposed):

 

1 hour ago, Marco Tecno said:

Pavel, very interesting. I still can't understand the reason for the reduced framerate when a higher bitrate is chosen. The higher the bitrate, the less work cpu has to do.

 

Btw, is there a way to create a shortcuts for video like this? :

Ev+fn= 4k, pro, pal, 25fps, 1x, dis

Ev+mobile= 1080, high, ntsc, 30, 0.25x, no dis

 

you can try this: http://***URL removed***/forums/post/57581215