[Becks21]

Does anyone has some info on the exposure-table for the K-5 video mode?

falconeye did some research on the K-7 and estimated the following table:
Falk Lumo: K-7 as a movie camera -- PART II: Controlling video recording

Is it the same for the K-5? Whats the highest ISO possible in video mode?

Thanks and Cheers!

[RonHendriks1966]

Well I did some video with the K-5 but did not spend enough attention to messure things. The K-5 uses definetly some more room in high iso settings. At least it uses 6400 but maybe even iso12800.

http://www.youtube.com/user/RonHendriks1966

[YT]http://www.youtube.com/watch?v=48YqCBFPvXg[/YT]

Image quality is better to my notice at FullHD, i didn't had a good feeling for K-7's 1536x1024 size.

[Becks21]

Thx.. nice low-light vid.. sadly the youtube-codec is blurring away any luminance-noise I fear..

I did some exposure-tests on my own:

1.) dimmed the lights in my room down until:
2.) I got a on-match exposure in Av with 1/50s f2.8 ISO3200
3.) then switched to video (1080p), set f2.8
4.) stopping further down would immediately lead to under-exposure (f-stops turn red)

So I doubt that video will go beyond ISO 3200 (like the K-7).. but I dont know if my testing was sufficient.. any ideas/comments?

Cheers!

[falconeye]

Thx.. nice low-light vid.. sadly the youtube-codec is blurring away any luminance-noise I fear..

I did some exposure-tests on my own:

1.) dimmed the lights in my room down until:
2.) I got a on-match exposure in Av with 1/50s f2.8 ISO3200
3.) then switched to video (1080p), set f2.8
4.) stopping further down would immediately lead to under-exposure (f-stops turn red)

So I doubt that video will go beyond ISO 3200 (like the K-7).. but I dont know if my testing was sufficient.. any ideas/comments?

Cheers!

Well, my finding for the K-7 was it uses the unexpanded ISO range, whatever was the actual ISO Auto range setting.

Assuming somewhat the same for the K-5, it would be relatively safe to assume that base ISO in video is ISO 100.

Then, one should videograph a dim enough target with a moving element. Ideal is an LED lamp swinging and hanging from a cord or something similiar.

Then, superimpose two consecutive frames. The ratio of length of the dark gaps to the bright LED traces (in between) gives the exposure time t.
t = px-bright / (px-bright + px-dark) * 40ms where 40ms is the frame duration at 25fps (PAL).

Once the exposure lowest exposure time is known (and base ISO) is known, everything else follows from exposure reciprocation.

1. Find the stop where the slowest shutter time is used. That's usually the time where a plasma TV stops banding in the resulting video.

2. Stop the aperture down to find the highest available ISO (before red blinking).

3. Open the aperture up to find the shortest available electronic shutter speed. You may additionally need a brighter target for this test though


I've not done this kind of tests yet. Another aspect worth to look into is noise.

My very first tests seem to suggest that the K-5 does NOT do row skipping in creating the video feed (i.e., line skipping only). There may be some sort of binning but I've to do more tests. If this turns out to be true, the K-5's low light performance should be able to rival the one of a D5mkII.

[Becks21]

Thanks a lot for your reply..

Well, my finding for the K-7 was it uses the unexpanded ISO range, whatever was the actual ISO Auto range setting.

Assuming somewhat the same for the K-5, it would be relatively safe to assume that base ISO in video is ISO 100.

Then, one should videograph a dim enough target with a moving element. Ideal is an LED lamp swinging and hanging from a cord or something similiar.

Then, superimpose two consecutive frames. The ratio of length of the dark gaps to the bright LED traces (in between) gives the exposure time t.
t = px-bright / (px-bright + px-dark) * 40ms where 40ms is the frame duration at 25fps (PAL).

Once the exposure lowest exposure time is known (and base ISO) is known, everything else follows from exposure reciprocation.

1. Find the stop where the slowest shutter time is used. That's usually the time where a plasma TV stops banding in the resulting video.

2. Stop the aperture down to find the highest available ISO (before red blinking).

3. Open the aperture up to find the shortest available electronic shutter speed. You may additionally need a brighter target for this test though

Probably I am going to have some time during xmas and new years for some more testing (or should I say reverse engineering?).. I wonder why Pentax doesnt just reveal some details in their documentation..

I've not done this kind of tests yet. Another aspect worth to look into is noise.

My very first tests seem to suggest that the K-5 does NOT do row skipping in creating the video feed (i.e., line skipping only). There may be some sort of binning but I've to do more tests. If this turns out to be true, the K-5's low light performance should be able to rival the one of a D5mkII.

Could you tell me what you understand in the difference between row- and line-skipping and why it would be important?

As for the pixel binning, I see two applications:
- suppress aliasing artifacts induced by under-sampling
- increase light sensitivity (virtually bigger photosites)

Am I missing something..? I am not that firm in this area..

Fröhliche Weihnachten und nen guten Rutsch !

[falconeye]

Becks,

row and line skipping mean that not all pixels are read out when creating a video frame. In general, it's not as simple as skipping. There will be a matrix to describe sensels read out and the matrix will have many zeroes. It's the population density which matters actually.

Cameras with a high density typically read out all sensels in a line but not all lines, the 5Dmk2 being a typical example. The term line skipping describes this.

Typical dSLRs have a lower density and a symmetrical matrix. I describe that by line plus row skipping. Typical matrices only read out 1/9 of sensels and the resulting effective sensor surface is relatively small, like a crop factor 5 P&S.

But a high density (like 1/4) could be made with a matrix more complex than line skipping. It would yield superior quality actually. So, row skipping isn't a bad thing in itself. Only if the term is used to describe a low read out density.

As you say, a higher density creates less aliasing and less noise. And less false color artefacts.

Current dSLRs actually don't have a high image quality compared to a still image. Esp. a still image scaled down to 2MP. Therefore, I consider the read out density the most crucial factor in judging video with a dSLR, more important even than manual controls. A dSLR with a low density yields worse quality than a camcorder. So, some DoF special effect set aside, why should I care about manual controls with it?

[Becks21]

Becks,

row and line skipping mean that not all pixels are read out when creating a video frame. In general, it's not as simple as skipping. There will be a matrix to describe sensels read out and the matrix will have many zeroes. It's the population density which matters actually.

Cameras with a high density typically read out all sensels in a line but not all lines, the 5Dmk2 being a typical example. The term line skipping describes this.

Typical dSLRs have a lower density and a symmetrical matrix. I describe that by line plus row skipping. Typical matrices only read out 1/9 of sensels and the resulting effective sensor surface is relatively small, like a crop factor 5 P&S.

But a high density (like 1/4) could be made with a matrix more complex than line skipping. It would yield superior quality actually. So, row skipping isn't a bad thing in itself. Only if the term is used to describe a low read out density.

As you say, a higher density creates less aliasing and less noise. And less false color artefacts.

Ok thx a lot for your explanation (had to read it three times actually

Current dSLRs actually don't have a high image quality compared to a still image. Esp. a still image scaled down to 2MP. Therefore, I consider the read out density the most crucial factor in judging video with a dSLR, more important even than manual controls. A dSLR with a low density yields worse quality than a camcorder. So, some DoF special effect set aside, why should I care about manual controls with it?

Fully agree.. IMHO, another influence in image quality would also be the video-compression-coder used.. now it "feels" like the coder brings back the motion blurring (I mean even when panning at minor angular speeds) and being able to manually push ISO would bring you nothing then..
this would also become more important if the FPS is increased, to lets say 60fps.. (as wished by some)..

[RonHendriks1966]

There was an artikle at dpreview. Just a little complicated to get it all good for me.


No direct linking, so you can look at the side.

The title was: Understanding Video compression & codecs

[falconeye]

IMHO, another influence in image quality would also be the video-compression-coder used.. now it "feels" like the coder brings back the motion blurring (I mean even when panning at minor angular speeds) and being able to manually push ISO would bring you nothing then..
this would also become more important if the FPS is increased, to lets say 60fps.. (as wished by some)..

Yes, that's true.

But IMHO the K-5 excels in this regard. It's MJPEG codec is old fashioned but refrains from doing motion compression.

Normally, this means poor compression quality for static content and the MJPEG codec has a bad reputation for this reason. But the K-5 features the highest bitrate for a consumer camera to compensate for this (). The result is a very good compressed video quality.

The only caveat is that K-5 videos should be transcoded to, e.g. MP4/AVC for long term archival.

[Becks21]

But IMHO the K-5 excels in this regard. It's MJPEG codec is old fashioned but refrains from doing motion compression.

Normally, this means poor compression quality for static content and the MJPEG codec has a bad reputation for this reason. But the K-5 features the highest bitrate for a consumer camera to compensate for this (). The result is a very good compressed video quality.

So you are not a fan of motion compressing encoders? Why?

I have seen videos of the HDC-TM700 which uses AVCHD, and the videos showing high quality encoded _static_ and _motion_ content.. as far as I can tell/judge.

So I think using that high bitrate is not a plus, indeed more a necessity. Speaking of effectivity vs efficiency.. If Pentax is going to support 60fps in the future, there will be no way around AVCHD in the end.

Anyways.. given the fact that MJPEG compresses very well (at sufficient high bitrates), I would prefer to compensate for motion blurring and have that artifically generated in PP.. so being able to manually raise ISO would come quite handy. Personally I dont care so much about noise in videos, even current BluRay movies have noise like hell in them, I guess to artificially generate "a sharpness feeling" for the eye.. but I may be wrong..

Besides, Pentax must use a different down-sampling matrix for video then for making 2MP pictures. The latter show a lot of jaggies in the picture, so down-sampling to 2MP with a software seems to generate a completly different results, or would it be the MJPEG which blurs the image and reduces the jaggies?

[falconeye]

Happy Xmas

I don't like motion compressed videos because they cause extra work upon import in my workflow. YMMV

The videos I just keep w/o plan to use in a movie project, I transcode to MP4 using the Turbo.264 HD hardware stick. I don't trust the hardware capabilities of current cameras to do decent H.264 encoding in HD and realtime. AVCHD is a dirty workaround IMHO (low bitrate).

A still camera doesn't use a downsampling matrix when creating low resolution Jpgs. Pentax being no exception. It just takes the full resolution image and resamples it for the lower size like Photoshop does. If you get jaggies, then your sharpening was too high. You have to lower it for lower rez output (because there is less softness).

Video is a different story though.

[TomGarn]

Anyone knows the details for the best possible rates and configurations
when you take your K-5 Material away from Premiere CS6 to give it away:
1. for broadcasting services and maybe further editing
2. for to upload it as H.264 to YouTube preserving the most of MJPEG quality

Thanks for a screenshot maybe - or some listings of those many different details
that can be choosen in Premiere - some I am still in doubt with about what to say.
You can create very big files - but when do you do to much and won't get a better quality ?