D5100 Bit Rate ?

All things embedded.
*NO FEATURE REQUESTS*

Re: D5100 Bit Rate ?

Postby nikkor » Mon Feb 24, 2014 7:51 am

But it doesnt make too much sense going over 65mbps because the d7000 upscales from 1642x924 to 1920x1080 in this very mushy way. So in adition to sub hd, there is this bad subsampling happening. So don't expect GH2 quality.

Nevertheless the improvement is great. Can't wait to try it on my d800.
Image

[edit] I have made another test. I have set up the d7000 and the canon 50D with the same lens (nikkor 85 1.4 D) same aperture (f8) and same speed. On the d7000 I'm using 65mbps and on the 50d I have used ML RAW and 75mbps.
The 50d is 15mp and records 1584x892 in Raw mode. Which is roughly the same as the d7000 does. The 50d has 1.6 crop factor, the d7000 1.5, so the canon has a little bit more "zoom".

First goes, 50d raw, then d7000 65mbps, then 50d 75mbps.

Image

I think the first two look similar resolution wise, the last one is worse (not only because of the colors).
nikkor
 
Posts: 138
Joined: Mon May 20, 2013 11:06 am
Been thanked: 44 times

Re: D5100 Bit Rate ?

Postby skiibojosh » Mon Feb 24, 2014 9:06 am

I'll just add to the side where everyone is already leaning: the d5100 is 8 bit 4:2:0 by the time it records. I also have to say- all of you are awesome! This is such an incredible feature! I'm testing the 54 bit rate on a D5100 w/ firmware 1.02. It works perfectly. I did some tests with 54 at 30fps and 24fps and then the standard (21) at 30fps and 24fps. I didn't notice as much of a difference until I colored the footage and then the 21mbs shots seemed to break quicker. It seemed that noise was more pronounced at the lower bitrates. The blocking was larger. This feature works great!
skiibojosh
 
Posts: 18
Joined: Thu Jan 31, 2013 11:30 am
Been thanked: 23 times

Re: D5100 Bit Rate ?

Postby coderat » Mon Feb 24, 2014 9:36 am

nikkor wrote:But it doesnt make too much sense going over 65mbps because the d7000 upscales from 1642x924 to 1920x1080 in this very mushy way.

How do you know this ?
We review scaling right now, so more information to come.
nikkor wrote:So in adition to sub hd, there is this bad subsampling happening. So don't expect GH2 quality.

Nevertheless the improvement is great. Can't wait to try it on my d800.
http://farm8.staticflickr.com/7321/12748826203_a61d8bd595_o.jpg

Sorry, but where is 65MBps ? What do we see here ?
So in adition to sub hd, there is this bad subsampling happening. So don't expect GH2 quality.

Does GH2 4:2:2? But this is not the end anyway.

Best regards,
coderat
coderat
Core Developer
 
Posts: 2230
Joined: Fri Apr 26, 2013 10:21 am
Been thanked: 401 times

Re: D5100 Bit Rate ?

Postby skiibojosh » Mon Feb 24, 2014 10:22 am

Maybe I spoke too soon- I'm not able to figure out if it's based on certain circumstances, but I have noticed that the 54mbs sometimes ends abruptly with a warning that recording is interrupted. There is no consistency as to how long the clip is before it stops (I have had it vary from 00:01, 00:01, 00:02, 03:31,and 07:31) Is there anything else I can tell someone about it that would help it? Is it possible that this is an sd card issue? I was using a SDHC class 10 which I imagine is too slow, but it's all I have with me.

At 24fps, it looks like the max time you can achieve is something close to 10m (9:57 is my longest recording). Hope that's helpful in some way.
skiibojosh
 
Posts: 18
Joined: Thu Jan 31, 2013 11:30 am
Been thanked: 23 times

Re: D5100 Bit Rate ?

Postby Dany » Mon Feb 24, 2014 10:29 am

First thing I would like to congratulate everyone here on working on these hack for the Nikon vdslr users.

The second thing is to take what Andrew reid says with a big big big pinch of salt. The guy is just about grabbing headline to drive traffic to his site. Look at his latest Nikon d5300 (supposed) review and the comments he got from it. It was just a rant and not one side by side or chart to see how the camera compared to other cameras. He is just about grabbing headlines, today he might write how the hack is evolving well and next day trash it like how he wrote glowing reviews of the D5200 and now trashing the D5300 a year later, while it is a better camera (no fix pattern noise and 60p). There is nothing factual about what he does/says and that is why on site like dvxuser and personal-view he is considered a joke (They call him voldermor on PV).
Dany
 
Posts: 1
Joined: Mon Feb 24, 2014 10:13 am
Been thanked: 0 time

Re: D5100 Bit Rate ?

Postby coderat » Mon Feb 24, 2014 10:44 am

skiibojosh wrote:I have noticed that the 54mbs sometimes ends abruptly with a warning that recording is interrupted. There is no consistency as to how long the clip is before it stops (I have had it vary from 00:01, 00:01, 00:02, 03:31,and 07:31)

So you need faster card. Class 10 is not enough. Because it is not only writing data stream, but also updating FAT, etc. So we are using Ultra- cards for such bitrate, but surely they are also all different. Or you use second option in patch: 38 MBps. This will also increase time limit (that is tied at the moment to 4GB).

Regards,
coderat
coderat
Core Developer
 
Posts: 2230
Joined: Fri Apr 26, 2013 10:21 am
Been thanked: 401 times

Re: D5100 Bit Rate ?

Postby ebstein100 » Mon Feb 24, 2014 10:56 am

nikkor wrote:But it doesnt make too much sense going over 65mbps because the d7000 upscales from 1642x924 to 1920x1080 in this very mushy way. So in adition to sub hd, there is this bad subsampling happening. So don't expect GH2 quality.

Nevertheless the improvement is great. Can't wait to try it on my d800.
Image

[edit] I have made another test. I have set up the d7000 and the canon 50D with the same lens (nikkor 85 1.4 D) same aperture (f8) and same speed. On the d7000 I'm using 65mbps and on the 50d I have used ML RAW and 75mbps.
The 50d is 15mp and records 1584x892 in Raw mode. Which is roughly the same as the d7000 does. The 50d has 1.6 crop factor, the d7000 1.5, so the canon has a little bit more "zoom".

First goes, 50d raw, then d7000 65mbps, then 50d 75mbps.

Image

I think the first two look similar resolution wise, the last one is worse (not only because of the colors).


First of all, correct me if I am wrong, I do not believe if field of view is same the resolution is same despite the same focal length. On the same camera with same lens choosing between small, medium and large jpeg produce different resolution image with same field of view. D5100 and D5200 with similar sensor size and field of view (crop factor) produce 16 and 24 megapixels respectively. Hear of line skipping but never of upscaling for 2 megapixel fullHD video from 16 megapixel sensor on factory default setting.

By the way thank you for the frame grabs. Comparing to 50D Magic Lantern raw, d7000 65mbps seems equal in resolution and dynamic range with similar highlight and shadow details. I guess 50D 75mbps is h.264 which has blown highlights and loss of details in shadows (less dynamic range and contrasty).
ebstein100
 
Posts: 46
Joined: Sun Jul 21, 2013 12:23 pm
Been thanked: 15 times

Re: D5100 Bit Rate ?

Postby skiibojosh » Mon Feb 24, 2014 11:06 am

coderat wrote:
skiibojosh wrote:I have noticed that the 54mbs sometimes ends abruptly with a warning that recording is interrupted. There is no consistency as to how long the clip is before it stops (I have had it vary from 00:01, 00:01, 00:02, 03:31,and 07:31)

So you need faster card. Class 10 is not enough. Because it is not only writing data stream, but also updating FAT, etc. So we are using Ultra- cards for such bitrate, but surely they are also all different. Or you use second option in patch: 38 MBps. This will also increase time limit (that is tied at the moment to 4GB).

Regards,
coderat



Thanks coderat! Do you think it would ever be possible to have the D5100 record to exfat? I was looking at some SDXC cards and saw that they supported exfat (which doesn't have that 4gb file limit). Allowing the d5100 to write exfat to a 64gb or 128 sdxc card would be an interesting feature to partner with high bitrates. I may try the 38 mbps! thanks again!
skiibojosh
 
Posts: 18
Joined: Thu Jan 31, 2013 11:30 am
Been thanked: 23 times

Re: D5100 Bit Rate ?

Postby ebstein100 » Mon Feb 24, 2014 11:09 am

skiibojosh wrote:Maybe I spoke too soon- I'm not able to figure out if it's based on certain circumstances, but I have noticed that the 54mbs sometimes ends abruptly with a warning that recording is interrupted. There is no consistency as to how long the clip is before it stops (I have had it vary from 00:01, 00:01, 00:02, 03:31,and 07:31) Is there anything else I can tell someone about it that would help it? Is it possible that this is an sd card issue? I was using a SDHC class 10 which I imagine is too slow, but it's all I have with me.

At 24fps, it looks like the max time you can achieve is something close to 10m (9:57 is my longest recording). Hope that's helpful in some way.


Maybe it will help. Magic Lantern users recording large Raw files to SD card have had better results with formatting SD card in camera and using card warm up before recording. By the way same class rating SD cards from different manufacturers have different max write speeds.
ebstein100
 
Posts: 46
Joined: Sun Jul 21, 2013 12:23 pm
Been thanked: 15 times

Re: D5100 Bit Rate ?

Postby nikkor » Mon Feb 24, 2014 11:12 am

Sorry I forgot to mention the details.

http://farm8.staticflickr.com/7321/1274 ... d595_o.jpg
the first one the d7000 with standart codec HQ. Second image is d7000 with 65mbps. Great improvement! But the higher you go the less you notice, so my guess 65, or a little bit more is the sweet spot.

Regarding the gh2. The GH2 has a 4:2:0 codec, like every camera out there (the only consumer options with 4:2:2 are the blag magic cameras, that use a 10bit 4:2:2 prores codec. And the GH4 which isn't out yet). But the GH2 gives a lot more detail because it actually does some sort of pixel binning to a 1920x1080p or close image.

The d7000 and d5100 use lineskipping. Both have a 16mp sensor which is 4928x3264 pixels with 3:2 aspect ratio. The 16:9 area would be 4928x2772. The camera has to convert this into a 1920x1080 image. The less processor intensive way is to lineskip. Every pixel is made of a matrix 2x2 ({R,G},{G,B}), so what the camera does is read just the first line {R,G}, then it skips the next one {G,B}, it passes to the next pixel line and skippes the first line of that pixel line {R,G}, but then it reads the {G,B},etc... this guy explains it pretty well.
http://falklumo.blogspot.com.es/2012/04 ... ction.html (this is for the d800 but the d7000 should do the same.
There are 2772 pixel lines, 5544 sensel lines (2*2772). So, it reads one line, skips two. 5544/3 = 1848. You need two lines for one pixel: 1848/2= 924 pixels. [Bullshit detector]
I wrote something stupid, it's simply 4928x2772, reads one, skips two, and again and again... 2772/3 = 924
So you end up with a 4928x924 image. The camera combines horizontal sensels in another way (not skipping but averaging 3 of them, you could ask someone over at Magic Lantern on how the canon do this as it should be very similar) and you get 1642. So in the end you have a 1642x924 image.

As you can see here http://farm8.staticflickr.com/7317/1275 ... 6a86_o.jpg I compare a raw video frame from the 50D (first one)which also does this frame skipping with the d7000 @ 65mbps (second image). If you zoom in on the 50D you can see these small magenta and cyan pixels at the contrastry edges of fine detail. This is consecuence of line skipping. When you look at the d7000 you can see the same thing, but nikon hides it with it's way of scaling up to 1080p and compression. Since the detail is similar I believe they both use the same method (both have almost the same MP).

I can't wait to see what you are working on :D, the bitrate helps a lot.
Last edited by nikkor on Sat Mar 01, 2014 4:24 pm, edited 2 times in total.
nikkor
 
Posts: 138
Joined: Mon May 20, 2013 11:06 am
Been thanked: 44 times

PreviousNext

Return to Firmware

Who is online

Users browsing this forum: No registered users and 2 guests