Rumored nikon d90 buffer upgrade

1080p should go a long way towards fixing the stair stepping problem, as they should have to bin less pixels during processing.

The easiest way for them would be to crop the 4288x2848 picture to 3840x2160 and then subsample or resize to 1920x1080. Area will be a bit smaller so there will be a crop factor of 1.1 compared to still images.
 
The easiest way for them would be to crop the 4288x2848 picture to 3840x2160 and then subsample or resize to 1920x1080. Area will be a bit smaller so there will be a crop factor of 1.1 compared to still images.

The image is already cropped to a 16:9 aspect ratio before any processing takes place afaict. Liveview will show you the top and bottom frame edges.

As for my own thoughts on price, I do think that $30 is unlikely, but that would make sense for the price of the part. I would expect that labour and shipping would bring it above $100.
 
the mathematician (lack there of) in me can't figure out if the expanded buffer will give the limited cpu power the ability to make 1080 a reality.. we shall soon find out. and like somebody else commented on nikonrumours, the upgrade will cost $$$.....


There is a misconception that D90´s mjpeg compression is made by a hardware compressor. Its software, in the same way as everything else as Picture Control, Active D-Lighting, AF in Live View, etc

I believe that as 2007 Nikon uses a single engine for their cameras (rather than multiple engines as before) its called Expeed processor, and I also believe that D90 has the same chip model used in the D3 and D300, that can work 4K 16bit at 9fps in RAW and JPEG images, so there is bandwidth for D90 to work 1K at 24

Probably Nikon´s next Expeed generation will push at the least the double of that (4x times is my bet)

Most limitations of D90, like the absence of an intervalometer, are software limited. But the major limitation of the D90´s mjpeg is it max compression set at 20m/s, so shooting the sea, or a forest, something with too many details and different shades creates ´blocks´and too many bandings

its possible for D90 to shoot 720P better and even 1080P, but lets see if Nikon is willing to give us that in the $999 price point
 
Last edited:
Hmmm. Well, for it to be worth it they would have to totally fix all of the issues regarding the image. Which tallies up to the stair-stepping for the most part. Don't know if they can do anything about the rolling shutter, but not all that worried about it to begin with.

A very clean 1080/24p would probably result in a damned-near incredible 720/24p, which makes the d90 more valuable from the get go.

Let us see!
 
I'm not sure what will happen with this release, but Nikon has the opportunity to be a game changer. They currently have no in-house video department to compete with. That means they could easily release a camera with 1080p and 24 frames per second. I like the fact that Canon uses H.264 compression and supposedly has less jello, so if Nikon can match them and then one up them with 24 frames per second, I will gladly give them all my money.

My hopes are for an updated D700 (full frame) in Spring with 1080p @ 24fps, reduced jello and better compression.
 
Well, that's another thing: The D90 needs a better format. Something that's better to cut with or hand off to someone else. Right now, AVI isn't all that hot. That's probably just me, I would prefer an H264 file.
 
Hmmm. Well, for it to be worth it they would have to totally fix all of the issues regarding the image. Which tallies up to the stair-stepping for the most part. Don't know if they can do anything about the rolling shutter, but not all that worried about it to begin with.

A very clean 1080/24p would probably result in a damned-near incredible 720/24p, which makes the d90 more valuable from the get go.

Let us see!

Yes, the rolling shutter is a CMOS architecture problem, only CCD (and film) shoots all pixels at the same time (the only manufacturer of CCDs in CMOS quality is Dalsa, which is the most expensive sensor in the market). Maybe adding an accurate accelerometer in the cam could help to remove digitally the rolling shutter in all directions, most still cams already have one (for portrait/landscape information) with the future image processors it is a plausible possibility

Well, that's another thing: The D90 needs a better format. Something that's better to cut with or hand off to someone else. Right now, AVI isn't all that hot. That's probably just me, I would prefer an H264 file.


H264 is the best to edit in Final Cut, very seamless at 1080, but for image acquisition interframe codecs is not a very wise step. Principally for post and effects, we needed something more like DVCPRO
 
Last edited:
There are very few differences between mjpg and dvcpro, at the same bitrate they are comparable. The problem with the d90 compression is the low bit rate, not the scheme. Avi is just a container btw, it doesn't have any effect at all on image quality.
 
I would like to see an increased bitrate, these movies do not take up that much space, i think they can squeeze out a little more data through the pipeline. I can fit approx an hour and some change of D90 footage on one 8 gig SD card.

I just hope Nikon does SOMETHING.
 
I would pay for a FIXED bitrate instead of the current VARIABLE... which inevitably eats the image on encountering blur or darkness.

We already know the bitrate can stay at its highest rate when the camera images a high DR/high detail scene..IT CAN handle the data... it should be no problem to tell the camera just to freeze it at the highest bitrate that it already has no problem doing...
 
I would pay for a FIXED bitrate instead of the current VARIABLE... which inevitably eats the image on encountering blur or darkness.

We already know the bitrate can stay at its highest rate when the camera images a high DR/high detail scene..IT CAN handle the data... it should be no problem to tell the camera just to freeze it at the highest bitrate that it already has no problem doing...

yep, thats the main difference between MJPEG and DVCPRO, DVCPRO is fixed to use in tapes (100m/s in HD?). Nikon is between 10~20m/s, and to tell the truth, I don´t think the major problem is a fixed bitrate, but the bandwith of the bitrate, it should be free to get any bitrate at a fixed Quality, that´s an advantage over the tapes. But we don´t really mind about file size, so I think a DVCPRO for us would be more reliable, but the D90 is meant for general public and they care about file sizes

But what do you mean by "IT CAN handle the data"? at 1280x720, the D90 compression destroys fine detail when there is too much information in the frame.
 
yep, thats the main difference between MJPEG and DVCPRO
mjpeg supports fixed bitrate fine, actually it's more commonly used that way.

am i the only one who remembers the dc30 or the media 100 stations? mjpeg is not some funky digicam format, it used to be the industry standard. :-) dvcpro sucks just as much as mjpeg in exactly the same way; you can never have enough bitrate. there are plenty of next generation intraframe codecs out there, like prores or wavelet based ones.

/matt
 
D90 compression destroys fine detail when there is too much information in the frame.
not really, as you can see from the tests posted in various threads the compression is actually pretty decent, it's the noise reduction, debayering, scaling, whatever that screws up the image.

/matt
 
not really, as you can see from the tests posted in various threads the compression is actually pretty decent, it's the noise reduction, debayering, scaling, whatever that screws up the image.

/matt

As I said before, try to film the sea with the D90, jpeg artifacts appear all over the place at 720P. And I´m not defending DVCPRO to use in D90, but a compression that behaves like it

an example of how the D90´s low bitrate bandwidth can destroy the image (1:1 720P PNG):
au8ly1.png
 
Last edited:
As I said before, try to film the sea with the D90, jpeg artifacts appear all over the place
you're jumping to conclusions. it's a fact that the camera produces video with artifacts and it's a fact that it uses mjpeg compression. connecting the two requires more than guesswork. i've shot maybe six or seven hours of footage, i know what it looks like. my point is that if you compress something else with the same codec and the same bitrate it doesn't look like that, so it's not the codecs fault, at least not by itself. see the "dqt" thread for more info.

/matt
 
you're jumping to conclusions. it's a fact that the camera produces video with artifacts and it's a fact that it uses mjpeg compression. connecting the two requires more than guesswork. i've shot maybe six or seven hours of footage, i know what it looks like. my point is that if you compress something else with the same codec and the same bitrate it doesn't look like that, so it's not the codecs fault, at least not by itself. see the "dqt" thread for more info.

/matt


I´ve made some math here: 20000bit/s 24p = 104 Kbytes per frame

this is a long exposure photo full of information I´ve made with D90 resized to 1280x720 (200%):

34gtt79.png


now look what happens when you convert this image to JPEG at 104 Kbytes (200%):

rqyey8.png


you see?
 
Last edited:
hey, at least they should make shooting 4.5fps stills possible for as long as you want... but at least a little longer... well they won't but we should just have someone hack it for us... there is a forum for hacking Canon consumer cameras like my A630, I wonder why those of you with DSLRs haven't found one like that yet... is there not one out there?
 
Back
Top