That thread shows the deficiencies of a compressed stream vs. a live uncompressed feed, which also happens to be at a higher resolution. Did you really think compressed HDV was going to look as good as uncompressed?

I have already addressed it over there, I am not surprised it is worse than uncompressed, but how much worse it is on the simple scene where the codec can render near it's fullest quality, it is a matter of context. I have followed up with an explanation over there.

I thought it might be interesting to examine the action and deficiencies of HDV on the scene which is what you guys are discussing (though, the chock scene is better place to check for motion artifacts which is what the thread, and Dan, was about). The best way to compare this is with an untouched/uncompressed frame. The frames were both recorded at 1440 resolution, and upscaled to 1920.

http://www.dvinfo.net/conf/showthread.php?t=52319


Here is talk about a bunch of h264 cameras coming out, Samsung's one has a bit rate of around 17-19Mb/s:

http://www.dvinfo.net/conf/showthread.php?t=58391


Kevin,

I've heard that TV/DVD movie companies have very expensive sophisticated compression setups and DVD movie companies can guide/vary the compression process through the dvd to maximise the look. So while a DVD might often still lack, because of the choices made they look good. On a camera though, it is good to have double that bandwidth of the consumer spec to star with. The extra bandwidth of XDCAM (often nicknamed HDV2 around the place) goes a long way to double HDTV bandwidth, and probably worked out to give just enough to give good results in most of the cases where HDV wouldn't, I would not be surprised if it has a broadcast quality Mpeg2 compression engine on board. But the problem with HDTV, I've looked at the channels and have never seen footage with that sheer volume of compression artifacts on Mini-DV/DVD, and I live in a 25p country. I think it is set to low to start with, and for a 19Mb/s stream I would much rather trust a h264 camera codec than a HDV camera codec. But then again most peoples TV's resolution is below the standard and hide artifacts through pixel combining.


Dan,

I'm with you in some ways. I often think about putting together a cheap low compression/no compression HD solution, but lack an engineer to do it with, and it includes parts of another confidential commercial oriented product that has to be done first, when I feel better. I've got a few things to get through before I can consider starting it though, and this Samsung camera might be good enough not to even worry about doing the HD solution.

An excellent suggestion. For those who have been ranting and railing against HDV ad nauseum, please find something positive to talk about. HDV is very much here to stay, and I will not allow format wars on these boards... such negativity is an utterly pointless, useless waste of cyberspace and time which serves no worthwhile purpose here at all. Please find somewhere else to vent. Thanks in advance,

Wayne, that link for the Hard disk cam has no mention of HDV spec. In fact HDV is a tape only format. The side of the cam has clear HDD writen on it which means it uses a hard drive. No mention of HDV or HD anything except from other posters. No new HD cam has been showen from JVC yet, and there has been no statement from them either.

As for your other thread.
"The frames were both recorded at 1440 resolution, and upscaled to 1920."

The uncompressed frames are captured at the full 1920. The HDV frame was captured at 1440. Then as you state, was upscaled. This would account for some of the softness in comparison. As well it should be noted that HDV cams are not all the same. While the Sony's have recieved deserved praise for their image/price, they offer the softest least detailed image of the HDV cams. You can't really blanket a format based on one product.

Sorry, you must be looking at the wrong link on the thread. Most of the links that the posters over there are talking about do mention and show a picture of what is supposed to be the HD Everio, and even mention HDV and 720p recording, except the last link, the poster has seem to got the wrong camera (he is looking at a standard def version by mistake. There is a link there to a thread that has somebody that actually was at a JVC dealers meet were they mentioned the HD Everio version. Most of the links are from some JVC press release/meeting. The story goes, that they are not sure when they should release it and are delaying it (I think, probably until they sell most of the HD100's to early adopters).

As far as I understand, the HDSDi output of the Canon only transmits a version of the original 1440 image, we can probably ask Alister. The softness of the HDV footage is a lot wider than 1/3rd of the pixel, looks more than two pixels.

Sorry for the confusion Ken, but I don't think I'm misreading it, and I think that ends my conversation on this thread. I wish reviewers would get serious about reviewing cameras/codecs in-depth, they see and Iceberg and write about it usually, no in-depth Analysis of what is beneath what is easy to test (I have, at times, thought of even making my own test charts to test fro various things through out the whole colour/luminance scale). I think professional engineers at TV stations must have a good laugh at some of the testing that goes on some camera sites. If I had the full setup and cameras here, I would love to do very in-depth testing. As I've shown, just getting values and pointing the thing at a correctly lit test chart will only scientifically test the tip of the iceberg, you really need to get hold of footage (preferably using it yourself) to get an idea of/experience the rest. If such an opportunity comes up I would like to have an accurate knowledge of all compression schemes, their artifacts and characteristics (When and to what degree they happen where) first. So, if anybody has links to everything. technical about camera image and codec faults, please let me know, better that one man swallow the nat and strain the camera than every man have to. Sorry Chris, if I've been a bit annoying, but I think their is a need for analysis of the camera codec performance, to see how well things will come up in practice.

Thanks

Wayne.

The HDSDI output of the H1 is full 1920 resolution. The H1 CCD's are 1440x1080 BUT the camera uses pixel shift to get full frame resolution. Have not measured the actual resolution, but it is certainly higher over HDSDI than HDV. If you take a 1920 frame and downsample it to 1440 it certainly has a very similar look to an HDV frame. There is a cineform frame on the other thread that was created from the HDSDI stream, the 8 bit Cineform codec is only 1440x1080 and this frame has the same softer look. It should also be remembered that even HDCAM is downsampled to 1440 and I have yet to come across any domestic display that can actually resolve more than 1000 TVL. Most HD LCD's, Plasmas or CRT's can't resolve more than 700 TVL.
To my eye HDV is excellent, It is more than capable of pushing the limits of the playback/viewing equipment and artifacts really are not an issue in most situations. It's not perfect, but it is pretty damn good.

I think many of the issues people see are more down to the fact that it's an 8 bit codec and thus has only 256 grey levels, ie quantization and stair stepping on areas of nearly flat colour.

except for digibeta all SD tape formats are 8bit ...
all SD & HD tranmission/broadcast are 8bit ..
pretty much all HD is 8bit to tape except for the 24p camera's that are FILM camera types ( viper, panavision genisis, dalsa )
most 35mm film that is scanned to digital is 10bit files ( not 16, 24, 32) .. and then transferred back to film 10bit ...

it looks like we're stuck with 8 bit for awhile and i'm not seeing any sign of 10bit as a delivery format ( other then the transfer to film) ...

I have spent some time comparing that cineform HDSDi frame at magnification to the original and it has a little blur, but nowhere near that of the HDV one. I read, probably the Internet news compression FAQ, that the tables in Jpeg lack enough precision to be give a accurate result (in the lossless/jpeg section) so there precision is increased for lossless compression to the extent that the error will not effect pixel values. Maybe this is related to the problem?

So your saying it produces 1920 true res rather than 1440 upscaled to 1920, or 1440+pixel shift unprocessed upscaled. Interesting. I saw the original post on the Canon HDSDi, that talked about 1440 res, that's where I got the figure from, but I accept what you are saying, sorry Ken. It's hard to believe that they would get this right, but I'm happy that it turns out out be.

I'm curious, what ever happen to the plan to use the Z1/s CCD's in the Canon.

BetaSP being analog dosn't suffer from many of the stair stepping and quantisation problems that most of the digital formats suffer from. However I would not for one moment want to go back to analog. Transcoding to 10 bit or higher for post production is getting easier and can improve the final image.

Canon do not quote the image resolution anywhere, simply that the CCD's have 1,560,000 pixels which is just over 1440x1080 and that pixel shift is employed which normally increases the resolution. Stills taken with the camera are grabbed at 1920x1080 so I think it is reasonable to assume that the camera head is operating at 1920x1080.