Sombody correct me if I'm wrong.

Are the techs with HDV consortium nuts?

HDV uses MPEG 2 based video stream at 19 to 25 megabits?
Every frame is an I-frame?
The audio is compressed with MPEG 2 (mp3 type) compression at 364k?

Somebody tell me this is perfectly OK!

Don't the basic laws of compression say that if you encode audio at that rate, you will drop 60-80% of the original sound that "they say" that we can't hear? Let's say that I apply a simple EQ to the HDV audio and render that to a new file...When I recompress the audio I will strip 60- 80% MORE audio from the original (already compressed) source!!

Even though the file is still 386k, it will sound like 128k. This is JUNK! Why wasn't straight PCM audio used?

How about the video side? If I apply some simple color correction, will I be left with a video stream that is equivilent to an 8 megabit MPEG 2?

Somebody tell me that MPEG 2 is a great format for editing! (Isn't that why we all HATE those DVD disk cameras?)

I know that HDV has a 100% I frames in the stream as opposed to 1 every 15 or so,...but darn,...MPEG 2 is good for a final distribution format,...not source capture.

It seems to me so far that HDV is a horribly "lossy" format...a big step BACK from DV on the editing side.

Somebody tell me that I am totatly wrong.

I hope that I somehow am.

CT

Yes, it is lossy, however, if I am not mistaken, so is the DVHS format (to a lesser degree) and it is plain awesome. Numbers are nice, but not a complete answer. As the camera is not yet available, I would reserve any judgements or conclusions until seeing the video first hand. As a teaser, what I have heard second hand is that it is very impressive. Suspend disbelief, and the knowledge of the means to the end (compression/numbers) and just evaluate what you see. I was very skeptical of MPEG 4 as in WMV9 for similar reasons, but was very pleasantly surprised when I saw a demonstration of Terminator 2 side by side with the DVHS version. They were indistinguishable at the same time as the WMV9 version having several times the amount of compression as the DVHS.

True, we will need to see how the video side of HDV develops. I have never played with MPEG 2 with data rates above 15 megabit. However, the audio part of this can be tested by most of us right now.

Start playing with a simple mp3 file. Take a nice CD rip at 386k. Take the file and ad a simple effect to it and then re-save it as a new (recompressed) 386k file. What do you hear? ( especially in the higher frequency sounds of the material)

Did HDV creators REALLY need that much of the 25megabits of bandwidth for video? They actually HAD to steel audio bandwidth for this?

Does anybody know if the HDV standard allows for linear PCM audio? Or has that been completely ruled out of the format?

Also, does the HDV standard natively save to .avi? or MPG/Mpeg.

CT

Some very good and provocative questions. I think it is to early to tell...at least, I am not able to answer them at this time. Anyone else?

I have not seen one single article in any forum on any website that addresses the "lossy" issue of HDV.

Anybody have any more thought to add to this topic?

Are there any serious compression theory people in the audience?

I was thinking about why they opted for an old MPEG 2 based codec over something that is far more effecient like an MPEG 4 based codec.

Can you imagine a Windows Media 9 or MPEG 4 HDV stream at 25 megabits? How much higher would the resolution have been if they chose a far more effecient CODEC like MPEG 4?

The only drawback I can see would have been the MONSTER processing power required to capture and edit that stream.

Anyhoo,...back to the real world. If I capture something to HDV. I dump it to an editor have some fun with it and save my "master edit" back to HDV ( first compression loss). I then take my "master edit" and maybe add a color correction to it that I might have not noticed the first time and save it again to HDV. ( second generation loss) I then transfer it to any other final distribution format. (third gerenation loss) Not to mention the audio left behind after all of this the would be totatly useless. Unless you enjoy "swooshing, swirling, phase" noises in your audio tracks.

I guess what I'm saying is that it's impossible to work and save in 100% HDV and not loose quality on every save. I guess their answer was to throw as much bit rate as possible into the video so that the losses would be "tolerable". I guess this is why Sony considers HDV to be a "consumer" product.

Hmmm,...why hasn't Canon joined the HDV consortium? What is their issue with it? (interesting)

I'm dying for more opinions on this. Somebody say anything. I'd even welcome a good "blasting" of my posts!

CT

Currently the only answers to these questions are from the JVC 1 chip camera. Soon (very, as Kerr Cook...the lucky bastard...is getting a tester this week) we will hear about the results with the FX1. Come December, consumers will have their hands on them.

Yes, it is lossy. The JVC uses 19.7Mbps MPEG2, the Sony (HDR-FX1 and HVR-Z1 "pro") are 25Mbps.

The audio WILL be encoded. MPEG1 Layer 2 at 384kbps. This is NOT "MP3" (MPEG1 Layer3 which is optimized for lower bitrates).

I didn't have time to do a lot of audio tests. I did do some "sounds of silence" to see if there was hiss, pumping, zoom/focus/head motor noise, etc. If I could have had a few more hours I'd have fed a good CD into the ext lin input, as well as a good mic and hooked up an XLR mic (using an XLR converter).

I was very very impressed. Yes, I could force some MPEG2 video artifacts, but I was really abusing it by asking too much. Zooming in and panning 90 degrees a second, etc. Pausing the playback shows artifacts... but if I raised the shutter speed manually the artifacts reduced greatly (and the pan became more stroboscopic).

The colours were incredibly rich. Fire, autumn leaves, colours at a carnival/fun day, my daughters's bright eyes, pastel/bright clothing, many shades of flowers and roses, etc. It did well with all of these ON PLAYBACK though the built-in LCD and Viewfinder were "off" a bit with being pale and lacking as much "yellow" as reality.

__________________
Sony HDV camcorder info (FX1, CX7, HC3 HC5, SR1/3/5/10, SR11, SR12, TG1, UX1, Z1) -> http://www.SonyHDVinfo.com

The "output" is MPEG2, but stored (at least the way I found to do it) in a .M2T file (for MPEG2 Transport). I could rename the files to .MPG and media players worked directly with it... mind you I had several HDV codecs loaded too though.

I am trying to edit the .M2T files to get them down (mine are 1 to 9GB!!!) so I can upload some clips.

I used CineForm's ConnectHD component "HDLink" to do the transfer (HDV over Firewire to the .M2T files on the computer. They sent me a Beta copy that supported the FX1's 1080i HDV.

A regular MiniDV camcorder CAN NOT read nor transfer the HDV even though it is the same bitrate (25Mbps) as DV.

There isn't a lot known about the HDV editing software yet. Many people are using a CineForm codec to make an intermediate AVI that they can edit in without loss.

There isn't a lot known about what "distribution" media to use. HDV is over 4 times the resolution of DVD video disks. Not knowing what will become the standard HDV distribution media (new Blu-Ray optical disks?), it is hard to say what encoding the output should/will have.

__________________
Sony HDV camcorder info (FX1, CX7, HC3 HC5, SR1/3/5/10, SR11, SR12, TG1, UX1, Z1) -> http://www.SonyHDVinfo.com

You lucky dog.

I'm psyched about the FX1 myself. I'm saving up some money right now and I figure I'll be in the market for a new camera around March of 05. I was originally looking at a VX2100 but I think I might keep saving untill I can get an FX1.

More about the HDV codec; I recently did a project in standard DV. However I made the project in reverse order over three months. I edited and rendered the last third of the video in August,...I edited and rendered the middle third in September,...edited and rendered the first third this month. I then took all three finished parts,..added some transition footage and a complete score and rendered my master DV file send sent it to Sony DVD architect.

Looking at the way I did this project in DV,..working this way would have been impossible in HDV given all the seperate renders and re-renders that were involved. ( not "impossible" but significantly lossy)

If you have the codec on your PC, can you take the raw footage that you shot and re-render it once and then take the second generation and re-render it again? I'm dying to hear about your results!

I have two weird questions and I think they are pretty heavilly rooted in compression theory.

1.) In theory,...could a DV capture program grab the HDV stream and convert it to a codec that is equivilent to DV but with HDV pixel dimensions? ( no real generation loss) This new "Hi-Def" DV .avi file would need to be 40 megabits (or more if you do the simple math)...BUT your recording would exist in a format that would survive multiple renderings. This would allow you to edit and render and re-render all you want. Your final project would be saved to the format of your choice and endure ONE generation loss only at this final stage. ( The audio of my dream "Hi-Def" DV .avi codec would of course be linear PCM )

2.) If this is not possible, could you completely overhaul an editor's video "engine" to allow for all functions to be completely performed in the HDV MPEG 2 compressed domain? That sounds bizarre but I'm not sure how to word it. Can all transitions and cuts and so forth be made so that they don't "re-render" per say,..but modify pixels within the compression calculations themselves? (Like,..never forcing a frame to leave the MPEG 2 world and back again.)

LOL,..question 2 sounds totatly weird...does anybody get my drift?

Regardless,..I'm glad the FX1 still shoots DV. Looks like DV will still have capabilities the HDV Codec can't compete with.

I'm very excited about the FX1 but the lossy nature of MPEG 2 drives me nuts.

I hope I'm firing some people up about this issue. I'm hoping we can get a vigorous debate going on this.

CT

The only one I know who can answer all those questions is Sean, our engineer. He can be reached at www.hollywoodstudiorentals.com/bb where you could direct him back to this board for comment.

__________________
Cheers and Best Wishes

Some good questions, Cliff.

That's a brand new board, MisterJulius! Best of Luck! You (and Cliff and Sean) are also invited to see SonyHDVInfo.com for more discussion as we "reverse engineer" and try to figure out what Sony has done with CineFrame24, MPEG2-TS (the 25Mbps MPEG2 over Firewire), etc.

__________________
Sony HDV camcorder info (FX1, CX7, HC3 HC5, SR1/3/5/10, SR11, SR12, TG1, UX1, Z1) -> http://www.SonyHDVinfo.com

I hate to tell you this guys, but DV is just as "lossy" as HDV. It seems to me that any situation that would cause HDV to re-render would have the same effect on DV. The only exception to this would be if you make an edit that breaks up a GOP - and then only the frames between the edit and the next I frame would re-render. (There is no way that HDV could be all-I-frames. That would make it no more efficient then the DV codec, and hence unable to record any more detail.)

You are never going to see a lossless consumer codec. Uncompressed NTSC has 5 times as much data as a DV tape can hold, and uncompressed HD would be up to 25 times as much.

All current (and likely, future) video codecs utilize compression to some degree. However, this fact does not imply parity....ie, there is lossy compression, and there is lousy lossy compression, and everything in between. All prefer the former, and it is a matter of compromise (file size, quality) and refinement over time. Overall this format is not a bad starting point, but there is room for improvement.

Hello,

Some input regarding some of what was said or asked.

HDV using a 6 frame GOP (group of picture) structure MPEG-2 transport stream compression format composed of one "I" frame (entire picture) and 5 "P" frames (predictive) that contain changes to the picture information in that original frame.

The JVC camera uses a lower bit rate for compression (about 19 Mb/sec) because of it's lower resolution, and the fact that it has less picture information to compress (1280 x 720 progressive frames) compared to the Sony (which records 1920 x 1080 interlaced frames) at 25 Mb/sec compression.

You can't really compare the DV codec to this HDV (MPEG-2 based) codec by using data rates or frame structure (it's like comparing apples and oranges somewhat). And the fact that their spatial resolutions are different makes for a difficult comparison.

Cineform transcodes the HDV video to their visually lossless HD video (using their codec tecnology) to enable more efficient editing.

Hope this was helpful and interesting.

Stephen

Thank you, informative and fairly concise.