Sony HDR-TD10 3D camcorder announced at CES

[Alister Chapman]

See this thread for more : http://www.dvinfo.net/forum/digital-...unced-ces.html

[Pavel Houda]

The inter-axial is 31mm. They don't expect anyone to edit it, according to the rep.

[Arnie Schlissel]

Not to be outdone:
JVC Full HD 3D Consumer Camcorder Is World?s First | JVC Consumer Press Releases

[Pavel Houda]

N.A. Withdrawn by P.H.

[Steve LaPierre]

The JVC camera seems to shoot in 1080x60i whereas the Sony is doing 1080x60p. I would think that with 3D the interlacing of image capture could contribute to depth problems for objects that are moving so the progressive scan should be superior. Of course neither of these cameras are going to produce the 3D that this group expects in general but at $1500 for the Sony it sure seems like a great way into this format for me.

[Steve Shovlar]

I think the Sony sounds great as a second camera. I film a lot of weddings and aim to offer 3D weddings this year, so as a pickup camera it makes much more sense than two Canon HV40's side by side.

I have put my Panasonic SDT-750 up o Ebay. Whilst its a great 2D camera for the money, the 3D was not great, as it's not true HD, the light loss horrific and no zoom!

[Pavel Houda]

I don't want to get into the i/p conversation, but I don't agree that interlacing somehow messes up the 3D, and I know first hand. Also, according to Sony rep, they store the video in MVC internal packing and only plan to output frame packed pixels (HDMI 1.4). Think of it as 3D Blu-ray player in playback mode. No USB, no directly editable video formats, like AVCHD, MPEG-2/4, side-by-side, dual stream, etc. They designed it for people who will take the video, plug the camcorder into their TV and play the captured video back as is. Knowing the hoops I have to jump in order to edit 3D properly, that maybe the right decision for "people" 3D camcorder. There will be enough headaches even this way. They will add a convergence adjustment on the final product, the prototype didn't have any. That means they'll have to do keystone correction internally. The rep said that they'll come up with 3D camcorder capable of outputting editable video later. That would not do for me. I cannot even post this without editing it 3 times.

[Bruce Schultz]

Originally Posted by Steve LaPierre

The JVC camera seems to shoot in 1080x60i whereas the Sony is doing 1080x60p.

It's been posted on another thread and forum that the Sony is recording 2D in 1080 24/60P and unfortunately 3D in 1080 60i interlaced. I hope that's wrong.

Always some Baloney with Sony.

[Pavel Houda]

Speaking of Sony announcements, I saw a Sony 3D camera at the CES 2011 which I never saw before. It can be seen here: YouTube - yt3d: 3D Camcorders @ CES 2011 - Side-by-Side, Cross-Eye , at about 7:26. Is anyone familiar with it? There was no marking on it.

[Steve LaPierre]

Originally Posted by Bruce Schultz

It's been posted on another thread and forum that the Sony is recording 2D in 1080 24/60P and unfortunately 3D in 1080 60i interlaced. I hope that's wrong.

Always some Baloney with Sony.

You're correct that it is 1080i not 1080p, I must have read the specs too quickly the first time. Still seems pretty good and at 60i the interlace vs progressive is probably a moot point as Pavel has suggested. The output being limited to "•Video Signal : 3D HD: HDTV 1080/60i; HD: HDTV 1080/60p, 24p, 60i; STD: NTSC color, EIA standards", is certainly a problem for this group but for an amateur lilke myself it might be OK. It appears to use Memory stick and SDHC, I wonder if that is stored in the MVC format?

[Adam Stanislav]

Originally Posted by Steve LaPierre

certainly a problem for this group but for an amateur lilke myself

But, Steve, you're a member of this group! ;)

[Pavel Houda]

The camcorder stores the data in MVC (bitmap+depthmap) packing and outputs it in HDMI 1.4a, which is 24p stream, which "this group" should like, unfortunately, it is "frame packed". I am not sure how to separate this into two separate, parallel HDMI (convertable to SDI) 24p streams, which should be desirable. Some hardware interface performing this function is possible, I just don't think it exists at this moment. Of course for serious amateurs this would be too expensive and complex. I think this camcorder is for wealthy people requiring an "instamatic" camcorder. The JVC GZ-TD1 is interesting as well, because they keep talking about the "popolar" sid-by-side output for 3D, which could mean that they'll not even have the frame-packed output of the Sony....

[Alister Chapman]

Never heard of bitmap plus depth map MVC are you sure about this? I would have thought it would be a normal MVC frame pack as per the BluRay 3D specs. In which case it should be pretty easy to extract the two streams as it's just a multiplex of the two full streams. AVIsynth with MVC decoder can de-mux a 3D bluray I believe.

[Pavel Houda]

I probably said it too popularly, but I called "depth" was trying to say more simply that basically MVC is trying to minimize the amount of bits that need to be stored or transmitted, so they do not store two fully independent frames, but use reference and the compressed predicted differences in both time and space. These cannot be directly used by editors. That is shown somewhat more clearly here: INTERNATIONAL ORGANIZATION FOR STANDARDIZATION . The result is saving of both storage space or potential transmission bandwidth (which of course is not the issue inside player or camcorder, but the storage space is). It is described in an extension of the h.264/AVC standard (Annex H), and it is in some aspects similar to AVC algorithms. The main point is that there aren't two full frames stored. It is not standard Blu-ray frame pack, until after it is decoded (decompressed), where after further formatting, it is sent as frame packed (HDMI 1.4a) stream, which consists of fully decoded L/R serially multiplexed and re-framed streams, packed into nearly standard video frame timing. This decompressed combined stream can be separated (the display devices are doing it), but as I said I am not aware about any commercially available interface device doing it right now, (with the exception of circuits and software embedded into displays and TVs) so that such stream(s) can be edited. The editors generally, internally deal with flavors of MJPEG compressions of the two separate streams, either done on the fly, or pre-processed, because the computer CPUs aren't usually powerful enough to do it so that editing can search randomly and play smoothly. That is why they sometimes use H/W assistance from the display H/W, which often uses the required algorithms implemented in H/W for other tasks. So what I should have said was that they are storing compressed prediction, backward prediction and correction between the two frames and in time as well, in a single stream, which to me sounded too complicated, so I called it "depth" information. The compression algorithms are trying to take advantage of the similarities of the two views to highly compress the frames. The stream outside the Blu-ray consists of decompressed pixels for both views in a single stream.
I don't think that the full solution, camera to editor to storage to display right now very economical to implement and support for consumer grade devices, and I don't think that it is easy. I don't think that it is possible to ask a consumer to go through all these professional "work flows". That requires too much technical knowledge. Turn key, simple to understand and support solution must be provided for that audience. That is why I think Sony is not implementing it on the first generation of the 3D consumer grade camcorders, and why Fuji never officially supported editing of the W1/W3 streams.
MVC is not just simple multiplex of the two streams, but the HDMI 1.4a decoded output is. (I am sorry, I am running into my English language limitations and talking too much here.

[Alister Chapman]

Thank you for the in depth solution of how a hardware based playback system would work. The bottom line though is that the files saved by the camera contain enough data to reconstruct the 2 discreet left/right video streams. So it should be possible to create a software decoder that can unpack the data from the files and extract the two L/R clips. That decoder, like any other codec could be transparent to user, so the workflow need not be any more complicated than existing workflows, assuming you don't want to manipulate the 3D itself.