[Toke]

Barry:”Especially when the alternative is MPEG-2, which has been around for over 12 years…”

What’s wrong with this industry if we would have to choose from two codecs both more than decade old?
How about mp4 (good enough for hdcam-sr), cineform, DNxHD, sheervideo, etc.?
I don’t see anything “wrong” with dv-codec, other than it has been implemented only with 8 bits of color,
but I would assume that compression research has developed a bit in last decade so that newer codec might
be a little bit more efficient.

Barry:”But have you seen VariCam footage? What exactly needs to be improved there? It’s absolutely fantastic.”

Can you tell what exactly is so fantastic in it? Do you still say that HDV is also fantastic?
Color grading HDV is quite impossible and it isn’t fabulous even with hdcam. Neither it is with dvcproHD.
8 bits are just not enough for acquisition. Everybody in still photography or motion film knows this.
Using same color depth for acquisition than for distribution leaves very little space for changes, which
leads very careful lightning on the set and that costs more time, which costs more money.
If you want to have top quality pictures.
Other thing is that when you can’t change the look at post, you’ll have to make more work before the
shooting for testing different looks.

Barry:[P2]”will in fact throw away the repeated frames.”

If so, that’s good. Even if it will record 1080p24 with pull-down, that’s only 20% wasted space.
I’m just wondering what really are the constraints of dvcpro codec?
With p2 it isn’t datarate any more? It isn’t the pixelrate either?
Is it just fixed compression rate of pixels*?
If these are not constraints any more why even pull-down?
Little expansion of the “format” every couple of years wouldn’t harm too much.
That’s what JVC is doing to HDV by implementing 24p.

With tape dvcproHD has had this magical boundary that everything has to be recorded 60p/100Mbps.
At the same time sony’s hdcam has been able to record with 5 different datarates.
Why the difference?

I think that main problem with these cameras are that they are still made for “full auto”-ENG use.
If professional still photographers have been able to shoot RAW for years, when it’s time for
us in the moving image to do the same? When do we get first high quality RAW-compression, where you
can make all gamma and color corrections in the post with high color bit depth?

With these new cameras, where last constraints of tape recording have gone away, wouldn’t it be nice
that user could choose the level of quality vs. capasity?
In still cameras you can select the level of jpg.

Compression could be fixed (like dv codec), but user could choose resolution
(480i, 480p, 576i, 756p, 720p, 1080i, 1080p), framerate (23.976 – 60), color depth (8 – 12 bits) and
color resolution (RGB, 422, 420).

Depending on a subject user could choose to record from 1 min of highest quality to 1 hour of the lowest
one to same amount of mass memory.
So far dvcproHD chips can handle 4 or 5 different compression rates. How about if user could choose which
one of them to use?
With progressive picture 4:2:0 is as good as 4:2:2 with interlaced picture, so that would be important
choise also.

So if I would need absolutely best quality for a one take commercial that will be heavily FX’d,
I could choose 1080p, 422, 12b, 48fps and 3.3x compression.
This would lead to 480Mbps, but it would be in boundaries of p2 and could also transferred
live through fw800 to computer. 15 second take would be less than 1GB, so if transfer rates
aren’t fast enough to p2 or hard disk this amount of data could be easily stored in-camera
memory buffer or computer’s RAM.

Another example would be long document shooting, which wouldn’t need any effecting and not even
consistent color timing. There I would choose 720p, 420, 10b, 24fps and 5x compression.
This would lead to 50Mbps.

Third case: consert multicamera shooting for live projection on big screen. Band also would like to
have the gig recorded only for viewing purposes and budget is extremely limited. Because of small
cameras handheld on stage 30fps is chosen for a bit clearer picture. So:
1080p, 420, 8b, 30fps, 6.6x compression.
This would lead to 75Mbps.

Lastly the kids birthday party for cd-r: 480p, 420, 8b, 24fps, 6.6x = 15Mbps (6 mins on one cd-r).

Cameras with bigger ccd/cmos chips might also have full resolution as an option.

How does all this sound?
I’d like to hear also other professionals (Barry, yours are of course always welcome) opinions from
different fields of shooting.

I may have given a bit over critic impression towards Panasonic (Jan), I have to say that if one
compares Panny to Sony, Panny has always been more user friendly than Sony which gives whatever it
likes from their ivory towers. Haven’t seen any sony’s camera compression codecs in sofware versions
(like digibeta & hdcam for FCP)…

So for 2nd gen tapeless cameras, all I am asking is totally user selectable combination of
resolution, color depth, frame rate and compression amount. All these options already exist in
the camera except option for choosing color resolution (422, 420 and maybe 411 if somebody still
believes the need for interlaced picture.)

3rd gen cameras might then have RAW-compressed option and 4th gen propably RAW-uncompressed (or losless).
(After a decade that might be economical…)

Very important feature would also be getting the identical data out of camera than recorded signal.
That’s needed when recording times exceed internal mass memory or its more convenient to use external
recording (eg. multicamera shooting, timelapse or highest quality).
With dvcproHD there could also be “legacy mode” that makes frame padded and pull-down signal for
old tape decks.

One thing I’m missing now with HD is 1/2″ cameras. One decade ago we had Canon EX1/2 (less than $5k then).
Now we would need that sensitivity to HD’s pixel sizes again.

Just my maybe a bit more than 2 cents…,
toke

*Quick calculation:
ntsc dvcpro: ((720*480+2*180*480)*8bit*30fps)/25Mbps=4.97664
pal dvcpro: ((720*576+2*180*576)*8bit*25fps)/25Mbps=4.97664
ntsc dvcpro50: ((720*480+2*360*480)*8bit*30fps)/50Mbps=3.31776
pal dvcpro50: ((720*576+2*360*576)*8bit*25fps)/50Mbps=3.31776
720p60 dvcproHD: ((960*720+2*480*720)*8bit*60fps)/100Mbps=6.63552
1080i60 dvcproHD: ((1280*1080+2*640*1080)*8bit*30fps)/100Mbps=6.63552

(1080i50 dvcproHD: ((1440*1080+2*720*1080)*8bit*25fps)/100Mbps=6.2208 ???
or: ((1280*1080+2*640*1080)*8bit*25fps)/100Mbps=5.5296 ???
720p50 dvcproHD: ((960*720+2*480*720)*8bit*50fps)/100Mbps=5.5296 ???)

[Toke]

Are you really saying that camera manufacturers has left out
electronic low-pass filtering with these non-changeable lense cameras?

Anyway, if you have seen pictures from that Italian modded fx1 with eng-lens,
it shows much more resolution.

And none of real interchangeable lens camera leans to low pass filtering of the lens.

You can put Zeiss Digiprime in front of any camera and see no anti-aliasing.
All filtering is made electronically after ccd.

[Toke]

It’s not just multiplying the pixels.
But the datarates do multiply, so you need very little
changes to the hardware, just building more pipes.

[Toke]

Where did you get that 6.7x realtime?
I read that it takes 4 minutes to transfer 4 GB and compared to realtime,
well it depends what you have recorded.

If you have recorded 100Mbps dvcproHD, then transferring will be less
than realtime. Which makes sense in a way, that if 2.5″ disks would be
fast enough for higher than realtime, they could be used without p2’s.

For faster on the field transferring there’s at least 4 possibilities:
1) 3.5″ disk, which eats batterys very fast.
2) 2×2.5″ in raid-0, increases chances of hd breakup
3) dropping the dupe frames in transfer (if it hasn’t been done in-camera)
4) 4×1.8″ in raid-5, this would be safe, power saving and large capacity
(180GB with todays disks), getting close to Kinetta…
Looking at iPod prices this might be in $2k area, but of course much
smaller volumes.

[Toke]

I’m sorry, if I have understood duping frames to p2 incorrectly.
I thought that was “verified info” somewhere in these chats…

[Toke]

Barry, dvcproHD-codec is just 4 dvcpro-chips in a same board, just like dvcpro was 2 of those.
So nothing new here in perspect of codecs for a long time.
Can you come up a logical explanation, why this new p2-cam does not throw away repeated
frames to better use of it’s very short recording time?

[Toke]

I’m not going to wait several years for some new holodisk, that can store whole weeks shootings and then suddenly brakes up or gets lost.
Dvd-r is WORM and already 25c/GB. Dual-layer will be as cheap next year and blu-ray year after.

Just thinking the storage capacity 8GB p2 isn’t even at the same level than decade old miniDV-tape.
Next year when we get 16GB p2’s it will get past that, but it will be tough choise comparing to xdcamHD,
which was just told to have 72Mbps and 50Gt capasity. If that stored data is actual picture data it will have almost double the bitrate at 24p compared to dvcproHD. Well, maybe the prices are also way more than double…

[Toke]

Fps of the camera is not about playback, it’s about recording time of each frame.
Hvx200 could have been a great cranking device, but if the rumors are true that
there will be different version for 60Hz and 50Hz countries, there might not be
so many speeds.
I can’t come up any reasons why the new ccd’s wouldn’t be able to use several
different frame rates. (23.976, 24, 25, 29.97, 30, 50, 59.94, 60, etc.)

Shouldn’t it be easy to have any kind of frame rate with solid state recording?
Or does the decade old dvc-codec somehow restrict the development.

[Toke]

Still assuming that ccd has similiar sensivity than 100A-ntsc,
1280×1080 would result pixel size of 9.4

[Toke]

Good question Graeme.
Lets calculate:

dvx100a (ntsc) has 380,000 pixels and chips are 4.4mm x 3.3mm.
Sensivity is about 640 ISO.
So you get 640 ISO with a pixel sized 38