Terms like uncompressed and raw video are often used confusingly. Uncompressed is often used to refer to the files from the camera when video from a camera is actually very compressed.

This is a very simplified ( cos I dont understnd it really well ) outline of what it is all about.

Uncompressed video is a beast and needs a mad bit rate. All the following calcualtions are a bit of a fiddle but give an idea of what it is all about.

Lets assume we have a video stream with a colour resoloution of 10 bits, only absurdly expensive cameras can do more than 10 bits, and 25ish frames per second.

The bit rate for HD works out to 1,250 Mbit/sec
The bit rate for SD works out to 200Mbit/sec

Your net connection is unlikely to be over 20mbit/sec. A decent hard drive can record at 500mbit/sec. The only way to record this much data this fast is to use huge arrays of raided drives - something some very expensive cameras do this. For example the kinetta has a whole bunch of 2.5 inch hard drives built into it, eight I think.

So what comes out of your camera - the one you can afford.

DV and HDV ( not HD ) use a bit rate for video of about 25 mbit/sec.

DV is compressed in a similar way to jpeg. Each frame is compressed as if it were a still - intrafrmae compression. Data is thrown away to make the rate manageable by the camera electronics and storage medium. The compression is lossy. Sound is left uncompressed.

DV is compressed about 8 to 1 .


HDV uses a more complicated compression technique that is similar to how MPEG4, WMV, H264, Divx, and Xvid works. The data rate is still 25 Mbit/sec but because of all the extra pixels compression is more aggressive and more lossy, but mitigating against this is that it is also more clever. Amusingly HDV sound is pants compared to uncompressed DV sound - it is 320 Kbit/ sec mp3 - shame.

HDV compression is interframe and uses something called I frames, P frames and a GOP ( nothing to do with US elections ). There are also B frames but they are simialr to P frames for the purposes of this discussion.

HDV compression relies on differences between frames. An I frame is a key frame and is a reference for the following P frames. The camera stores the difference between the I frame and the P frame. Clearly this will often be only a small amount of data as often frames of video are very similar meaning that we are throwing data away but not loosing too much from the picture - well not ususally.

HDV 1080 uses a GOP - group of pictures - of 15 - this means that one frame in 15 is a key frmae with the following 15 being I frames that contain just the difference data from the earlier I frame. HDV 720 uses a GOP of 6 and IMHO is far preferable becuase of this and will foten look BETTER than 1080.

The end result of all this cleverness is that HDV is compressed 50 to 1 - yes - 50 to 1 - aint it amazing that so much data can be lost and the picture still look so great ?!?!?

BUT - no one rides for free on the compression express....

Due to this complex structure HDV needs lots more power to edit. DV is easy to edit on a 2 gig pentium but a modern dual core is almost essential for HDV.

A glitch in the data stream, a drop out, on DV means a few lost frames. A drop out with HDV can be more serious, if you loose an I frame the following GOP, all 15 frames are very likely to be lost or damaged.

This is not a drastic problem as drop outs are rare but thay may be more catastrophic with HDV.

ACVHD often uses a lower bit rate, a longer GOP and also a variable bit rate making this format practically uneditable as it requires heaps of power to unmash it - more than HDV.

Other problems with interframe HDV / AVCHD compression are that it hates motion. The effect can be seen is an exaggerated form in video compressed for the net. Any shots of high energy fast motion will result in loss of resoloution as the huge differences in the I frame and P frames means that the 25 Mbit / sec of data gets spead very thin. Static shots look best in HDV with the picture gently degrading on fast motion. DV is interframe and does not suffer from this effect - but obviously the res is usually lower as it's a SD format.

It aint fashionalbe to say it but motion artifacting is real, but dont get too upset, only occurs when the whole frame is in motion - Motion Artifacts

Lots on DV - DV
Lots on HDV - HDV
Compression Guff - Video_compression
Bit rates source - Blackmagic Design: Support Detail

The Kinetta - kinetta: by filmmakers, for filmmakers

You data whore.

All that technical jiggery pokery goes right over me. I work on this - if it looks good, smells good and tastes good, it's probably good enough for me.

Sorry if that sounded dismissive Mark. I'm sure many people with paw over this and wet themselves with data rate excitement. Well done for putting it out here.

__________________
Crane operators use powered heads....

Get a crane on your shoot...

Yea - well I do get excited by such things but as you say if it looks right it is right.
I just wanted to clear up some common misconceptions - very much the fact that your camera at home is producing very compressed video and that what we buy in the shops is right at the bottom of the performance envalope.

It does that admirably well.

__________________
Crane operators use powered heads....

Get a crane on your shoot...

Good article. I think when most of us refer to saving uncompressed video we mean uncompressed more (than when it comes off the camera).

Yes - I think you are right - BUT - everyone should be clear that you gain nothing by rendering to uncompressed, you cant invent data.

All you get is a full drive.

True! ANd you end up with a full hardrive real quick. Actually I have two capture cards that I use with one of cameras (analogue) and for capturing from VHS. One will let me capture in many formats,including uncompressed, while the other uses hardware compression) direct to mpeg-2. I find I use the direct to mpeg-2 all the time now because:
a) At a high bitrate I don't see that much of a difference.
b) It's more reliable encoding because it's taking the load off your entire system.
c) it saves time and hardrive space.
d) I'm going to encode everything for DVD in the end anyway. Might as well do it right at the start.

This might be a good time to throw in a reminder that not all encoders are created equal. That is to say, the codec you're using in one application may not have as good an output as another. They vary. Some handle motion beter than others, some handle transitional effects better than others,etc. Mpeg-2 I believe struggles with slow motion - period. The quality of your original clip has to be very good to survive the encoding process.

And you are absolutely correct - you can't invent data.

So true about codecs.

I usually use the panasonic DV codec but a few moths back I tried cineform DV - it was suprisingly poor - 5 times slower and did odd things to the picture. Swapped DV artifacts for noise.