views:

5831

answers:

6

I'm looking for an order of magnitude estimate for expected on-disk file size for 1 hour of H.264 encoded HD video transcoded from HDV (HD on a MiniDV tape). I want to archive approximately 100 hours of such content and want to figure out whether I'm looking at a big hard drive, a multi-drive unit like a Drobo, or an enterprise-level storage system.

Edit

To clarify from several good comments, I am developing the software that will use this archived video and will also manage the transcoding from HDV to compressed format (using QuickTime on OS X). This question is for gathering hardware requirements. The video is from HDV and the transcoding does not have to happen in real time. I will not be streaming the video, so I'm looking for compression that maintains as much information as possible form the HDV content. The original HDV video is interlaced, but I can deinterlace it if that would help either file size or quality of the H264 compressed output.

+2  A: 

Around 4gb/hr is quite common.

krosenvold
+2  A: 

For a good quality x264 encoding of 1060i, done by a computer, not a mobile device, not in real time, you could use a bitrate at about 5 MBps. That means 2250 MB/hour of encoded material. Recommend you deinterlace the footage and compress as progressive.

evilpenguin
1060i? I think you've combined 1080i60 (1920x1080 pixels, 30fps interlaced, giving 60 fields per second).. although I couldn't agree more with "deinterlace the footage" - interlacing must die..
dbr
Oh, just noticed the question mentioned 1060i first!
dbr
If the footage is deinterlaced, would the filesize decrease or is deinterlacing just for quality of the image?
Barry Wark
Depends on the framerate. If the framerate is the same, interlaced footage displays half the info that progressive footage does. Interlacing works by refreshing odd and even lines alternatively, that's why most interlaced cameras that produce good quality have twice the fps of progressive ones.
evilpenguin
But interlacing is a result of the way old TV CRTs display information. On a computer display, be it LCD, plasma or modern CRT, interlaced footage looks really jagged and bad. Most media players like VLC deinterlace automatically at playback if they detect interlaced footage.
evilpenguin
Bottom line: If you want your videos not to look bad, deinterlace. If they take up to much space, change the quality, change the resolution (720p is sufficient, it well compressed), but stick to progressive footage no matter what. Interlaced footage is not worth it and is really just a relic.
evilpenguin
And another thing: most codecs behave better with progressive footage, because they evaluate shapes and how they change over time. These evaluations are made by comparing neighboring pixels in matrixes. Interlaced footage really screws the whole algorithm.
evilpenguin
+3  A: 

It is whatever size you want it to be, the only thing that changes is quality. If you intend it to be played back on a non-PC device (or a slow PC), you may need to respect a certain profile (standardized set of compression settings that ensure a fixed device can play back the content).

You can see the main H.264 profiles at Wikipedia

While it is highly subjective (and highly dependent on the content being compressed), it is claimed that H.264 can achieve the same quality as DVD MPEG2 using half the bitrate.

David
+3  A: 

It would be a couple of gigs per hour.

MPEG-4 (of which H.264 is a sub-part) define high quality as around 4Mbps. which would be 1.8GB per hour.

This can vary depending on the type of video and the type of compression used.

Lehane
+5  A: 

You need to know the bitrate.. Then it's simply bitrate (bits per second) multiplied by number of seconds..

HDV is 25Mb/s, so non-transcoded it would be (I think)..

25 mbits * 3600 sec = 3.125 MB * 3600 sec = 11250 MB*hr ~ 11 GB per hour

The same applies with h.264 footage, although the above might not be as accurate (being variable bitrate and such)

I want to archive approximately 100 hours of such content and want to figure out whether I'm looking at a big hard drive, a multi-drive unit like a Drobo, or an enterprise-level storage system.

First, do not buy an "enterprise-level" storage system (you almost certainly don't need things like hot-swap drives and the same level of support - given the costs)..

I would suggest buying two big drives.. One would be your main drive, another in a USB enclosure, and would be connected daily and mirror the primary system (as a backup)

Drives are incredibly cheap, using the above calculation of ~11 GB/hour, that's only 1.1TB of data (for 100 hours, uncompressed).. and you can buy 2TB drives now..

Drobo, or a machine with a few drives and software RAID is an option, but a single large drive plus backups would be simpler.

Storage is almost a non-issue now, but encode time can still be an issue. Encoding h.264 is very resource-intensive. On a quad-core ~2.5GHz Xeon, I think I got around 60fps encoding standard-def (DVD) to h.264 (compared to around 300fps with mpeg4). I suppose that's only about 50 hours, but it's something worth considering. Also, assuming the HDV is on tapes, it's a 1:1 capture time, so that's 150 hours of straight processing, never mind things like changing tapes, entering metadata, and general delays (sleep) and errors ("opps, wrong tape")

dbr
9000 megabits gives 1125 megabytes.
joelr
Opps, thanks joelr/Barry
dbr
thanks, the HDV information is where I should've started.
Barry Wark
The math is a bit wrong. 25 mbits * 3600 sec = 3.125 MB * 3600 sec = 11250 MB*hr ~ 11 GB per hour. Uncompressed, as you say.
evilpenguin
Another small tip: if you're buying a machine for processing video, go for Intel Core i7. I've compared a PC with i7 to a Mac with quad-core Xeon and the i7 is definitely faster. And I think it's cheaper.
evilpenguin
Note: HDV is compressed, but it's compressed in a way that introduces only mild visual artifacts and allows for efficient and accurate seeking. 100h of uncompressed 1080p30 video would require 1920*1080*3channels*30fps*3600s/hr*100hr ~ 61TB.
Mr Fooz
A: 

hi,

I'm a friend of keeping the original files, so that you can still use the archived original ones and do new encodes from these fresh ones when the old transcodes are out of date. eg. migrating them from previously transocded mpeg2-hd to mpeg4-hd (and perhaps from mpeg4-hd to its successor in sometime). but all of these should be done from the original. any compression step will followed by a loss of quality. it will need some time to redo this again, but in my opinion it's worth the effort.

so, if you want to keep the originals, you can use the running time in seconds of you tapes times the maximum datarate of hdv (constants 27mbit/s I think) to get your needed storage capacity

Andreas Hornig