The difference between 720p hdv and dvcPro HD???

[Huiy Tang]

What's the difference between 720p HDV and panasonics 720p DVCPro HD?

[David Newman]

At 720p24 the JVC has less compression and higher luma resolution -- one in favor of JVC, although HXV200 has a high chroma res in the vertical direction. The HXV200 can compress to 720p60 whereas the JVC can only do 720p30 -- one in favor of HXV200. I say "compress to" as both camera can do 720p60 if you capture into a PC solution like Prospect HD, Wafian HR-1, and of course some non-CineForm products also, but most will not do this. The two cameras are very different, particular projects and shooting styles will favor one over the other.

[Guy Barwood]

I don't know about DVCProHD being more compressed than HDV at 24p. DVCProHD is well known for having less compression than HDV.

DVCProHD has less pixels (960x720 vs 1280x720) but each pixel is more accurate (less compression).

I did a comparison of a still 6MP image conversion to a frame of HDV and a frame of DVCProHD and DVCProHD won conclusively to me, In video however, the compression articacts that HDV 720p does have are not neally as obvious as in a still frame capture.

If you want to see the images for yourself look at
http://www.glasseye.com.au/hdv_test

Be warned though the page is a few MB to load and it isn't hanging off the fastest connection.

[David Newman]

"DVCProHD is well known for having less compression than HDV." -- this is only in internet forum world where nothing is "well known". People only assume that this is the case because 40Mb/s seems to be a bigger number than 19Mb/s. I would chose a good 19Mb/s MPEG2 stream over 40Mb/s I-frame DCT for most footage types. I don't think your tests reflects what the cameras are doing, as there are so many controls for MPEG, one implemention of MPEG can't readily be used to simulate another. A common MPEG sequence a 'B' frame is typically half the size of 'P' which is half that of an 'I' (conservative ratios based on real-time world data from the bit stream of the JVC cameras.) In the JVC's 6 frame GOP (IBBPBB) the average bitrate for an I frame is can be determined by substituing the the ratios of I to Bs and Ps to get

I + I/4 + I/4 + I/2 + I/4 + I/4 == 19Mb/s divided by 4 (4 GOPs per 24fps)
2.5I = 4.75Mb per GOP.
I = 1.9Mbits per frame (normalized to I frames.)

Therefore over the 24 frames the MPEG2 sequence has an effective bit-rate of 45.6Mb/s, which is higher than 40Mb/s. The MPEG2 I-frames and the DVCPRO-HD frames are equivalent in compression efficiency. For scenes with less motion the MPEG2 results get even better, with a lot of motion the DVCPRO-HD results are better. So there is no hard rule, other than in most dramatic shoots (we are talking 24p here) the JVC will more likely than not have fewer compression artifacts.

Also regarding "DVCProHD has less pixels (960x720 vs 1280x720)", this is only valid for luma data. The total number of pixel cooefficients are the same, as shown here:

960x720 4:2:2 is 960*720 + 480*720 + 480*720 = 1382400 pixels.
1280x720 4:2:0 is 1280*720 + 640*360 + 640*360 = 1382400 pixels.

Now typically chroma is compressed more, so that can favor DVCPRO-HD. Then again as JVC luma samples are closer together, improving redundancy favoring 1280 MPEG. You are probably get the idea that this a complex subject. In the end, one number being bigger than another really doesn't mean much.

[Guy Barwood]

I brought up the subject of the quality of the MPG2 compressor being an important aspect to HDV products well over 12months ago yet no one has seemed interested in this point up until your post. I know full well how the quality of compression engines can effect the quality for which is in theory the same compression codec. We all see it every time convert DV to MGP2 for DVDs...

Having said that, the real world test of mine showed such a difference quality of the two codecs that even a substantial improvment in the implementation of the HDV codec would only just start to bring the fine detail back in that MPG2 @ 19Mbps dropped while DVCProHD didn't.

Of course if you have some opossing real world evidence (real frames etc) please feel free to provide it (ie if you say it ain't so, lets see your frame captures to proove it). I'm sure I wouldn't be the only one interested. I'd love to see that HDV can capture the detail that DVCProHD captured.

I dare say, once the HVX200 is in peoples hands there will be a few good direct comparisons of the compression and quality vs the HD100. As cameras designs go, for the most part the HD100 looks the clear winner, as codecs go, if I have a choice, DVCProHD seems to me to capture the most real detail.

PS: As far as I am concerned, a pixel is a pixel, not a luma or chroma sample for example. Thats why I said DVCProHD has less pixels, which it does, but each pixel on average has more detail. What you were comparing is not pixels, but samples.

[Steve Mullen]

All this reminds me of the Panasonic - Sony fight over HDCAM vs DVCPRO.

A few years later I don't know anyone who cares.

[Guy Barwood]

That doesn't make the discussion of HDV vs DVCProHD any less relevant today.

[Barry Green]

Originally Posted by Steve Mullen

All this reminds me of the Panasonic - Sony fight over HDCAM vs DVCPRO.

A few years later I don't know anyone who cares.

Anyone who edits it on a desktop cares. With HDCAM, you can't -- there's no way to get at the data, no software codecs for the editors, etc.

[Alex Bowles]

The big differences between the formats themselves are found in the data rates (2.5MBps vs. 12.5MBps), the color sampling schemes (4:2:0 vs. 4:2:2), and the compression types (intraframe vs. interframe.) The effects of these differences all show up in post.

While higher data rates almost always mean a superior picture, the improvements may not be worth the dramatic increase in drive space needed to accommodate DVCproHD (44GB/hr vs. 9-12GB/hr. needed for HDV) or the increased render times needed for graphics and effects work done with a higher grade format.

If you do like working with layers of video, graphics and other rendered effects, a good rule of thumb is that your project drives should have triple the space needed by your raw footage, and that your drives should be mirrored if data redundancy is at all important. Needless to say, you're going to spend a LOT more on storage for a system that can accomodate several hours of DVCproHD material vs. HDV. You can limit this expense through the use of low res proxy files for offline editing, but doing so means expending more effort on media management. The time value of money should be the deciding factor here.

Superior color sampling is important if you plan to do any serious work with effects, as a compression scheme that limits data rate by sampling less color information (HDV at 4:2:0 vs. DVCpro at 4:2:2) is not going to serve you well in the post suite. This is especially true of chroma-key work, which gets much harder than it needs to be if you opt for HDV, and can also be an issue if you're needing to color correct footage from various cameras that, for whatever reason, weren't set up to match when the footage was shot, but need to match in the finished show.

Intraframe vs. Interframe is a big deal when it comes to tape dropouts. In a long GOP scheme (i.e. 1080i HDV), a single dropout will wipe out a 1/2 second of footage, and even a glitch in a shorter GOP (6 frames in 720p HDV) can get ugly, whereas a dropout in an interframe format like DVCpro can be limited to a single frame - something which can be fixed fairly easily. Of course, the less robust MiniDV tape stock used by HDV cameras is far more prone to dropouts than DVCproHD stock, so you're increasing the likleyhood of problems at the same time you're making their impact more severe. You should also be advised that frame-accurarte inserts to tape are not an option with a format like HDV which relies on intraframe compression. This is an old-school technique, but is still something to be aware of.

Beyond the differences between the formats themselves, you should also consider the mechanical differences between the cameras that support one or the other. Typically, DVCproHD cameras have bigger sensors and better optics, both of which can make a dramatic difference in picture quality - especially with regard to controlling depth-of-field. However, the cameras also tend to be bigger and more expensive, which effects everything from the amount of power they require, to the day rates on rentals (camera + monitoring) and the corresponding insurance requirements. You'll also spend a lot more on the decks used to get your footage into your edit system (for example, you can rent a JVC HD-100 camera for $250-325/day, and use the on-board deck for capture via firewire vs. $900-1200/day for a Panasonic VariCam rental + $600/day for a AJ-1200 deck to edit with.) If you're shooting a movie, DVCproHD is a much safer bet, whereas reality TV can do just fine (if not better) with the highly portable HDV cameras - several of which can be had for the price of one Varicam.

One a final final note on how to get the best picture possible, regardless of format: hire a good gaffer. Nothing will fix bad lighting, whereas great lighting can mitigate nearly any limitations presented by your chosen format. Even in this brave new world, it's still garbage in, garbage out.

Hope this helps.

Alex

[Stephen L. Noe]

Originally Posted by Alex Bowles

The big differences between the formats themselves are found in the data rates (2.5MBps vs. 12.5MBps), the color sampling schemes (4:2:0 vs. 4:2:2), and the compression types (intraframe vs. interframe.) The effects of these differences all show up in post.

You'd need to compare the uncompressed frame's, not the compressed datarate to be more accurate.

Originally Posted by Alex Bowles

While higher data rates almost always mean a superior picture, the improvements may not be worth the dramatic increase in drive space needed to accommodate DVCproHD (44GB/hr vs. 9-12GB/hr. needed for HDV) or the increased render times needed for graphics and effects work done with a higher grade format.

Once again, most people do graphics on an uncompressed timeline. Speaking for Avid Liquid, you get the choice of render/fuse codecs: MPEG2 (MP@HL) or RGB(AVI) or 2VUY (uncompressed).

Originally Posted by Alex Bowles

If you do like working with layers of video, graphics and other rendered effects, a good rule of thumb is that your project drives should have triple the space needed by your raw footage, and that your drives should be mirrored if data redundancy is at all important. Needless to say, you're going to spend a LOT more on storage for a system that can accomodate several hours of DVCproHD material vs. HDV. You can limit this expense through the use of low res proxy files for offline editing, but doing so means expending more effort on media management. The time value of money should be the deciding factor here.

I'll agree. Also note that uncompressed is uncompressed whether it's HDV or DVCProHD. The render space for FX will be approx the same when working on an uncompressed timeline.

Originally Posted by Alex Bowles

Superior color sampling is important if you plan to do any serious work with effects, as a compression scheme that limits data rate by sampling less color information (HDV at 4:2:0 vs. DVCpro at 4:2:2) is not going to serve you well in the post suite. This is especially true of chroma-key work, which gets much harder than it needs to be if you opt for HDV, and can also be an issue if you're needing to color correct footage from various cameras that, for whatever reason, weren't set up to match when the footage was shot, but need to match in the finished show.

No doubt that 422 is better keying. HD-100 is capable of 422 10bit out it's component jacks at full 1280X720. (Thanks David Newman for supplying that information). Also HDV uncompressed keys very nicely.

Originally Posted by Alex Bowles

Intraframe vs. Interframe is a big deal when it comes to tape dropouts. In a long GOP scheme (i.e. 1080i HDV), a single dropout will wipe out a 1/2 second of footage, and even a glitch in a shorter GOP (6 frames in 720p HDV) can get ugly, whereas a dropout in an interframe format like DVCpro can be limited to a single frame - something which can be fixed fairly easily. Of course, the less robust MiniDV tape stock used by HDV cameras is far more prone to dropouts than DVCproHD stock, so you're increasing the likleyhood of problems at the same time you're making their impact more severe.

Dropout's have not been a problem when sticking to the same tape stock. I have not seen any complaints about dropout's. Have you guys?

Originally Posted by Alex Bowles

You should also be advised that frame-accurarte inserts to tape are not an option with a format like HDV which relies on intraframe compression. This is an old-school technique, but is still something to be aware of.

This depends on the NLE. Liquid is a frame accurate Native HDV NLE. Also there are other solutions that transcode to intermediate codecs that are frame accurate as well.

Originally Posted by Alex Bowles

If you're shooting a movie, DVCproHD is a much safer bet, whereas reality TV can do just fine (if not better) with the highly portable HDV cameras - several of which can be had for the price of one Varicam.

I disagree and there are some features in production using the HD-100 now.

Originally Posted by Alex Bowles

One a final final note on how to get the best picture possible, regardless of format: hire a good gaffer. Nothing will fix bad lighting, whereas great lighting can mitigate nearly any limitations presented by your chosen format. Even in this brave new world, it's still garbage in, garbage out.

Always good advice to work with an expert.

[Jiri Bakala]

Originally Posted by Stephen L. Noe

I disagree and there are some features in production using the HD-100 now.

Well, it all comes down to two things: budget and suitability.

If your budget allows for 35mm film, you'd be a fool to shoot on video. No matter what anybody says, visually, film still has advantages over even the best HD available today. Most everybody (except for Rodrigues and Lucas) shoots film if they have the money for it (and all its consequences; dailies, transfers, general post, DI, etc.).

The fact that there are currently feature productions being shot on the HD100 doesn't make it the holy grail. Which brings up my other point, suitability. There are projects that are simply not suitable for film or big cameras or large crews or whatever. They are best served by a small video camera regardless of the budget.

And one more note, a good gaffer is great but on a film set it is still the DP who calls the shots and sets up the lighting style/look/whatever you might want to call it.

[John Mitchell]

Originally Posted by David Newman

"DVCProHD is well known for having less compression than HDV." -- this is only in internet forum world where nothing is "well known". People only assume that this is the case because 40Mb/s seems to be a bigger number than 19Mb/s. I would chose a good 19Mb/s MPEG2 stream over 40Mb/s I-frame DCT for most footage types. I don't think your tests reflects what the cameras are doing, as there are so many controls for MPEG, one implemention of MPEG can't readily be used to simulate another. A common MPEG sequence a 'B' frame is typically half the size of 'P' which is half that of an 'I' (conservative ratios based on real-time world data from the bit stream of the JVC cameras.) In the JVC's 6 frame GOP (IBBPBB) the average bitrate for an I frame is can be determined by substituing the the ratios of I to Bs and Ps to get

I + I/4 + I/4 + I/2 + I/4 + I/4 == 19Mb/s divided by 4 (4 GOPs per 24fps)
2.5I = 4.75Mb per GOP.
I = 1.9Mbits per frame (normalized to I frames.)

Therefore over the 24 frames the MPEG2 sequence has an effective bit-rate of 45.6Mb/s, which is higher than 40Mb/s. The MPEG2 I-frames and the DVCPRO-HD frames are equivalent in compression efficiency. For scenes with less motion the MPEG2 results get even better, with a lot of motion the DVCPRO-HD results are better. So there is no hard rule, other than in most dramatic shoots (we are talking 24p here) the JVC will more likely than not have fewer compression artifacts.

Also regarding "DVCProHD has less pixels (960x720 vs 1280x720)", this is only valid for luma data. The total number of pixel cooefficients are the same, as shown here:

960x720 4:2:2 is 960*720 + 480*720 + 480*720 = 1382400 pixels.
1280x720 4:2:0 is 1280*720 + 640*360 + 640*360 = 1382400 pixels.

Now typically chroma is compressed more, so that can favor DVCPRO-HD. Then again as JVC luma samples are closer together, improving redundancy favoring 1280 MPEG. You are probably get the idea that this a complex subject. In the end, one number being bigger than another really doesn't mean much.

David - you're obviously very uptodate on the codecs:-) I had forgotten that DVCProHD actually required less space in the 24P format - in the back of my mind I've always got that 100Mb/s figure, but I guess that's only for the 60fps flavour.

I'd be interested to know how technically Cineform Intermediate compares - obviously it's a 10bit codec so the colour precision is higher, but what is the colour sampling scheme, Mb/s etc?

[David Newman]

CineForm compression doesn't do any sub-sampling tricks, so a 1280x720 4:2:2 is encoded at 1280x720 4:2:2 whereas DVCPRO-HD encodes this as 960x720 with an effective 3:1.5:1.5 color sampling. But then CineForm compression wasn't designed with the needs for fixed tape bit-rates in mind, so we are comparing a fixed-bit-rate codec to a variable bit-rate constant quality codec. For 720p24 DVCPRO-HD is around 40Mb/s which is very low, for 10bit 720p24 CineForm we vary between 60-90Mb/s (25% lower for 8-bit.) DVCPRO-HD is designed for tape not for post production. DVCPRO-HD is a fine aquistion format, but just like with HDV, there are issues for its use in post production. The page Panasonic doesn't like : http://www.cineform.com/technology/H...lysis10bit.htm

One more note: CineForm is not limited to the two or three resolution modes that most fixed bit-rate codecs like DVCPRO-HD and Avid DNxHD seem to have. We have products that allow film resolution up to 2048x1556, we even tested up to 4096x3112, so resolution between SD and 4K are available at any frame rate you care to run. Sorry I really like this stuff. :)

[Steven White]

I always like reading your posts David.

I think the real "HDV @ 720p vs. DVCPRO-HD @ 720p" question comes down to the individual cameras. As David's breakdown of pixel count and effective bitrates deomonstrates, the codecs are virtually identical at 24p.

The real questions are: how much does the camera cost? Do the images have enough lattitude? Is the lens any good? What's the form factor like? Are there any technical issues (i.e., split screen, noise, gain) that will cause one camera to be problematic where the other one won't? etc. etc.

-Steve

[John Mitchell]

Originally Posted by David Newman

CineForm compression doesn't do any sub-sampling tricks, so a 1280x720 4:2:2 is encoded at 1280x720 4:2:2 whereas DVCPRO-HD encodes this as 960x720 with an effective 3:1.5:1.5 color sampling. But then CineForm compression wasn't designed with the needs for fixed tape bit-rates in mind, so we are comparing a fixed-bit-rate codec to a variable bit-rate constant quality codec. For 720p24 DVCPRO-HD is around 40Mb/s which is very low, for 10bit 720p24 CineForm we vary between 60-90Mb/s (25% lower for 8-bit.) DVCPRO-HD is designed for tape not for post production. DVCPRO-HD is a fine aquistion format, but just like with HDV, there are issues for its use in post production. The page Panasonic doesn't like : http://www.cineform.com/technology/H...lysis10bit.htm

One more note: CineForm is not limited to the two or three resolution modes that most fixed bit-rate codecs like DVCPRO-HD and Avid DNxHD seem to have. We have products that allow film resolution up to 2048x1556, we even tested up to 4096x3112, so resolution between SD and 4K are available at any frame rate you care to run. Sorry I really like this stuff. :)

So if I'm reading you correctly Cineform relies more on the power of the processor for the max resolution etc?