[This article was first published in the April, 2006, issue of
Larry's Final Cut Pro Newsletter. Click here to subscribe.
Updated June, 2007; August, 2007; January, 2008; March, 2008; April, 2008; June, 2008.]

Update - June 2007

With the release of Final Cut Studio 2, Apple's ProRes codec, and my continuing education about HDV, I realized I needed to modify this article.

HDV has a very definite place for some productions, especially when you need low-cost HD images which won't require a lot of color correction, compositing or projection to large image sizes.

HDV cameras are continuing to fall in price, which makes them a great way to experiment with 16:9 framing and high-definition images without breaking the bank. HDV can create some very cool looking pictures.

For a relatively small amount of money, you can shoot an HD picture. You also have a variety of frame rates to select from, including 23.98, 25, and 29.97. Some cameras offer additional rates beyond these three. Plus, as mentioned earlier, the default aspect ratio for HDV is 16:9.

However, for me, HDV's long conform times, lack of color information, inability to support chroma-keying and incompatibility between camera manufacturers reduce HDV's attractiveness.

That's why the release of Final Cut Studio 2 is significant. With the advent of the Open Format Timeline, we can now easily combine HDV images from different camera manufacturers. As well, we can transcode HDV to ProRes (with a significant increase in file size) which removes time spent conforming as well as improving general overall color handling. Finally, we can select ProRes for rendering, which reduces render times with no apparent image quality loss.

While HDV still has limitations, as outlined below, HDV in conjunction with Final Cut Studio 2 can be very attractive for many productions.

HDV Weaknesses

However, HDV does have weaknesses:

• This compression groups several pictures into a "group of pictures," called a GOP, rather than each picture being it's own entity. 1080i HD groups 15 images into one GOP. 720p groups 6 images into one GOP.
• This GOP method of compression means that HDV is not accurate for timecode or frames when capturing or outputting. (Editing HDV inside FCP is frame-accurate, however.)
• HDV uses extreme color sampling, resulting in very poor color keying, color correction, or compositing results. (HDV uses 4:2:0 color sampling, the same as a DVD. Here's an article that explains it in more detail.)
• HDV is enormously compressed, creating the possibility of significant motion artifacts when the camera is moved, or zoomed, quickly. HDV uses MPEG-2 compression, the same as a DVD.
• Because of this compression, and the overall image size, HDV renders take significantly longer than DV. (Render speeds are directly dependent upon the speed of your computer.)
• HDV needs to be conformed, or rebuilt, into a consistent GOP structure before it can be output to tape or exported to a file. This conforming can take an exceedingly long time. (For instance, conforming a 30 second sequence consisting of five shots, took over 10 minutes on my PowerBook. Conforming a complex hour-long sequence could take several hours on a G-5.)
• HDV can only use Print to Tape, not Edit to Tape, due to needing to conform the image prior to output.

One lesser limitation is:

• The HDV image is 1440 x 1080, which does not precisely match either the 720p or 1080i format (though I've discovered that several HD formats use non-square pixel aspect ratios).

For all these reasons, we need to select HDV as we would any other video format -- when it is the best choice for the job. Additionally, we also need to separate how we CAPTURE the image from how we EDIT the image.

Using HDV for capture

HDV is an inexpensive way to capture HD images, provided you are not doing any chroma-key or color work. If so, HDV is the wrong format to use because the compression process removes almost all the color information from an image. And once the image is recorded to tape, the color can not be rebuilt into being better than HDV.

If color or motion quality are important, you are far better off shooting in DVCPro-HD. This codec, which takes about 4 times the disk space of HDV, uses significantly better color sampling, compression and image quality.

Using HDV for editing

Based on what I've learned during these last several weeks, here are my new criteria:

• If money or disk space are more important than time or color, HDV is a good choice.
• If color or time are more important than money or disk space, HDV is not a good choice.

What I'm recommending to my clients now is that if they choose to use HDV for image acquisition (again keeping in mind the problems with color and motion artifacts), they should not capture HDV into Final Cut via FireWire, instead they should convert from HDV to DVCPROHD or ProRes during capture.

(DVCPro-HD is one of the native formats of Panasonic's P2 camera. ProRes 422 is shipped with Apple's Final Cut Studio 2.)

This requires an HD capture card, from companies like AJA or Blackmagic Design, to do the cross-conversion on the fly during capture.

DVCPROHD and/or ProRes provide the following benefits:

• Frame-accurate timecode during capture
• Color sampling of 4:2:2 (at least 4 times better than HDV)
• No time lost to conforming
• Can be edited on slower computer systems
• Support for both 720p and 1080i images
• Frame-accurate edits when laying back to tape

Capturing an HDV signal to DVCPROHD or ProRes requires:

• A camera that shoots DVCPROHD natively, such as the Panasonic P2, or...
• An HD-SDI connector on your camera or deck, such as the Canon XL H1, or...
• HD component video out of your deck or camera, running into an HD SDI converter from AJA, Miranda or others.

The idea is to take the HDV signal, convert it to HDSDI, then use your capture card to convert it into the DVCPro-HD format for editing.

Here's another interesting idea. The Canon XL H1 has an HDSDI port on the camera. If you record to a hard disk, such as the FireStore, directly from this port, you can record in a much less compressed format that HDV itself. Plus, you bypass the entire capture process because your file is already digital on the hard disk.

This means you can use an HDV camera for image acquisition, yet record image quality far superior to HDV.

Now, using either the P2 or Canon XL H1 in this way opens up the whole issue of how to deal with tapeless video. And I have major problems with all the current solutions proposed by Panasonic and others for archiving footage that did not originate on tape.

For now, I only recommend tapeless recording for those images that do not need to be archived for long periods of time. My hope is that we will see more robust solutions from a variety of vendors at NAB this year.

This whole process of shooting and editing HDV is very much a format in transition -- and archiving is only one of the major issues still in need of a better solution than exist currently.

Still, I wanted to give those of you thinking about working with HDV some options to think about as you strive to get more work done and still maintain quality.

As always, I invite your comments.

- - -

Update: This HDV article generated a lot of contrasting comments that I want to include here.

Graeme Nattress corrected me on HDV image size:

Larry, you wrote: "The HDV image is 1440 x 1080, which does not precisely match either the 720p or 1080i format."

And neither does anything else even remotely affordable. To get 1920x1080 you need to go to D5 or HDCAM SR!!! HDCAM is also 1440x1080 (as is most broadcast HD) and DVCProHD is 1280x1080.

Just think of 1440 as "anamorphic" (which it isn't really) and you'll be fine. It's a non-issue, just non-square pixels.

720p HDV is, of course 1280x720, which is what it should be, square pixels. DVCproHD is 960x1280, which again, is reduced.

Larry replies: Sigh... Wasn't HD supposed to REDUCE all this format confusion. Thanks for the correction, I'll stop worrying about image size.

Gardner Reynolds writes:

My only comment [on your HDV article] is the Z1 captures (if you know what you are doing) some smoking images in any mode and is great to have in the bag with other DV cameras. I agree it's a format in transition but at least a camera in that price range gets us started and warmed up on the HDV platform. Thanks for the good articles.

Tom Wolsky agrees:

You might want to restate it when you write: "The HDV the compression process removes almost all the color information from an image."

I don't think it's accurate to say it removes almost all the color information from an image. You make it sound like it's going to practically desaturated.  It sounds as if the color rendition of HDV is very poor, which is just not so. The color reproduction on most HDV cameras is outstanding. HDV is not, I don't think, substantially more difficult to chroma key than DV, or chroma keying DV in PAL, which uses the same color space.

Graeme Nattress provides a technical perspective:

[This color compression is] no more extreme than DV (chroma for 4 luma, just in a different pattern). Interlaced 4:2:0 is nasty, but this, as always only effects chroma resolution, not quality. You shouldn't see chroma sampling causing colour correction problems - it's the excessive compression that gets in the way here.  And remember, HDCAM is 3:1:1 with only 1 chroma for 3 luma, so hardly much better :-(

Tom continues:

You might want to explain how convert from HDV to DVCPRO HD during capture. I didn't realize that AJA and Blackmagic Design can convert to DVCPRO HD. I don't see anything on their web sites to indicate that they can capture to DVCPRO HD. They have DVCPRO HD hardware acceleration and output support, but I think they capture to uncompressed HD. Big fast RAID drives. You're not in the HDV league at all any more. This is the old use DV for acquisition and then bump to uncompressed route. Great if you have the bucks and your production has the budget. 

Larry replies: I just called tech support at AJA and confirmed that they can convert HDV to DVCPRO HD in real-time using their Kona 2 or Kona 3 capture cards.

Tom goes on:

As for exporting for DVD, I wouldn't downconvert on export, but export a self-contained native HDV file and have Compressor do the SD MPEG-2 encoding. The H.264 I'm sure works well but it's probably really, really slow as all H.264 compression is quite time consuming. Then doing another pass for MPEG-2 into DVD seems like an extra long step.

Larry replies: My point is that when you are working with HDV, you need to allow time, lots and lots of time, for conforming. Conforming must occur whether you are outputting to a QuickTime movie or to a video tape.

On the issue of whether converting from HDV to DVCPro HD is a good idea, opinions diverged. I respect Graeme too much to ignore his comments:

Graeme Nattress writes:

[Should you convert HDV to DVCPro HD?] No. Never! Why? Adding compression on top of compression is just bad. DVCproHD is way too compressed. It's full of artifacts even straight of a varicam. To add that compression on top of HDV just makes a mess.

Answer - edit HDV native - it's easy on a decent mac, and then just change the compression right at the end of editing to "uncompressed", do a final render before going out to master tape. Again, you'd never use HDV as a master - even one compression back to HDV looks awful.

Only use DVCproHD if that's what you shot, or are going back to DVCProHD tape. If you're recording live from SDI, uncompressed HD, or PhotoJPEG 75%, are much superior.

Ed Yost comments:

Up until two weeks ago I was working on the DVD documentaries for a very high profile film in which all the behind the scenes footage and most all interviews where shot in HDV. As the film progressed the DVD producer started to add NTSC footage shot on digital beta, NTSC DV, and DVCPro HD 24p. The project became an incredible mess and still is.

Unfortunately many of these producers pretend to be know-it-alls and don't really take advice when given. Sure wish they would take your seminars.

Larry replies: Thanks for all your comments. I'll update this as more come in.

Update - August, 2007

Norbert Jeger writes:

I was looking at some recently captured HDV footage and comparing the captured HDV footage from FCP with Avid Xpress Pro and Avid Media Composer.

FCP always saturates the footage with the Apple HDV codec. So that means the FCP footage is not as its on the tape.

Maybe you have some experience with this issue already? It seems that the HDV footage captured with FCP is only made for computer monitors and not for watching on TV after...

An answer from you would be helpful.

Larry replies: Norbert, I have not done a comparison of HDV footage between these three applications. I don't believe Apple is doing anything to alter the chroma levels, but there are a lot of engineers who read this that may be able to suggest a solution.

However, keep in mind that you should never judge any video by looking at your computer display.

Norbert replied:

Of course i was not only looking at my computer monitor.. also on TV. Unfortunately FCP seems to increase saturation after capturing the HDV footage. Attached find a comparison picture of FCP and Premiere/Avid.

You can see there are differences if you capture HDV footage over Firewire to FCP... I wonder if this is true if you capture through a Blackmagic Design Decklink or an AJA IO?

Larry replies: I don't know. Let's see if anyone else does.

UPDATE - Aug. 22, 2007

Tom Wolsky writes:

Not only higher saturation, but higher contrast as well, which is actually more pronounced in the image offered than the saturation. Good job Apple! The Premiere/Avid image looks flat and under chromaed. (BTW, is this Premiere/Avid on a Mac or PC?)

Justin Ascott writes:

I'm a lecturer in digital media based in London, UK. I've been using your Lynda.com tutorials for FCP 5 which have proven extremely helpful. I'm videoing with the NTSC Canon HV20 camera (HDV 1080i 24p). It's very unclear however whether I should use the 1080p24 or 1080i60 setting in Audio/Video Settings > Sequence preset. I have captured using 1080p24 which appears to work however some HDV discussion sites suggest using 1080i60. Are there benefits to using one sequence preset rather than the other?

Can I suggest you advise subscribers about this format's severe limitations.

Larry replies: Justin, we've been talking about HDV in this space for a while now.

UPDATE - Aug. 22, 2007

My general recommendation is to always capture HDV in the format in which it was shot. However, in this case, Adam Lloyd Connell supplied better information.

One thing I noticed with this month's newsletter to do with Justin Ascott's question about the Canon HV20:

I have one myself. It is not a progressive frame camera. The sensor records the video which is encoded in NTSC or PAL. The Cinema Mode and the HDV Progressive frame mode can be used to create a progressive look, ie, the look of 24/25p, but the actual signal that is recorded to the tape, and then output when you capture in FCP or playback on TV is in fact a standard 60i/50i signal that has just been de interlaced. Therefore, to capture whatever footage you have recorded with the HV20, you need to use FCPs 1080/50i or 1080/60i preset.

UPDATE - December, 2007

Iain Anderson sent his alternate view of HDV.

I heartily recommend editing HDV footage natively. A number of people have expressed the opinion that long-GOP MPEG footage is inherently hard to work with. It's not -- it's very much the same as editing with DV, and Unlimited RT works just as well. You certainly don't need to render every edit.

I recently cut a short film in HDV on a MacBook Pro, and it was a painless experience which produced high quality results. I used FCP6, which did make the process somewhat easier. As the film was part of the 48 Hour Film Project, I had less than a day for editing and delivery to mini DV tape. A shame to see all the HD quality disappear at the end, but the process worked well.

The advantages of HDV:

1. No extra hardware needed to import footage. Any machine capable of running the current version of Final Cut Pro can cut HDV. It's painless to edit, and while effects can take time to render, the vast majority of effects work with Unlimited RT just like SD footage. The only issue I had with rendering was when overlaying several streams of video with feathered cropping.

2. No transcoding needed. You can't get better quality by recompressing. In fact, you can make it much worse. DVCPRO HD, for NTSC frame rates, uses a different frame size (1280x1080) to HDV (1440x1080). If you have to transcode, ProRes 422 or AIC are better choices.

3. Low data rate. It's HD with the DV space/data rate requirements of ~13GB an hour, while transcoding at least triples your data rate and space requirements. We've just about hit the point where you'll be able to keep all the HDV you ever shoot permanently online by buying a cheap hard drive every few years.

4. Easier media management. When you capture HDV natively, footage is automatically split into separate files based on when you started each recording session. No more huge files from students capturing files in hour-long blocks and easy deletion of bad footage.

5. Near-instant HD screeners. HDV footage needs no recompression to go onto an HDDVD, and every current Mac can play HD-DVD's made with DVD Studio Pro burnt to regular DVD discs. This is incredibly powerful; it's about the only way to deliver HD footage without expensive hardware. You will need to conform HDV footage to do this, however, bringing us to...

The drawbacks of HDV:

1. The 4:2:0 colors. Not ideal, but the same as PAL DV. If you've shot in HDV, the damage has already been done, but it's really not that bad. The latest Sony XDCAM EX uses the same colour space. This doesn't have to limit the quality of your colour correction, though -- see 3 below.

2. Long conform times when finished. This is an issue only when you've finished your edit and need to export back to HDV. The conforming process takes about as long as a slow render, say, a few times real time. If you're not finishing to HDV, not a problem. While you're editing, not a problem.

3. Recompressing to HDV can introduce artifacts and is slow. True if you're colour correcting or applying other filters, but there's a fantastic solution built in to FCP 6. Choose Sequence Settings, and look under the Render Control tab. Change the render codec to Apple ProRes 422. Specifically for HDV and XDCAM HD/EX, this feature renders effects to ProRes instead of HDV. No quality loss, no extra time spent recompressing during editing.

HDV isn't perfect, but I'm constantly amazed by the quality I get out of my HV 20 for the price and how easy it is to deal with. It's possible that the poor opinion of HDV PQ in the industry is partly due to the PAL/NTSC divide. Here in Australia, we only have to record 25 frames per second at the same data rate that NTSC models have to record 30 frames. Those 20% extra frames per second might be pushing the compression just a bit too far.

Larry replies: Thanks, Iain, for allowing me to share this.

UPDATE - Jan. 2008

Here's my 2-cents, says Iain Anderson:

The Canon HV20 *does* capture progressive footage, but it stores its footage in an interlaced stream. The NTSC model captures 24P but stores it as a 60i stream with 3:2 pulldown that should be removed before editing. (This is tedious and time consuming -- get the PAL version if you plan to do a lot of progressive work, especially if you're in a PAL country!) The PAL model captures 25P stored as a 50i stream with no pulldown that can be edited directly in FCP. You can capture with the 1080p25 Capture Preset, but footage is imported as 1080i50 clips. If you want to edit true progressive footage with FCP you'll need to change field dominance of captured clips to "None" before editing them into a progressive sequence. More details here:

http://funwithstuff.com/therules/2007/10/hv20-pal-1080p25-workflow-solution.html

Some sample images from HV20 progressive footage, shot at home here in Australia. Final Cut seems to have issues exporting high quality still images, so these have been copied and pasted from QuickTime Player into Photoshop, resized from 1440 to 1920 wide, then saved as 100% quality JPEG. Gamma shifts may have occurred, and the limitations of the 4:2:0 colour space are evident around the apple the possum is eating.

http://funwithstuff.com/pics/kookaburra1.jpg
http://funwithstuff.com/pics/kookaburra2.jpg
http://funwithstuff.com/pics/possum1.jpg

Larry replies: Iain, thanks for sending this in.

UPDATE - March, 2008

With the release of Final Cut Pro 6.0.2, we can now transcode HDV directly into ProRes 422, provided your computer is fast enough.

Plus, we can change render settings (Sequence > Settings > Render tab) so that both HDV and XDCAM footage renders in ProRes.

After experimenting with both of these settings, I recommend them both. HDV renders 35% faster when rendering to ProRes and editing in ProRes is faster, more accurate, and higher quality than editing HDV natively.

If you have not yet experimented with these two techniques, I suggest you give them a try.

HDV has been a hot topic in several issues of this newsletter and, recently, Professor Uli Plank, of Hannover Polytechnical University, in Germany, wrote to clarify some of my statements:

In writing on HDV-editing you are using the expression "long GOP" (like many other authors do). But may I ask what a short GOP would be? 6 vs. 15 frames? After all, everything you wrote applies to JVC's HDV recordings with shorter GOPs just as well.

So, I'm not sure if this expression isn't more confusing than enlightening to our readers. After all, in intra-frame compression there is no "group of pictures", not even a short one. GOPs only exist in inter-frame compression, and they can have different lengths.

I'd suggest using intra-frame vs. inter-frame to keep the compression schemes apart.

Regarding the precision of batch capturing HDV: While it's true that this is a critical mission, FCP does a remarkable job on it when your drive is in good shape and you don't use tapes more than once. From my tests with burnt-in timecode I can say that FCP was always dead on regarding the first frame, only the tail was longer at times, but that wouldn't cause serious problems.

Further down you wrote:

"Additionally, HDV supports only two audio channels."
I'm sorry, but that's not right, Canon for example has models with 4 channels of audio. But this complicates things even further, since other cameras and recorders don't read it and FCP doesn't support it in any straightforward fashion.

And then you wrote:

"If you are creating a project that will end up on DVD, shoot HDV, then down-convert to DV during capture (out of the camera)."

From my years of experience with DVD, I wouldn't recommend this for two reasons:

• As our technical editor already stated, most camcorders don't do a great job on down-converting. FCP does it better and MPEG Streamclip is excellent at it if you don't have access to a Snell & Willcox [hardware converter].
• Using DV between HDV and DVD is giving quality away, in particular with DV in NTSC with it's 4:1:1 color sub-sampling, since DVD is using 4:2:0 and the change in sub-sampling makes things even worse. Ingesting HDV as ProRes or DVCProHD is much better than this, you can always down-convert in I/O hardware or later in software and it yields visibly better results on the final DVD. Working in DV50 (aka DVCPro-50) would be the better alternative if one has FCP 5 only.

For monitoring HD, if one doesn't even have the dough for a Matrox MXO but owns a decent HD flat-screen, Blackmagic's Intensity Pro is a viable alternative.

Larry replies: Uli, thanks for writing. Sorry it took me so long to respond.

I agree with most of your comments as I continue to learn more about this format. HDV, XDCAM HD, and AVC HD all use compression based on a group of pictures (GOP), and different recording frame rates work better at different GOP settings. So the use of inter-frame (using a GOP) and intra-frame, make sense to me.

When I wrote this article, I believed that HDV cameras did an acceptable job of converting HDV to DV. I've since learned that this is no longer the case. Most HDV cameras do a very poor job of down-converting.

Consequently, I've changed my recommendations to:

• If you are working in an exclusively HDV project, then shoot HDV, edit HDV natively, output HDV, then convert it into the format you need for final distribution.
• If you are integrating HDV with other HD formats, convert it into ProRes 422 during capture and standardize your editing on ProRes for editing and output.

As always, I enjoy hearing other opinions and will share them in future issues.

More Thoughts on HDV

Recently, Gary Adcock of Studio37 in Chicago and I got into a discussion about HD and HDV that I thought you'd like to read. And, let me state at the beginning that I totally defer to Gary's knowledge about HD. Having said that, here's the thread.

A reader asked Gary:

We are getting into HD, specifically, we are developing a series for National Geographic which we plan to shoot on HDV.

1. What is the best way to capture the material? HDV native? Apple ProRes?
2. What would be the best way to transfer the material to post-production facility that has the xdcam tape?
3. How to I make the sound? we need to deliver stereo and Dolby-e. Does FCP 2 do dolby-e?
4. Since NG here broadcasts from server, can I deliver the movies as compressed data, without using Tapes?
5. Do I need more hardware (We have no capture cards in our studio, we use firewire only, but I have access to Blackmagic or AJA cards.

Gary replied:

• It is my understanding that Nat Geo will not accept anymore than 10 minutes per hour of HDV content in any HD show
• No station or network I have ever heard of accepts XDCAM discs or media for a deliverable - HDCam SR is THE universally supported delivery format.
• Nat Geo HD does not accept files for deliverables - Tapes only.
• Dolby-E files need to be created in a Dolby sound suite- and cannot be handled by FCP in any way, shape or form.
• Yes you will need more hardware- Kona cards are the accepted standard since AJA hardware is now found in Discreet, Avid, and FCP systems

Larry then wrote to Gary:

From my understanding, HDV is acceptable as source media provided it is converted thru a Snell & Wilcox converter into a standard HD format, such as HDCam SR. So, while conversion to ProRes for editing is preferable, it may be possible for him to edit natively in HDV, provided he budgets for a hardware conversion process prior to final delivery on tape.

Gary replied:

It is my understanding that even with Snell and Wilcox's Alchemist or Ukon systems or the more widely known Terranex, many of the cable networks are still not willing to allow more than a limited amount of HDV content per show per hour. I have not seen a delivery guide that lists more than a few minutes for HDV content at all, while some even limit P2 and XDCAM HD for the same reason (ie: compression ratio of master acquisition).

Shows like Baghdad ER for HBO were completely shot in HDV, however the workflow was established as properly converted and edited as uncompressed prior to the delivery of the show, all camera masters were converted via Terranex at the same time as being cloned to HDcam tape. HBO, Discovery, Nat Geo and others still do not qualify the conversion process -- only the camera originals and all of them do limit HDV content (currently that is still less the 25% of the total viewing time's content) for shows being delivered.

None of the networks I have mentioned allow for camera native editing of HDV or DV content in their original state according to their "written" spec.

Larry adds: Thanks, Gary, for sharing your knowledge. Your comments just reinforce how CRITICAL it is to look at the entire workflow and make sure that what you are shooting can be edited and output in a format acceptable to your distributor.

UPDATE - April 8, 2008

Gary Childress adds:

I noticed in the newsletter a section on HDV and broadcast delivery formats. HDCam SR is NOT a common delivery format for final masters. It's possible that it has been adopted for some uses, but regular HDCam and Panasonic D5 are the common delivery formats for television, not SR.

Larry writes: Thanks Gary. I didn't know this.

Larry Jordan is a post-production consultant and an Apple-Certified Trainer in Digital Media with over 25 years experience as producer, director and editor with network, local and corporate credits. Based in Los Angeles, he's a member of both the Directors Guild of America and the Producers Guild of America.

The information in this article is believed to be accurate at the time of publication. However, the author assumes no liability in case things go wrong. Please use your best judgment in applying these suggestions.

The opinions expressed in this article are solely those of the author. This newsletter has not been reviewed or sanctioned by Apple or any other third party. All trademarks are the property of their respective owners and are mentioned here for editorial purposes only.

Links to my website home page or this article are welcome and don't require prior permission.