BBC (UK) View on Z1 and HDV

[Mark Grant]

So what exactly is 'wrong' or lacking with the in-camera down conversion?

For one thing, you're going from 4:2:0 HDV to 4:1:1 or 4:2:0 DV. If you edit HDV, then downconvert, it should give you a similar quality to 4:2:2 SD footage at the final downconvert since the HDV footage has about twice the color resolution of DV footage.

Also, the editing system probably has far more power than the chip that's downconverting to DV in real-time in the camera, so it should be able to produce a better result.

I'm not convinced that there's a huge difference, but editing HDV and then downconverting to DVD does look a bit better to me than editing downconverted DV and then compressing to DVD.

[Steven Gotz]

At present however the HDV format is not compatible with most popular editing systems like Avid.

Avid is about the only one that can not deal with it. That is what I meant by behind the times. That and the fact that the article was written before the release of the Z1.

As revealed by our tests, shooting in HDV mode and down converting for SD delivery can produce higher quality pictures, but only if the down-conversion is achieved with more expensive downconversion devices.

This is partially what I was disagreeing with. I think that Premiere Pro does a decent job, and Spot says that Vegas does an even better job. They are not what I consider to be "more expensive downconversion devices".

But then again, I had no complaints the one time I downconverted in camera. But since I only did it once, and that was before I had Aspect HD or even Premiere Pro 1.5.1, I don't particularly care about the in camera solution.

Your response indicated that you agreed with the BBC and I can not say that I do. But I would never shoot DV with my HDV camera for any reason that I can think of at this time except for one. Perhaps if I was a second shooter for someone with a DV camera and I needed to turn my tape over to them. Other than that?

[Thomas Smet]

When the camera does the down convert to SD I'm sure there is no pixel sampling filtering going on. This might give you not as good results. I actually wrote a filter that gives me almost perfect 4:4:4 RGB SD images from the camera. So far I have been only able to test it on still images from the internet but it works great. I don't think we will ever see a better option for SD work. I can do the same thing with 1280 x 720 but it just isn't as good.

[Steven White]

Yeah Thomas, similar experiences here, though I did it all with After Effects, rendering the 1080i to 480p60. The results were astonishing! I've since compressed it to DV, and it looks awful in comparison, likewise to DVD... But gosh - the uncompressed 480p (even 480i) from HDV is gorgeous stuff.

-Steve

[Robert Young]

Steven
I'm a little confused as to what you are recommending.
Clearly, one should acquire in HDV, then--
1) downconvert in camera, capturing SD, and edit in SD, or
2) capture the m2t to system, convert to SD using Premiere Pro (will PPro do that???) and edit in SD, or
3) capture m2t using Cineform Aspect, edit with PPro 1.51 in Cineform IC, output to SD or DVD, etc.
Which workflow, in your opinion, gives the "best looking" SD final product?
I haven't taken the HDV leap yet and I want to be sure it's really worth it. All of my delivery is SD DVD. I do not want to go thru all the HDV upgrade expense, workflow hassel, etc. to end up with DVDs that look like they could have been shot on my PD 170.
Thanks
Bob

[Steven Gotz]

Robert,

4. Capture in Premiere Pro 1.5.1 (or on weaker systems, use HDLink to go to M2T and then CFHD AVI in two steps).

Edit HDV all the way. Produce a HD WMV. Play it on the Linkplayer2. Enjoy it a few times, and then realize you must downcovert for old Uncle Andy. Sigh. That's OK, at least your wife and her cousin Susie saw it in the original. Sigh again.

Rather than just export to MPEG2-DVD, I bring the finished CFHD AVI into a DV project and then Pan/Scan to get the best portion into the 4:3 frame that I can. Why not letterbox? Good question. It's a lot easier, but Uncle Andy has bad eyes, and the bigger the image on his 19 inch TV the better (his kids are too cheap to even get him a 27 inch). Why not just center cut like the Z1 will do for you? Because I am a lousy cinematogrpher and didn't plan that far ahead. Not all of the best stuff is in the middle of the frame.

Here is one reason to work like that. I shot a school choir. Little kids. I got three takes and there was no way to use additional cameras because I was not told far enough in advance. I needed to downconvert for their televisions at school. So I used the 16:9 wide shot for the entire song, moving it up to the top of the frame to make room for the lyrics at the bottom. I put a black matte behind the text for reasons that will become clear if you try this. I made the first version of the video just that way. Then I made a second version. This time, I zoomed in and moved around to show closups of each kid in each row. The next video I did the same thing, but panned from the other direction. There were a few weeks left of school, they play it on Fridays, and each one for the rest of the year was slightly different. It completely confused the kids as to how I did that. If you haven't tried it, messing with the heads of elementary school kids is great sport if you do it carefully.

The point is that if you edit a very large frame in a small framed project, you have many, many choices that you would not have if you downconverted in camera.

How many times have you wanted to zoom in on a shot and couldn't without pixelization. Well now you can. Over 200% zoom and you are just getting back to the original frame resolution.

[Robert Young]

Steven
Doing the pan n scan while still in the large format sounds like a great technique to get more miles out of your footage. Definately an advantage of the HDV workflow.
The entire subject of {Z1 HDV>SD DVD out vs. PD 170>SD DVD out: is there really a difference?!?} seems to be very controversial. I have read so many conflicting opinions on the 4 different HDV forums I follow. The BBC paper in this thread seemed to conclude that Z1 SD looks better than PD 170 SD in 16:9. I can buy that- Z1 chips are a little bigger. But BBC doesn't address this other issue at all. People seem to have many technical reasons for concluding that one is better/ worse/no difference. i.e. Z1 HDV>CFHD >DVD is 4:2:0> 4:2:2>4:2:0= gotta be better; and other arguments are to the contrary conclusion.
I'm really interested it the basic eyeball test. Since it all ends up as 720 x 480, the question is: does the Z1 HDV workflow result in SD DVD that has noticably more color depth, dynamic range, or some other intangable benefit that causes the audience to look at that 42" plasma screen and scream "Whoa...this is great!!! This is MUCH better than your last one... we want more, more, MORE like this!!!" , or not?
I would really welcome any first hand eyeball opinions.
Thanx
Bob

[Graeme Nattress]

If you convert, in camera, down to DV from an HDV shoot, then that could look slightly worse than shooting DV as the video has gone through two compressions. However, from what I'm told, if there is a difference, you can't see it.

The BBC were not downconverting to DV, but on very high end broadcast gear to Digital Betacam, and hence were not subjecting the video to lots of compression in the conversion process.

If you downconvert to DV in software, then you're loosing the benefit also. What you should be doing is downconverting to uncompressed SD (or very lightly compressed SD) and that should produce visibly superior results, in terms of chroma sampling etc.

Because you've downsampled, this should make the image sharper and less noisey. This can only do good things for going to DVD. Whether, on DVD you can see the benefits of higher chroma resolution though, is debateable, but you should see benefits over NTSC DV more than PAL DV due to the clash between NTSC DV 4:1:1 and DVD / PAL DV 4:2:0

Graeme

[Steven Gotz]

Robert,

When you say it all ends up as 720X480 I cringe. Why would I not just shoot with a higher quality SD camera if that is all I needed? My goodness, there are certainly better tools available. I could pick up relatively inexpensive field monitors, use less expensive storage. etc. And maybe get better low light capabilities. And better DOP from bigger chips?

The 42" monitor? Garbage, in most cases. I want them to look at the 60" Plasma or LCD or DLP and be shocked! To heck with SD. I am distributing HDV whenever possible. But, yes, even I must downconvert for Uncle Andy. But not for real customers, thank you. Real customers get a Linkplayer as part of the contract. There must be a dozen of them throughout the training center I produce stuff for on my "real" job.

I imagine that there are tests showing that the PD170 is a superior DV camera. And there are probably tests showing downconverted HDV to be better. My guess is that it depends on a lot of factors. I, however, have made my bed with HDV, and I plan to lay in it.

Graeme is a lot more qualified to discuss some of these conversion issues, as are many others. I can only say that I am happy with my choices.

[Graeme Nattress]

I don't think that using the camera downconvert to DV is going to produce a bad DV picture by any means, but I don't think it's extracting the maximum SD quality that the camera is capable of.

As for PD170 v FX1 / Z1. In 16:9 the HDV camera is an obvious winner as 16:9 on the PD170 is rather poor. For 4:3 SD, the difference might be more subtle, and I'd expect the only difference to be that the PD170 has a little better low light. Overall picture-wise, the more modern HDV should again look better.

Graeme

[Douglas Spotted Eagle]

Having done conversions with the Z1 via several routes, I can easily express that the software route is by far, far, far better than any other routine.
I have:
1. Converted component out to SD/SDI resampled to DV, subsequent MPEG SD
2. Converted component out to HD/SDI downsampled in software, subsequent MPEG SD
3. Converted HDV 25Mbps to DV in software, , subsequent MPEG SD
4. Converted from cam to DV, , subsequent MPEG SD
5. Converted Component out to SD in 3rd party hardware box, brought in as DV to editing application.

My conclusions:
1. Sony Vegas converts better than anything I've tried.
2. Premiere converts nearly as good
3. Final Cut can't manage the m2t, but it doesn't convert the HD/SDI ingest nearly as well as Vegas or Premiere. Interestingly enough, iMovie does a better conversion than FCP does (at this time)
4. AVID doesn't convert (at this time) as well as either Premiere or Vegas.
5. In EVERY single case with these applications, the camera down convert was superior to what the software or hardware with component in could do, excepting Vegas or Premiere.

These tests are all based on naked eye viewing a variety of image types. I'm no where near as technically astute as Graeme is, my math blows. But I know what my eye sees, and frankly at the end of the day, that's all that matters. But of course, that's just my opinion.

[Graeme Nattress]

Douglas, have you tried doing any downconverts in software from HDV to uncompressed SD? I think that would look even better still for going to DVD.

The issue with FCP is that it's scaling engine is rather poor, being bilinear rather than the bicubic it should be. However, it's possible to decent downconversion if you add a little gaussian blur to the image before it gets scaled down, as that, in effect, is forcing the bilinear to function more like bicubic, taking sample points from beyond just the immediate pixels in question. If you don't do this, you get really nasty aliassing artifacts in FCP as there is effectively no filtering going on in the downconversion. Good filtering is necessary for good results, as this is part of the averaging process that reduces noise and aliassing.

Graeme

[Douglas Spotted Eagle]

Oops! That's what #1 is supposed to be.
Yes, I have. Of course 4:2:2 SD is better than anything when going to MPEG from there. I've done this with 2 different encoders, Cinemacraft and the Sony implementation of the Main Concept. Both are outstanding. I did these tests not for any reason other than wanting to be "informed" when both writing one of my books, and also to be able to speak to clients. Like I say, my math sucks so I haven't written anything up on my works. BTW, I love your piece on compression, very nicely done. Even an idiot like me can read it.

[Graeme Nattress]

Thanks Douglas. Sounds like theory is right in line with practise, which is excellent.

Graeme

[Robert Young]

Douglas
When you say you've gotten the best results converting HDV to SD with Vegas/Premiere; are you importing (capturing) the m2t stream and using a Cineform type IC, or importing the component out signal thru SDI? I'm not exactly sure what you are getting out of the camera to convert to SD .avi with Premiere. It's important to me because my existing system can handle m2t/CFHD/avi bandwidth, whereas SDI/uncompressed formats/etc. is going to be the next level up hardware-wise.
Also, if I am reading your post correctly, you are saying that the Z1 in-camera conversion to SD firewire out looks almost as good to you as the Vegas/Premiere in-system conversion. And that overall, all of the above "look" better than material shot in PD 170 SD.
Am I getting this right???
Thanx
Bob