Yes, Ruslan, it is a discussion board, one where when people have time to have discussions, they do so. At the moment, I'm in the middle of an HDV Solutions training tour that encompasses travel from the Pacific Rim to New York, to Taipei, to Singapore, and 18 points in between.
So, in order to intelligently demonstrate why both Graeme and myself believe that software can provide a better solution, I'll need to take a fair amount of time to write the article, provide images, and comparisons. I don't have time to do so right now.

Bottom line from my opinion, without taking the time to write up an article, is that if you feel the camera can do a better conversion than Sony Vegas or Adobe Premiere, you're clearly doing something incorrectly. Search these boards, you'll find *several* discussions backing up my opinion, from a wide variety of users. If that doesn't satisfy you, then when this tour ends in July I'll try to find the time.
Please remember that participation in these boards is voluntary, and no one is paid for the information you are provided.

I guess the lack of replies to my challenge about the inferior quality of software downconversion of interlaced HDV to SD means that most of you agree with my statements that Z1's downconversion is actually better. Well, that's fine by me...

P.S.: I thought this was a discussion board and not a forum where all you get is one-liners that you're doing something wrong (where there's obviously nothing wrong with my software export to SD), instead of exploring the issue. Oh well...

P.P.S.: Thanks to Thomas Smet for his opinion though.

Boy... If only I had known, I'd have replied straight away... But then I have the FX-1e and not the Z1. Many FX users would have - like myself - not responded because we don't know for certain whether the processor works identically on both models.
And, as Douglas has pointed out... writing anything beyond mere opinion involves the sort of time that just isn't always available.

Now for my 2c worth...

The logic behind downconversion through software; and more specifically using a 'lossless' intermediate codec, revolves around the quality of the 'real-time' downconversion of the already compressed MPEG2 stream that must be carried out by the camera's internal chip. I'm not aware of any specifications released by Sony on how the chip performs the downconversion. Perhaps you are.
Regardless; the number of times the video must be re-sampled to achieve the final SD product via the internal chip is greater than if you use the HDV original and good quality software algorithms, which don't have to re-sample in 'real-time'.
i.e. subject>lens>chip>processor>tape>processor dv>capture>re-encode.
or subject>lens>chip>processor>tape>capture>re-encode.
or subject>lens>chip>component capture>re-encode (requires capture hardware)

The idea is to reduce the number of pathways that the image has to pass through until it reaches the final product: and barring that, to use some means of maintaining the highest quality of image until the final step. Digital isn't that different to analogue in that regard.

As an FX-1e (and HD-10u) user, I can honestly say that Cineform's CFHD codec has provided my 720x576 progressive DVD renders with the sort of quality that I just couldn't get close too by using the raw m2t's, let alone from the cams DV downconversion. I managed a DV capture by mistake once... and that was enough.

It may be different for the Z1... I don't have one to test. If you have one, you've already got the best HDV camera available, so you are streets ahead of the vast majority of SD/DV camera users. I'm sure that's not such a bad feeling.

Just a tought: can you actually edit the footage in HDV in FCP5, export to tape as HDV, and then let the camera do the downconversion to sd?

Just a tought: can you actually edit the footage in HDV in FCP5, export to tape as HDV, and then let the camera do the downconversion to sd?

Not that I'm a Mac owner, but the process would be do-able....it certainly is on PC: but why would one bother to follow such a pathway? The quality of the result wouldn't justify the extra steps as far as I can see...

Going back to the camera with a 'finished project' is great for quality evaluation (via-component) and archiving at 1080i, but I wouldn't downconvert such a project from the tape to achieve a DV end product.

This is a very interesting discussion. But I am getting a bit lost here with the various claims.

It starts off with a report by the BBC's technical staff saying that HDV converted through an expensive box produces superior SD. Now it does not specify if this superior SD is in fact 4:2:2, as some have hinted. Just as an aside for those here, who don't live in the UK, the BBC is a public funded organisation in that every person with a tv in Britain has to pay 250 USD$ per year to them, or go jail, that's the law. The result is that the BBC recieves some 5 billion USD$ a year in funds, that's a lot. The Sony reps in London call the BBC the biggest goldmine on the planet, for example they once ordered 4000 pd150s in a single morning. What I am saying is you must be aware that all BBC technical reports can be seen as an attempt to rally funds for more expensive kit, that's the way BBC culture works, they can afford anything, they have the cash to do so.

But back to the HDV. Douglas, as I read your post about downcoverting, are you saying that putting HDV through After Affects gets you 4:2:2 DV? Correct me if I am wrong? I have seen other simliar posts to this effect. Some say FCP 5 will do the same...

Anyway, what I would like to know is, if this is the case, how does this software magic work exactly. It's one thing talking about better quality, but is it really 4:2:2.

For the the record having looked at the in camera downconvert from my own FX1, it looks no different to ordinary DV footage, only with more compression. As far as I am concerned, if it don't look any better, it aint.

...
What I am saying is you must be aware that all BBC technical reports can be seen as an attempt to rally funds for more expensive kit, that's the way BBC culture works, they can afford anything, they have the cash to do so.


That makes no sense. Why then make statements that HDV is an acceptable format for SD production? Why not insist everything be shot on Varicam or HDCAM units. I think you misunderstood the statement. I meant that on HDV cams, HDV downconverted is better than basic DV in that particular camera, but only when using a superior downconverter rather than in camera. It's not claiming that HDV downconverted is superior to (for example) Digibeta cameras, MUCH more expensive kit that the BBC already has.

These technical reports may serve as campaigns for more expensive kit, but they are primarily guidelines for independent producers (who make something in the region 75% of the BBC's output) and regional production offices who may be intending to use HDV, and want to know what the best practice is for optimising image quality in SD.

As for outputting to 4:2:2. well, it depends, if you output to a codec that supports 4:2:2 (DVCpro50 for example) then it will be 4:2:2. If you output it to DV then it will be SD 4:1:1. (4:2:0 in PAL countries). There is no such thing as 4:2:2 DV

This is a very interesting discussion. But I am getting a bit lost here with the various claims.


But back to the HDV. Douglas, as I read your post about downcoverting, are you saying that putting HDV through After Affects gets you 4:2:2 DV? Correct me if I am wrong? I have seen other simliar posts to this effect.
.

You definitely missed something in there, there is no such animal as 4:2:2 DV. There is DV, which is 4:1:1 in NTSC and there is DV, which is 4:2:0 in PAL. Then there is 4:2:2 uncompressed, which is not DV. I'm converting to Sony 4:2:2 YUV through a variety of means such as the Decklink or on direct conversion. In other words, hardware vs software.

Ok I'm getting the picture here, but as I understand it DVCPRO 50 is 4:2:2. I know I am probably missing a few steps here though?? It take this is simpy uncomressed DV then?
Does the colour information in downcompressed HDV then match the amount of colour info in DVPRO 50? Or am i going in circles?

Thanks for replys. Is there then a simple way to get this DVPro 50 quality from converting HDV via software, I mean is it on the horizon yet? Can I buy it for under 1000$ Is it do-able.

Dylan,
The BBC have actually discussed doing news in HDCAM etc. So have some larger US and Japanese networks. The problem is this. It's more diffcult to edit in the field on a laptop. This is why DV25 is so popular in the news business. Also most news bulletins use lots of agency and overseas footage in their reports which all comes in DV25/Betcam/DV50 standard. So this would look out of place in HD news.

But HD news will come one day...

ps. I think independent producers do 25% percent of BBC output, not the other way around as you mentioned.

My feeble understanding of the issue/challenge of HDV to DV conversion so far is :
HDV is a Hi Rez image with 4:2:0 color sampling. To get a better SD image than native DV the goal is to retain as much color information as possible, even while reducing the resolution to 720x480.
For NTSC, direct conversion of HDV to DV (in camera, for example) will drop the color from 4:2:0 to 4:1:1. If you then output to DVD, the color is resampled again, now from 4:1:1 back to 4:2:0. This is represents a quality hit from the original.
If you convert HDV to an intermediate, or uncompressed codec, color sampling goes from 4:2:0 to 4:2:2. This doesn't add any chroma that wasn't there already, but you aren't losing any due to the conversion. Likewise, resampling to DVD 4:2:0 should be realitively lossless.
However, for PAL, HDV is 4:2:0, DV is 4:2:0, and DVD is 4:2:0. To me, this implies that in camera conversion to PAL DV may be less lossy than its NTSC counterpart possibly simplifying the workflow.
Do I have any of this right??
Are there any Luma issues with conversion workflows as well: HDV>DV/DVD vs. HDV>IC/uncompressed>DV/DVD?
Bob

Robert, that's pretty much the case.

Graeme

Thanks everyone for replies.

Douglas, I completely understand that you're a busy man. It's just that I wanted a little more explanation as why you thought my downconversion results were not valid. When you're back in July, I'd still like you to try that 60i M2T clip you have on your site with a surfer walking on the beach, convert it through Z1 to DV (or better through component Z1's 480i downconversion to lossless, if you can capture that way). Then convert the same M2T in Vegas to 480i, and pay attention to fields rendition. Let's try to put all the theory aside and just see what downconverst more naturally.

Yes, I absolutely dislike the way the bright colors, especially reds, look from a DV codec. They look particularly bad and stairy when viewed through component outputs on LCD or plasma TV's. That's where software downconversion to lossless codec wins 100%. However, I'm sorry, but the interlaced downscaling in Z1 (put colors aside) is far superior to any software I've tried. Software wasn't bad, it was quite good doing downconversion, but Z1's version wasn't just good, it was transparent, it didn't look like something was scaled. It looked as it was shot in 576i/480i.

So this is my quest now: How to achieve the interlaced downconversion rendering quality of Z1, combined with color rendition of a software lossless codec? Any ideas? Is the analog component input capture card the only answer? Z1's component downconversion to 576p or 480p from HDV is simply stunning.

As for deinterlacing (SD for now). I don't know why people say that software is always best at it. Even highly praised Magic Bullet, with whatever possible tweaks, could not give me as good deinterlacing as a $2k Panasonic EDTV plasma that deinterlaces in real time. This plasma retains the most resolution and least artifacts, while Magic Bullet was struggling on many horizontal lines. Therefore I disagree, I think that dedicated hardware is still for the most part the best for whatever it's designed to do. Software is still lagging behind. Software gives you great flexibility, but not always the best result.

Please anybody, try that surfer clip and see it for yourself..... First of course you'll have to upload that M2T clip back to your Z1/FX1. Also I'm not really sure if Z1 and FX1 use the same downscaling technique, so Z1 test would be preferred.

Thanks.

Ruslan.

...so the question for mac users now is - 1) if fcp 5 has the capability and/or - 2) what then is the best codec to use as intermediate ? i've got a studio upgrade already coming but should I also invest in Lumiere ? This question aside - all I've done is edit downconverted footage thru my Z1 and then run it thru Graemes' filter and I am very pleased . Yes it's dv but it looks way better. I've compared equal footage from my dsr-pd100a next to the z1 downconverted .I've looked at it on my hd 17" lcd and projected at 1024x768. There's no comparison . I haven't yet burned a dvd so maybe that's where the kicker is - otherwise , for whatever their reasons , BBC will be using a great camera .

Ok so, if HDV 4:2:0 is converted via lossless software SD 4:2:2, there's actually no real difference in picture quality. That's what I thought.

And I don't mean 'it just looks better' statements from the BBC or others, this is just being subjective. Hype. People going gaga over new tech. etc. No offence intended to anyone, of course, merely my humble opinion, and I am open to being corrected on this.

There's no techinical or scientific basis for downconverted HDV becoming superior SD, right? Or am being to controversial here, but's that's what I seem to be reading as the bottom line.

Anyway I'll take a bet in ten years time - FCP 12, or Premier 20 - will have software that will make DV or its progenitor look just like high end film colour. Laughing? just remember what things were like ten years ago, could you imagine editing high end stuff on your laptop-?

Does BBC use only PD150/170, or they use also DVX and XL2?

Radek

There's no techinical or scientific basis for downconverted HDV becoming superior SD, right? Or am being to controversial here, but's that's what I seem to be reading as the bottom line.

Have we been reading the same thread? I know English is a beautifully intricate language with nuances that can beguile a novice into erroneous assumptions...

There's also no technical or scientific basis for ignoring the logic that a higher rez source (HD/HDV) should provide a result of at least equal quality to a native SD original. That's assuming the processes to achieve that result are understood by the operator carrying out the downconversion...

There's no techinical or scientific basis for downconverted HDV becoming superior SD, right? Or am being to controversial here, but's that's what I seem to be reading as the bottom line.
I have followed this topic on several different forums for a while now trying to figure out the answer to this. There are quite a few production pros and super techie posters who are really digging into all angles of HDV. These people are like Lewis & Clark, hacking thru the underbrush, trying to get the map drawn. Me, I'm like a tourist. I just want to know where it's safe to drink the water. The few clues I've gotten are as follows:
1) If you want the best SD DV possible, don't use a $5,000 camera. Use a 2/3"chip pro camera with pro lenses.
2) If you are exploring what the Z1 is capable of in SD final output, shoot in HDV, capture m2t, convert to 4:2:2 IC, or capture component output as high level codec, or uncompressed. These images, when downrez to 720x480, are pretty uniformly reported as being of noticably higher quality than DV shot on the Z1.
3) If your projects are for delivery on NTSC SD DVD, as mine are, it makes theoretical sense, and is apparently true in the event, that by avoiding color downsampling to 4:1:1 between the HDV acquisition and the final 4:2:0 DVD output, you will end up with better looking DVD video than you would have obtained from shooting DV in camera. I haven't heard how that works in PAL. There my not be as much difference.
4) I think that the ultimate purpose of the Z1 is to make HD products for people to watch, not SD on steroids. Unfortunately, for the moment, it seems like the camera has arrived before the delivery system. When HD DVD is finally up and running the Z1 HDV will produce video that is head and shoulders above standard DV format.

For me, the question is whether or not to jump in to this now, thinking that I may only get a marginal improvement in my end product, or wait for the technology to mature a bit, and to see what options we'll have for delivery.
Bob

Fair enough. And I do understand the concept of downconverting HD in a simple way. But as a keep saying, what does 'noticable better quality' mean? More chroma or luma, for example, like dvpro50.

It's just when I view HDV, it looks like ordinary DV.

If anyone has the time, can they post comparisons for judgement...just a thought.

interesting thread....

if you downsample to 720*540 you will get 4:4:4 from a HDV source, provided you save in a lossless c*dec

more info here:

http://www.canopushdv.com/conversion/index.html

see page bottom

ps. I think independent producers do 25% percent of BBC output, not the other way around as you mentioned.

John. sorry, guess I got my numbers backwards.

No problems, I guess I'm pretty backward myself at times :)

John, thanks for that link. That does put the discussion it in a better light. Though frame grabs and moving vid do tend to be very different things. The HDV downsamples do look sharper, certainly compared to the dsr250 SD, as they have there. I think DSR uses the PD 150 chip, and I think most agree the FX chip does produce cleaner images than the old PD/DSR insides.

It would be more interesting if they were actually comparing FX video shot in normal SD, not in-camera downconvert via HDV for a truer comparison???

Hey John, interesting link.

Does anyone besides me find it funny that while Canopus mentions the possibility of down-converting to 4:4:4 SD, their codec only goes to 4:2:2 SD? They're throwing away more information than they have to. While they quote that they do it in "the best possible quality", this is completely incorrect.

-Steve

http://www.canopushdv.com/conversion/index.html
OK guys, I looked at those M2V clips. And guess what, if you think I'm doing something wrong with my software downconversion to SD, well Canopus must be doing something wrong as well, because those horizontal lines look awful!

I mean yes there's less mosquito noise and slightly better color rendition in Canopus version of downconversion, but come on, one must be blind not to see that FX1/Z1's downconversion produces more natural looking lines without softening the overall image. Simply try viewing this on interlaced PAL monitor (yes it's in PAL). I viewed it through RealMagic Hollywood+ MPEG2 card and in-camera downconversion looks way better on the eye than Canopus' version.

Try to compare sony_tram.m2v to edius_tram.m2v and look for the nasty horizontal artifacts on those long steps in the lower part of the frame, especially when zooming out. In-camera version looks almost completely natural, while Edius' does not. And this is the kind of artifact that no filtering will be able to remove. It seems like FX1/Z1 does some really good motion estimation technique when downconverting, while software doesn't do it at all or does it really poorly.

The difference is very obvious in favor of in-camera downconversion. So I'd say, if you want to bypass DV codec, you have to capture SD from component FX1/Z1's output into lossless codec for best results. But the in-software downconversion looks plain bad on those detailed shots. If you think that it's Canopus' issue, well, I was getting the same exact results with Vegas. That certainly is not an acceptable result for me.

So, can anybody at all see what I'm talking about? Anybody???

P.S.: You could also burn those clips to a PAL DVD. However, you'll have to reduce the bitrate first, as those clips are beyond the allowed bitrate specifications for DVD-Video. You can either recompress with MPEG2 encoder or transcode using some free transcoding application such as DVDShrink.