I know that some answers may be NLE specific, but in this "neutral" territory, I'd like to hear any tips towards maximising quality on SD - DVD.

It's going to be a while till all our clients are clamoring for Blu-Ray output, and so far I've not been knocked out with the finished quality, after shooting 1080P that looks stunning prior to being rendered into widescreen SD MPEG2.

The worst problem I've had is the old flickering around thin horizontals, and in Vegas the "Reduce Interlace Flicker" helps, as does rendering at Best quality to help with the down-rezzing, but I'd love to hear what others are doing in this direction.

Same with me - the SD DVD's produced via software are not acceptable to my standards.

The best solution that I have found out till now is this:

Shoot 1080 50i (I'm in PAL land) - transfer to pc - edit HD footage using Premiere CS3 + Axio LE - export and burn to Blu Ray via Encore if requested - for SD DVD - connect component out of Axio break out box to component in of my Sony 1800 DVCAM deck - play out the edited footage of the edited HD timeline whilst recording onto a DVCAM tape - real time instant hardware very good downconversion - then capture DV footage back and encode the SD DVD.

This might sound a bit complicated and so I'm going to try out the following workflow. I already have a Matrox RT 100 xtreme so:

Shoot 1080 50i - transfer to pc - edit HD footage using Premiere CS3 + Axio LE - connect SDI output from breakout box to SDI input on a Kramer FC20 (this converts SDI to Firewire) - connect Kramer to RT 100 via firewire - play out HD timeline whilst capturing real time straight to RT 100 as m2v file - real time instant conversion to m2v ready to be authored by Encore.

I haven't yet tried out this last workflow but I'm sure it will work out like the previous one. I'm going to use the Kramer since my Rt 100 has only firewire in (apart from the useless composite and y/c) - if it had component in I would have avoided the Kramer conversion.

Unfortunately when one sees the 1080 footage then any SD DVD will look abysmal. I actually had a good look at the SD DVD's that I used to produce by my DSR-300 and found out that resolution was in fact reduced but I never really notoced. With the advent of HD, this lack of resolution is now glaring in our faces.

Thanks Brian - interesting to hear about your hardware based approach.

Prior to this I was shooting HDV with a Z1 and the resulting DVDs looked fantastic - I was hoping that EX1 shot material would be even better, and I still am ..... surely I don't have to start shooting HDV on the EX1 to achieve this - but maybe I do!

Peter that is strange! I have no experience of HDV - so you are saying that HDV to SD is OK. So what is happening in the transfer of 1080 to SD? So do you think if we shoot in HDV mode on the EX1, we will get better looking SD DVD's? Any comments on this from other users?

Peter that is strange! I have no experience of HDV - so you are saying that HDV to SD is OK. So what is happening in the transfer of 1080 to SD? So do you think if we shoot in HDV mode on the EX1, we will get better looking SD DVD's? Any comments on this from other users?

Good question Brian

I don't know any technical reason why HDV converted to SD-DVD should look better than Full HD to SD-DVD - just observing that so far it has, at least here.

The Reduce Interlace Flicker setting I referred to does make a considerable improvement, but it may be at the expense of a certain amount of resolution - at SD that doesn't seem to matter.

Part of the problem may be that I've become used to seeing all the detail and definition of 1920x1080 on a 24" screen, and watching SD, particularly on a CRT screen, will always be a let down.

There's two problems:

1) Line twitter. With interlace video the vertical resolution cannot exceed around 80% of the theoretical maximum or horizontal edges will winkle. This is dealt with in interlaced cameras by line pair averaging. On the upside you gain an improvement in noise or sensitivity. You can see this in those cameras that do both I and P, I gives you more sensitivity.

2) Aliasing. Nyquist places a limit on the maximum frequency that can be presented to a sampling system. For DV the clock is 13.5MHz and one way the frequency (aka resolution) is controlled is by the use of an optical low pass filter (LPF) in front of the imager. The design of that seems pretty tricky, ideally you'd want a square edge at the cutoff frequency but thats hard to achieve. Let too much detail through and the camera will produce images with aliasing, make the cutoff too low and the image will be soft.

During down conversion from HD both these issues might need to be wrangled. Most HDV cameras don't have enough resolution to pose a challenge, moreso if you're shooting interlaced. The V1P sure could produce these issues and there was much discussion about it here.

With Vegas my fix for both HD stills and video has been the gaussian blur or median FX. You should apply these before downconversion and usually you only need them in the vertical direction. As for the design of the OLPF in a camera, getting just the right amount can involve a bit of trial and error.

Thanks Bob,
I will try it tonight

Most HDV cameras don't have enough resolution to pose a challenge, moreso if you're shooting interlaced. The V1P sure could produce these issues and there was much discussion about it here.

Not that I'm stealing this thread, but in the light of Bob's explanation (and indeed the line twitter has been a real problem with PAL area versions of the V1), two things come to my mind:

- as this involves only the vertical resolution, it shouldn't make a difference whether one is down-rezzing from HQ or SP

- why there's no line twitter in the EX1, even though it's reported to have an even higher vertical resolution than the V1E/P.

Bob?

Not that I'm stealing this thread, but in the light of Bob's explanation (and indeed the line twitter has been a real problem with PAL area versions of the V1), two things come to my mind:

- as this involves only the vertical resolution, it shouldn't make a difference whether one is down-rezzing from HQ or SP

- why there's no line twitter in the EX1, even though it's reported to have an even higher vertical resolution than the V1E/P.

Bob?

1) Not really, the horizontal resolution could cause aliasing. Also who knows what differences exist in how the encoder works at HQ and SP without checking both with a res chart.

2) Who said there's no line twitter in the EX1?

Suggestion. Take the ISO 12233 res chart (Google for it, it could be copyright so I'll not attach a copy here) then animate it with Vegas (or any NLE), see what happens.

1) Not really, the horizontal resolution could cause aliasing. Also who knows what differences exist in how the encoder works at HQ and SP without checking both with a res chart.

This I cannot comment at the moment; I think though it's a very important aspect of the SD DVD creation workflow, deserving its own discussion (see below).

2) Who said there's no line twitter in the EX1?

I did :)

Seriously Bob - my shed's fine horizontal lines that you have been tortured with so many times as my main "testing object" with both the V1E and now the EX1, don't twitter at all even if I engage bobbing ON in an MPEG player (nor do they with straight playback from the camera using component)!

BTW, as I am planning to record a lot of live classical music concerts with my EX1 and - while keeping HD masters - deliver them on SD DVD's for quite a while, this discussion is of a great interest to me. Should it prove to be easier to deliver high quality SD from the EX1's SP than from HQ, it'd be a very important incentive to keep my DR60 drive and record SP to that, rather than invest into more SxS cards and get more HQ stuff, but only to get more frustrated at down-rezzing! So, to me at least, sort of a strategic decision to make..

Any input will be appreciated.

Should it prove to be easier to deliver high quality SD from the EX1's SP than from HQ, it'd be a very important incentive to keep my DR60 drive and record SP to that, rather than invest into more SxS cards and get more HQ stuff, but only to get more frustrated at down-rezzing! So, to me at least, sort of a strategic decision to make..

Any input will be appreciated.

I second Piotr - I too am getting frustrated with the EX1 down-rezzing from full HD. Will SP (aka HDV) mode produce a better looking SD DVD?

My from HQ to SD resized DVDs look fine.
As sharp as commercial DVDs with less artifacts - they look nice on Flat LCD TVs and even on old-fashioned CRT-TVs.


I shoot in 1920x1080 25p HQ with Detail=Off, which is very important!

regards Dennis

Ok guys, worst case: Nature :D

I'm NOT satisfied with the results.
The sources are looking very good though.

http://rapidshare.com/files/96728426/HQ_SP.zip.html


left is HQ 25p and right is SP 50i


regards Dennis

Dennis, interesting that your shot shows tree branches. I shot at Brooklyn Botanic Gardens some weeks back with EX1. Lots of naked tree branches shot at 1080p30.

Every attempt at down-conversion looks HORRIBLE.
Viewing on Samsung 46" 1080p LCD HDTV
Using component out of Sony SD DVD player with "Progressive" button (no upconvert).

Line Twitter on branches and aliasing on angles very bad.
I'm on Mac so I tried "Rick Young" method (posted elsewhere) and used source in Compressor 3 to create SD DVD. I also tried the EX1 1080p30 sequence as source for Compressor 3. That actually looked a bit better but still pretty bad.

Maybe adding a blur filter is a solution in such scenes?
There's definitly too much detail for SD.

regards Dennis

Wow, everybody please keep on posting with your results as I am looking at this camera for a purchase.

SD DVD will be the delivery medium for a few years to come at least, whether we like it or not.

It is troubling that this camera can look so good in HD but making an SD product does not turn out well.

The shoot in HD and deliver in SD is the workflow that all of the camera manufacturers talk about.

Has anybody tried exporting as a tiff sequence and letting Photoshop do the downsampling?

I tried some other resize filters with avisynth.
Bicubic (soft) looks best, less linetwitter/Aliasing and not unsharp like gauss and very similar to Photoshop Resizer (bicubic..).

Here's the result:
http://rapidshare.com/files/96755900/From_HQ1080p_withBicubic.m2v.html

The problems being experienced are not unique to the EX1. Certainly the EX1 is probably the first HD camera most of us have that pushes the resolution high enough to really cause us grief however that in no way is a problem with the camera, it's a general 'How To Downscale HD to SD' problem.

If you're trying to crack the problem on your NLE the best appraoch is not to use footage from the camera at all, the appearance of the problem is content dependant. Start with res charts or a high res still from a DSC, something with fine detail and sharp edges. Do a slow vertical pan using your NLE. Try downscaling that to SD. Ideally you need a workflow / solution that'll work regardless of content. I spent a lot of time dealing with this issue with HiRes stills well before HDV cameras ever appeared. I must say that to get the very best results on a SD PAL DVD I spent many hours hand tweaking the problem parts of each image while I watched the results on a interlaced monitor. Obviously this is not a viable solution with video.

Perhaps the best answer lies in the hardware downscalers that broadcasters use although they're well outside the reach of most of us. Maybe what we need is a better tool that does in software what those hardware boxes do, unfortunately that could mean very long render times. From discussions years ago with some engineers the problem is implementing a SinC function that doesn't require excessive iterations i.e. takes forever to process every frame.

So I must ask since I am not working at HD resolutions yet:

If one knows SD DVDs will be their primary distribution, is it not worth upgrading to HD yet?

Mainly due to the difficulties in getting a great looking SD DVD from HD source material.

I have heard quite a bit about this problem in the Edius forums.

Thanks for your input.

I know that some answers may be NLE specific, but in this "neutral" territory, I'd like to hear any tips towards maximising quality on SD - DVD.

It's going to be a while till all our clients are clamoring for Blu-Ray output, and so far I've not been knocked out with the finished quality, after shooting 1080P that looks stunning prior to being rendered into widescreen SD MPEG2.

The worst problem I've had is the old flickering around thin horizontals, and in Vegas the "Reduce Interlace Flicker" helps, as does rendering at Best quality to help with the down-rezzing, but I'd love to hear what others are doing in this direction.
Hello. I have recently upgraded from a Z1P to EX1. Set up,1080p 25 Pal. profile settings Sony Default and Doug Jensen Vortex recommendations. Both look fine. Work flow into FCP v6. Easy Setup-XD CAM EX 1080p25 VBR. Export-XD CAM EX 1080p25. The rendered QT .mov looks stunning when expanded to full monitor size (23 inch display) the bottom line, I place the movie into iDVD resulting in a really degraded flickering jagged edge (strobing call it what you like) image. It's common to both ambient & studio lighting. It is most apparent when the camera is panned or tilted, when the target moves, not so bad. I have tried to down sample by copy & pasting into a new FCP Sequence with a 1080 720 format, all with negative results. I view the results on a HD 1080 720i 42 inch screen. A point to consider, when viewing the Vortex DVD sample footage of 1080p60 on the same 42 inch screen, it looks terrific. What am I doing wrong? I attempted to contact Sony, it's impossible, being referred back to the supplier. I don't know what to do next, there has to be a simple common sense solution. Thanks. Ron Coker.

http://www.kenstone.net/fcp_homepage/ex1_sd_output_young.html

Works pretty well for me. Although the best method would be to use a matrox mxo and downconvert from the raw dvi output to a mini-dv deck and then reimport from the mini-dv and burn to dvd. You could always output to a stand alone dvd burner through the mxo's s-video connection as well.

My original post was made after watching a DVD on a CRT TV - lots of twittering lines, jagged edges and colours looked wrong.

After applying Reduce Interlace Flicker in Vegas, things started improving, and now I've watched the DVD on a 24" LCD things are getting quite acceptable - not perfect but quite impressive. I realise that many viewings will still be on CRTs, but that will lessen progressively from now on.

Thanks for lots of interesting reponses. One thing I regard as important is ease of conversion - I'm not keen onthe idea of always having to render more than once, or outputting then recapturing - what I'm after is having the original HD footage on the timeline and rendering directly from that to DVD format.

Vegas is already managing that quite well, and I now intend experimenting with as few of the other suggested tweaks, such as median filter and maybe a tiny touch of vertical gaussian blur, to hopefully arrive at an optimum workflow.

... and of course the EX1 settings may have an even greater effect. Lately I've been shooting 1080/25P, but I should try some others ....

I have to say that it's astonishing that it would be so difficult to get acceptable looking SD from XDCAM or HDV source material. Is this really possible? I've read (I think) all the threads on this, and getting even close to an acceptable SD DVD requires all sorts of acrobatics.

I ran into this problem just today-- a mixed FCP timeline with SD footage from a DVX and downscaled (in FCP) HDV footage from a Z1. The DVX footage looked great, the HDV footage was a mess, with stairstepping everywhere.

Z1 footage downscaled in-camera looks great. But then I lose the ability to scale and crop at HDV resolution.

I know that someone will say, "you should use the 'so-and-so' method", but it really seems like we shouldn't need an arcane "method" to achieve what is essentially a basic task in post.

(Harumph harumph)

Interesting thread.
I've, recently, tested a workflow whereby I do all my editting in Edius HD, rendering the final version to an intermediate HD like Cineform, Canopus HQ, PicVideo, or HuffyUV. I input the intermediate to VirtualDub and use the resizer plugin, which allows the selection of different algorithms, depending on the source material, to downsample to SD. I then feed the downsampled output into TmpGenc. I've had the best results, yet, with this workflow. Minimal stairstepping and flickering. The mpeg2 output looks extremely good, at least on my 32 inch Sony HDTV.

Downsampling is a problem that dates back to Digital SLR development days. Lanczos, bilinear, and trilinear downsampling techniques have been around for a few years. Only VirtualDub seems to have all these choices implemented. Flicker is the result of interlaced video interacting (harmonic beating) with TV raster. It can be all but elliminated by going progressive all the way.

It's counter-indicated to blur HD content for the sake of reducing twitter/flicker.

Bill, I have to ask:

If you shot a DSR-450 and an XDCAM HD (in HD) side by side at the same time knowing you were going to produce a widescreen SD DVD, which camera would you choose?

Does the quality of the HD downconverted look worse, equal or better than an SD camera?

Because I am equally shocked.

When I take a photo with my 5D at 12mp and downsample the image to 800 pixels, it retains its integrity. The image does not have a bunch of "junk" added to it.

I would have thought that aside from interlacing, there would be nothing else in the way of HD sourced material always looking better than SD sourced material.

Maybe we are early in the cycle, but HDV has been out for at least a few years.

I am glad you found a way to get decent results, but outputting separate files, downsampling, and encoding in separate programs sounds like a lot of time and storage coming from a DV workflow of chop, chop and go straight to DVD.

Tim...

Afraid I don't have the experience with a DSR-450 to answer your question.

Digital SLR's, like a 5D,have the advantage in that they don't have to display motion. While modern dig vidcams don't have a raster, per se, they still constuct an image with horizontal rows of pixels.It's the motion, combined with row frequency that causes the twitter. Perhaps if the CCD/CMOS pixels were spaced alternating up and down rather than in a linear row, things would be less flicker prone.

One other note, there is a deflicker plugin for Virtual dub that works quite well. If you have an interest, google, msu_deflicker. We seem to be in the midst of technology development. For the time being, piecework rendering is optimizing the work path. One of these days, someone will integrate it, but, not yet.

With Varicam jobs for my former employer, I found that our best looking SD downconversion came by running the project through our KonaLH cards, and laying the project off to tape. This method always produced far superior results to running the project through Compressor. I left that position before I had much of a chance to test this workflow in FCP Studio2 so I don't know if Compressor has gotten better at handling downconversion to MPEG2's.
Past experience tells me to do a hardware downconversion when possible.

I fully agree with Ethan - that's what I did. Output to tape and then re-capture. Very silly workflow but it's MUCH better that software conversion. Or else have 2 edit stations linked together instead of outputting to tape. This method is quicker and real time.

Tim...

Afraid I don't have the experience with a DSR-450 to answer your question.


Well, how about two equal quality & chip sized cameras, one SD, one HD.

Which one would you chose if you knew a SD DVD was your output source?

No brainer...I'd pick the HD cam.

Thanks for your reply.

It is reassuring that even with all of the talked about hoops to jump through to get a clean SD DVD that you would still choose the HD camera.

Hopefully there will be a software-solution showed at the NAB to convert the EX1-HD clips to SD in a good quality. AT IBC when I asked AJA and Adobe about handling EX1 files the first answer I got was "EX-what"? lol Only Mainconcept told me about their Codec.

Think now they all have understand that EX1 will be a big market for them. Still hope to see a good working software-solution for this problem. Maybe ProCoder4 or simililar. Also an update of the Mainconcept pro HD Version would be great. I'll aks them all at NAB again.

I don't like the idea to export with an AJA-Card (sadly I don't have such a card now) to a DVCAM-Camcorder and then to capture it again. That sounds like an expensive and time wasting workflow for me. Really strange that Real-Time Hardware solutions are better than non-realtime software solutions.

http://www.kenstone.net/fcp_homepage/ex1_sd_output_young.html

Works pretty well for me. Although the best method would be to use a matrox mxo and downconvert from the raw dvi output to a mini-dv deck and then reimport from the mini-dv and burn to dvd. You could always output to a stand alone dvd burner through the mxo's s-video connection as well.

I have got great results using this method.

First serious shoot with EX1, over 1 hour long shot. Went straight from HD to SD mpeg-2 and onto a 16:9 DVD using Vegas 8 Pro with DVDA 4.5. Due to the subject nothing in any frame that'd cause line twitter so I had an easy time of it I guess.
Despite forgetting to switch TC out of Clock the clip stitched back together seemlessly. I only shot 50i SP, all fitted onto my 2x 8GB cards.
Double headed the audio with Edirol R-4. All synced with around a 2 frame error after 1 hour. Dead easy to compensate with Vegas.

Would it help to shoot with "detail off"?

I've been doing a fair amount of trials. Indeed, shooting with DETAIL ON, levels "0" will produce a fair amount of twitter/flicker. Once I disable DETAIL, twitter is gone. Recover sharpness in post. Don't fall for the easy trap, ie turning detail on.

I ain't getting any twitter or flicker,it's just all smooth,with pans and everything. I'm using your picture profile Bill,detail on.
1080/50i or 720/50p,i render as interlaced,upper field first.Burn with DVD-A.

I've watched on a CRT tv,top of the range Sony from about 5 years ago,and a Sony 40 inch LCD. Picture's good on both.I can't see any difference between 1080 or 720 once it's downconverted to SD. 720 is a lot easier to edit though in vegas.

Paul.

Paul..

I'm shocked..."ain't" ??

Not the Kings English.

Oh ok,well one is not getting any twitter or flicker on ones television set. :-)

Ps,we've had a queen now for quite a while,not a king !!!
Ha ha.


Paul.

hehehehe

Con su permiso

Everyone should look into the Matrox MXO. I've been using it with my Macbook Pro with incredible results. I usually use the MXO to connect to a 1080P monitor for my HQ 24p footage. When I'm done the project I then use a sony dsr11 to lay the video off the mini dv. Basically the MXO does all the heavy conversion via it's hardware. You connect the MXO via s-video the dsr11 and it does a very good downconvert. Then you can re-import the footage and burn the dvd.

The only other method would be to do what real post production houses do. When they can't run the footage through a professional hardware scaler they usually export the time line as uncompressed SD and then do the mpeg conversion in compressor. This method is the second best to hardware however.

Everyone should look into the Matrox MXO. I've been using it with my Macbook Pro with incredible results. I usually use the MXO to connect to a 1080P monitor for my HQ 24p footage. When I'm done the project I then use a sony dsr11 to lay the video off the mini dv. Basically the MXO does all the heavy conversion via it's hardware. You connect the MXO via s-video the dsr11 and it does a very good downconvert. Then you can re-import the footage and burn the dvd.

The only other method would be to do what real post production houses do. When they can't run the footage through a professional hardware scaler they usually export the time line as uncompressed SD and then do the mpeg conversion in compressor. This method is the second best to hardware however.

This is what I have found out to be the best way - doing a hardware downconversion - but to go thru y/c (s-video) you are losing a lot.

I hate to be so cynical but those doing software downconverts, short of presenting some "different" workflow, aren't seeing twitter because they're just not using video that shows the issue.

You will NOT see it on most of your video UNLESS you have THIN HORIZONTAL LINES exacerbated by some type of non horizontal movement.

In my case it's very obvious with barren horizontal tree branches with slow zoom. You will NOT see it on talking heads, shooting pebbles, etc. The same tree branches look fine when the shot is held. My source is 1080p and I keep it progressive on down convert.

Craig,i've panned left to right,up and down,slow and fast,past walls,windows and trailers with mesh sides,i ain't getting any twitter or lines or anything,just smoothflowing footage. I've purposely tried to find twitter etc but can only find it if i render wrong. I render main concept mpeg-2 for dvd-a,upper field first,the same is for 720/50p and 1080/50i.

Paul.

It's not pans, it's zooms. A pan is likely on a similar horizontal plane. The image must change horizontally and must have VERY THIN lines.

This is the footage but you will NOT see it since this is H264.
http://thirdplanetvideo.com/CineAlta.html

That 2nd shot with the zoom out from the Japanese architecture in the water. Those tree branches in the background twitter like crazy. Also the horizontal lines on the roof of the building overlooking the water a few shots later. Other shots like the close up of tree branches are good. It's those THIN Tree branches at a distance. They're probably only a scan line (or less!).

I can probably upload the MPEG2 Elementary Stream so anyone can burn to DVD on Mac or Windows and see this.

BTW this problem isn't specific to the EX1. On another forum a user is seeing the on the HVX200 shot at 720p24. He also said it's more obvious in "nature" shots. I mentioned the trees I had shot with zoom and he said, that's what he's seeing.

What I think we need is some "intelligent" line blurring much the way a professional compression app deinterlaces by using edge detection and applied to moving areas only. Those kinds of deinterlaces only act on certain areas to keep resolution detail high. We need something similar for HD to SD conversion. It must be out there otherwise you'd be seeing this on commercial DVDs.

If you're not seeing it then maybe main concept is blurring the lines but if you look at my video above I think you'll see what kind of source I'm talking about.

Basically a horizontally thin line that moves from scanline to scanline at 1080p downrconverted to SD (even kept progress) seems to be too thin to resolve properly.

Craig,i've panned left to right,up and down,slow and fast,past walls,windows and trailers with mesh sides,i ain't getting any twitter or lines or anything,just smoothflowing footage. I've purposely tried to find twitter etc but can only find it if i render wrong. I render main concept mpeg-2 for dvd-a,upper field first,the same is for 720/50p and 1080/50i.

Paul.

This is what I have found out to be the best way - doing a hardware downconversion - but to go thru y/c (s-video) you are losing a lot.

The only problem is that if you're going to a deck that only offers a composite or s-video in then I would choose s-video. Of course you would chose sd-sdi or component if the deck you're mastering to has it.

Those who are using Compressor 3 for the MPEG encoding, have you played around with the anti-alias settings on the downconversion?

I'm new to that program, but would have assumed that this might help.

One interesting result from trying to finesse the SD Downconvert from my last shoot with Vegas 8.
Applying an Unsharpen Mask FX prior to the downscale produces very visible twitter and aliasing. Applying it after the downconvert no such problem. It's good that one can do all this in one pass with Vegas, I'm rendering straight from the project T/L to 16:9 SD mpeg-2.

One interesting result from trying to finesse the SD Downconvert from my last shoot with Vegas 8.
Applying an Unsharpen Mask FX prior to the downscale produces very visible twitter and aliasing. Applying it after the downconvert no such problem. It's good that one can do all this in one pass with Vegas, I'm rendering straight from the project T/L to 16:9 SD mpeg-2.

Bob, perhaps I'm too tired and sleepy after all day work - but don't quite follow; how do you do all this in one pass, if the Unsharpen Mask is applied AFTER downconversion? Doen't downconversion require a separate render? Please elaborate on your steps; TIA>