This “Guide” here, started as my own answer to a post I did some time ago. It end up to something that I wish to share with everyone interesting on this kind of conversion, without the use of avisynth. I’m a PAL user and I wrote all this with that in mind. But maybe our NTSC fellows can benefit from it too.
The situation: Most music channels broadcast 4:3 while the majority of the music videos are widescreen. The same happens with various European movie channels and lately, with some of the current US TV series.
The project: Crop the black bars from those letterboxed transmissions and create from that source the best possible anamorphic 16:9 result (mpeg 2, 720 x 576, ready for DVD)
The source: DVB Satellite or Terrestrial channels from all around Europe. Most of them have a framesize of 544x576 but that varies a lot. No analogue capture here, direct digital capture only!
As you understand, this project needs heavy filtering, resizing, etc. I searched hard and from what I found, I can suggest 2 methods of doing it.
The best method
The best method is the combination of Virtualdub and Demon's video Enhancer. In short terms, load your source to Virtualdub, crop the black bars from the top and the bottom of the screen (144 vertical lines, 72 on the top and 72 on the bottom), delogo – if you wish - what's necessary and save the result to a new mjpeg of huffyuv avi. Then, load this file to Demon’s Video Enhancer, set the output framesize to 720x576 and hit “convert”. You can also add many Virtualdub filters to Video Enhancer, and I suggest that because it runs the filters twice as fast as Virtualdub itself!
Video Enhancer, gonna upscale the cropped picture to whatever you set manually: If you set to 720 x 576 the result gonna be an anamorphic picture beautifully upscaled from any source. Then, feed the result to your favourite encoder, set the input as 16:9, hit encode and that’s it.
This method might be the best to do this, but it is slow, even for my Intel C2D6600... Video Enhancer along, needs 45 min to upscale a 4 min Video, so this solution, for me at least, is out of the question. Don’t mention the HD spaces that required for the intermediate files. Maybe it is the solution for those very special projects... But for the rest, I needed faster alternatives! So here is the alternative method!
The alternative method
The alternative method requires Virtualdub mpeg2, specific Virtualdub filters (the links are at the bottom of this page) and TMPGenc Plus 2.5 encoder. Of course, you can use your own favourite encoder. But I suggest TMPGenc Plus for this, since you can do some tricks with it and enhance even more the result! Now, let’s see the steps of this alternative method
The Virtualdub steps
Since now, my alternative method needs a specific filter chain, in a specific order.
- First, I load my DVB mpeg 2 files to Virtualdub mpeg2. I don't use Virtualdubmod, because somehow the preview is not as good as on Virtualdub mpeg2. Personally, I use the old 1.5.10 (built 18270) version.
- Then, I load the necessary filter msu_smart deblocking. Latest versions fixed a problem that appeared on the edges of the picture, so it's time to update (if you didn't already). This filter must be always first on the filter chain otherwise it won’t work.
- The third step is the step that nobody really likes: De-Interlace ...
This step is necessary, because enlarging interlace video is impossible “as is”. You can resize vertically without issues, only when your source is progressive. If not, you have to de-interlace.
After many tests, I end up using various de-interlace filters for different occasions.
The Alparysoft Deinterlace filter, seems to be the “overall” better solution. It is a shareware plug in for Virtualdub and the demo they offer, just adds a small logo on the bottom right of the picture, which can easily cropped later and nobody gonna see it. Of course, adding a small logo on the screen, reduce a bit the speed, but it is unnoticed on my system. This filter seems to work “perfect” with PAL interlace sources, but when you de-interlace a PAL interlace broadcast from a source that previous converted from NTSC material, you may have motion jitter. With simply words, don’t use this filter when you convert music videos from the American market.
A great (freeware) alternative is the “Smart Deinterlacer Filter” by Donald Graft. It does what Alparysoft’s filter do, sometimes better sometimes worse. Seems like it's the best alternative for broadcasts based on VHS sources, so if you are a fun of late 80s and 90s music videos, this is the filter for you.
A third great alternative is the Deinterlace "Muksun" filter by Stolyarevskiy Sergey. It is the only de-interlace filter that made it always without blending something on my tests. Unfortunately, the quality price to pay for this is high: It doesn’t make it well on scene changes. But for me, this filter seems to work better when there are added graphics on video and it seems to do less motion problems on converted Videos from NTSC sources.
The good news are that the European DVB channels broadcast in various ways and there is a great chance to reconstruct the progressive source, without really de-interlace your source. Especially with the “made in Europe” music videos of the last decade, more than the half are possible to make them again progressive, before anything else.
There are three possibilities to avoid the necessary de-interlace step:
a) The obvious one: The channel is progressive. Have luck to find one!
b) The channel broadcast to something that I call "pseudo-progressive” (until someone points me out of how this is truly called). In this case, you have something that looks “Progressive”, but in reality there are 2 fields without the offset position between them. You can notice some kind of interlace artifacts in the left of your screen, easily cropped later, while you resize (upscale). This “pseudo-progressive” broadcast, is very common to the French channels and especially those services related with “Canal +” one way or other (most of the SECA/Mediaguard and Nagravision subscription packages in Europe, are related one way or other to Canal+). Those channels seems that they do some kind of hardware 2:2 pull down (PAL to Film and speed up to 25fps) on their own sources before they broadcast them, or they have hardware filters we don’t even imagine that exists. The point is, we don’t need to de-interlace those broadcasts!
Keep in mind that the very same channels broadcast converted NTSC to PAL material truly interlaced, as they do also with videos from VHS/SVHS or Beta sources (70s,80s and early 90s videos mostly). So see careful if the video you wish to convert to 16:9 looks progressive. And since most r’n’b’, rap and rock videos are made in USA, if this is your music taste, de-interlace is necessary (and expect some motion jittering too…)
Music channels like the MCM and M6Music variations, UK’s MTV Dance, Viva Germany, Poland and Austria and JIM Belgium, mostly broadcast their videos to that “pseudo progressive” way.
c) The channel broadcast to “forced” interlace. This turned to be very common those days especially to music channels like Germany’s “Interactive Music 1 – IM1” and Belgium’s “TMF”. Those channels send their signal to the broadcast terminals in a progressive form (possible through a network) and the broadcaster turns them to interlace before air them. You can easily turn those broadcasts back to their progressive nature, by using the “Progressive Frame Restore” filter, created by Simon Walters. So, remember always to try this filter before other de-interlacers! You may not need to de-interlace after all.
There is also a situation, which some channels, broadcast to a form that I call “switching”. In short terms, they switch their broadcasts from Progressive to Interlace and back to Progressive all the time, for various reasons that I can’t explain in this guide. You can read about this DVB issue in a more detail here: www.uwe-freese.de/software-projekte/virtualdub-filter/TBDI_Doc/index.html
To handle this, I end up using the “Smart Deinterlacer” filter by Donald Graft.
- After all those steps, it’s time to crop. I use "null transform", a virtualdub's internal filter for this. I do all the necessary adjustments so to keep the aspect ratio correct, since rarely something actually is "true" 16:9 and I don't wish to crop something from the picture. This step can be tricky, but since you do all the other steps and this hobby is about patience and perfection (yes, sure...) I do the necessary adjustments here, so to fill all the picture on the 16:9 canvas with the less possible black boarders. Just remember: You have to crop 144 total vertical lines - No more, no less! It can be more only if you also crop a bit on the side so to apply a slight "zoom". Since this step is done after the de-interlacing and before the resizing (upscaling), it doesn't harm the aspect, if you are careful!
- The third step is to de-logo what's necessary, inside the active picture area. That may need the combination of DeLogo and Logoaway filters. Delogo is overall a better filter and you can choose manually on which frames to be applied. Unfortunately, this filter may not work at all when the logo is between the active and the non-active area of the screen. On those cases, I use Logoaway, since I don't know any other filter to do a better job on this. Unfortunately, this filter crashes when I set manually specific frames I wish to use it and for that reason I need to have it always "on". I still hope and pray for a better "delogo" filter for virtualdub. Our friends from Russia (MSU) promised us a better solution, but their solution is on a beta stage for now (buggy and slowwwww).
- After this cropping, is a great time to add the Static Noise Reduction filter, by Steven Don. But only slightly: I use it to "3". On noisy sources, I may use other alternative noise filters: there are plenty of them, as you know. Video DeNoise 2.0 does a great job, so MSU_De_noise. It's up to you!
- The next step, is what this guide it's all about: The resizing! The source is not interlaced anymore, so you can upscale the picture to 720x576 as progressive. I use Precise Biqubic 1.0 for this: IMO it gives more sharpness that the other options
- Now, it's the last touch: I may use a bit the Dynamic Noise Reduction (I set it to 3) filter if the source is grainy. Raerelly, I may use the "Frequency Suppressor of the Noise" filter by Stolyarevskiy Sergey. "Why?" you may ask: Well: The original source of many channels, remains VHS or DV: The channels digitise those sources realtime on mpeg 2 and broadcast them. Or, they have digitised once there archives and they made a digital jukebox so to broadcast them (It is the case on most music TV channels, when they broadcast Videos made before 1999). Or, the TV station signal goes analogue to the transmission center and it is digitised and multiplexed there for the digital broadcast. The point is that at one step of this process something analogue is made and that means noise. This filter manages to kill this noise! It's not exactly the kind of the noise our eyes see: Is more about a kind of noise the encoders "see" and when they convert again our projects to mpeg 2 or whatever, they act "weird" and add macroblocks...
Did I mention it before? I hateeee macroblocks… So, I try to keep them to a minimum when I use filters like “FSN”.
- That's it with Virtualdub! It's time to hit "frameserve" and start encoding... But hey: This is not the end!
The TMPGenc plus steps
Call me old fashion, but I still use TMPGenc plus 2.5. Don’t ask me why, I just… like it!
Anyway, I have to do some more few steps manually here, that helps a lot this convertion.
- First –of course – we load the dummy file from Virtualdub. Then, we set the input source as 16:9 display (not 16:9 PAL). You may notice, that TMPGenc 2.5 detects the frameserved file as interlace for some reason, while it is obvious progressive now. Well, leave it that way, it doesn’t harm something.
- Then we go to the filter’s Tab and we select the “sharpness” filter. We set the HORIZONTAL slide to a value 62 and less but we left the vertical slide to zero. That tiny step improves great the output picture, especially if your original DVB source was something like 544 x 576 (the rule of the DVB broadcasts in Europe those days). If your source was 704/720 x 576, you may not need to add horizontal sharpness.
- Now a step that it is not necessary, except if you are perfectionist! As you know, TMPGenc allows you to not really crop the input source, but masks a selected area you determined. To be able to do this, you need to set the “keep the aspect ratio (2)” method on the output. I use this feature to mask (crop…) the left and the right black bars, when they exist. At the same time, TMPGenc “centers” the picture, which on some cases, it might be a very good thing.
That step helps slightly the overall picture quality, since no bitrate is wasted to those masked boarders (it is a “trick” that has it roots back to my VCD/SVCD, when every kb/s counted…). Also centering the picture is a plus.
So, in case you left some boarders during the virtualdub cropping so to retain the picture aspect, you may use this “trick” so to mask them finally for sure and center the picture 100%. After all, why to waste any amount of bitrate for black boarders?
- OK, it’s time to hit “encode”. Just remember to set the output to “progressive” (your 16:9 LCD/Plasma TV gonna love this!). And check using the “preview” of TMPGenc if everything is OK!
On my brand new Dual 2 Core 6600, a 4 min video is ready in about 14 minutes: Blame Virtualdub for this. But I set and encode 2 files at the same time with TMPGenc, so in practice I have two 4 min videos at 15 min for real: It’s like 7.5 min each if you think about it! Not bad, indeed!
Very rarely, when the source of the TV channel is obvious a VHS tape which digitised on the fly "as is", I use the amazing rmPAL filter, which kills the well known VHS phase alteration colour shift (a PAL issue). Always add this filter second to the msu smart deblocking one!
msu_smart deblocking: http://compression.ru/video/deblocking/smartdeblocking_en.html
Alparysoft Deinterlace: www.alparysoft.com
Deinterlace "Muksun": http://acobw.narod.ru/deinterlace.html
Progressive Frame Restore: www.geocities.com/siwalters_uk/autodint.html
Time Based DeInterlacer V1.0: http://www.uwe-freese.de/software-projekte/virtualdub-filter/TBDI_Doc/index.html
Static Noise Reduction: www.shdon.com
Dynamic Noise Reduction: www.shdon.com
Frequency Suppressor of the Noise: http://acobw.narod.ru/fsn.html
Video DeNoise 2.0: www.risingresearch.com
rmPAL: www.barzgaran.de/files/rmpal1.zip (Thank you nbarzgar!)
P.S.: A crazy idea
Since the output is progressive at 720 x 576, mpeg 2 is not necessary! You can use mpeg 1 for your output of your projects. Then, re-mux those mpeg 1 files as mpeg2 and “patch” their headers with DVDPatcher to 16:9, so to trick your Authoring application! Just remember to follow the GOP limitations the DVD Video standard determines during the encoding, exactly as you do with mpeg 2.
Using mpeg 1, might help those who wish to do DVDs in a totally freeware way, since the MPEG-1 free encoders are plenty! I don’t know if that idea works and I don’t have the patience to test it, but the theory says that ain’t gonna be any problem!
© SatStorm for VideoHelp.com
Revisited Version – 23-05-2007
First Version: 1.0 – 18 Dec 2006
+ Reply to Thread
Results 1 to 30 of 31
La Linea by Osvaldo Cavandoli
You refer to "Demon's video Enhancer" - it is not in the list of links at the bottom - ?
May be talking about this: http://www.xentrik.demon.co.uk/ ?
FWIW… Something I’ve played around with but only found sometimes useful is changing the size flags of mpg2 video. Doubling the height using just flags causes the video to be rendered at 200% in something like WMPlayer, but in V/Dub the original width is maintained, & deinterlace can then be done using resize, Chris LaRosa’s 2:1 vertical reduction, splitting the fields (like with V/Dub’s deinterlace filter options) etc. Changing the flags you’re not adding anything to the video – you’re just using different code to get to the same place as a deinterlace filter – so depending on your source it might help… might not. Since I use DGIndex & AVISynth to get the mpg2 into V/Dub, there’s also the opportunity to add extra filtering (not available as a V/Dub plugin) in the .avs file.
That said, in the US with NTSC it’s not always worth the bother going anamorphic to remove the letterbox, since it can be very difficult to get decent progressive frames. Sometimes you can remove the extra frames inserted to get 29.976 fps by using IVT (like in V/Dub), but a lot of broadcast content isn’t based on film’s 24 p. And progressive frames that show a lot of ghosting [blended motion from separate fields] can make your video seem to stutter.
nice guide SatStorm.....
I have som thoughts about the conversion....
I some times do this conversion but I use Virtualdub (MOD) AviSynth and HC encoder.....
I live in PAL country so I wil talk about this format
I record alot from sat... Iam using Dreambox and CIFS..
And many prodivers use the format 544x576 and some even smaller.
This problem is solved by resizing the video to 720x576 (if the video is in 4:3) format.
This is done with a avisynth script and a MPEGDecoder.dll by NIC..
The script is loaded into HC encoder...You can always try to change the aspect flag
but to get rid of the problem i just encode the whole video file....
If the video is in 16:9 format but in side a 4:3 frame... I would use VirtualDubMOD.
Use the filter nulltransform to get rid of the blackbars and resize to 720x576..
Start frame server and link it to HC encoder via a small avisynth script...
There can be much talk about the quailty of the video and which filter to use.
But the quilty cannot be better than the original. So if you are capturing a VHS movie
it wont be better than the original maybe enhanced but not better...
And the video captured digtally and usally have a bitrate about 2Mbits to 6Mbits wont be better
if you set the bitrate higher than the original.
There is one situation then you should set the bitrate higher than the original and
that is when you resize a 544x576 video to 720x576.. If the original bitrate is low
you must set the bitrate higher because the videoframe is bigger and there for needs
higer data rate...
If you dont have a flat tv its better to keep the video into interlace mode....
When you convert into 16:9 widescreen there are some thing to think about...
So a fullscreen 16:9 PAL should be 720x405 pixels..
So if you video hight is smaller than 405 pixels you must leave some
black bar at the top and botton, else will the aspect be wrong...
1.78:1 = 720x405
1.85:1 = 720x389
2.35:1 = 720x306
2.40:1 = 720x300
So dont just remove the black bars if the final video hight is lower than 405 pixels...
you are forgetting to incoporate calc for PAR (PAL = 1.090909 h/v) and 720vs704 overscan padding (lose the padding before resizing, add the padding after).
so 16:9 PAL should be 720x432 (i.e. 72 pixels off each top + bottom)
- 1.85:1 = 720x416
2.35 = 720x326
2.40 = 720x320
otherwise, i follow where you're going.
- 1.85:1 = 720x416
Basicly, you can't enlarce interlace without mess up your picture. That's why de-interlace is needed. It has nothing to do with the flat screens.
I also have a dreambox but I don't use it. I prefer my old trusty nokia 9600 and I also use skystar 2. I recieve satellites from 90 East (Yamal 201) to 45 West (Panamsat 1). That's why my nickname is "SatStorm"
Resizing horizontal lines it's not an issue: You keep the source also interlace. But cropping vertical lines and scretch vertically the picture so to much 720x576 again, is a pain.... Especially with interlace sources. And because you add Vertical lines is more of an upscaling than resizing.
I use filter chains to improve, enchance and fix issues. And I prefer virtualdubmpeg2 because it works better with my dvb sources. Virtualdubmod also do the same thing, but I don't like the preview it offers plus it crashes easier in my PC, when I use filters.
Filtering DVB sources really improves the results: MSU Smart Deblocking eliminates the macroblocks and using the Video DeNoise 2.0 after the vertical resizing does miracles. I almost stopped using Static Noise Reduction!
thanks for all replies....about my thoughts about the conversion
happy new year
I have to perform the opposite task, and since I am a newbie, I wonder if you can give me any guidance.
I have some VHS recordings that I want to put on DVD, but these are 4:3 material embedded in a 16:9 frame. That is, they have vertical black strips left and right. That came about because they came from a Humax 5400 that was set up to output to a 16:9 telly, but the channel was transmitting 4:3.
So, what I want to do is to crop the black bars left and right, then resample the central column so that it fits into a standard HxV pixel count, and flag it as 4:3. I am looking for the easiest way to do that. I have available an AVerMedia DVB-S PRO that I use for Analog capture, Pinnacle Studio 8, Nero VisionExpress 6, and ProjectX 6 (whose features are mostly beyond my comprehension so far).
Once I (or you have solved this problem, there is an corollary: sometimes this channel carries 16:9 material embedded in its 4:3 transmission, which is then displayed on my 16:9 telly. So in that case, I have black bars top, bottom, left, and right, with a 16:9 image in the middle of the screen. I would very much like to resample that as well, to generate a full-screen 16:9.
Can anyone point me to relevent postings / documents / software ?
Thanks in advnace.
Good guide, but just as a sidenote it *is* possible to resize and keep the interlacing. Look at unfolding the fields side by side, applying the resize, and then combining them again. This can be done using the standard deinterlace function in virtualdub, although may require more than one lot of input / output thus increasing the time to get the result.
nick1977, in a basic way you're right. But one thing you're not taking into account is the quality differential between processors (deinterlacers/resizers/FRC/scalers) that do motion-compensated work using multiple dimensions.
IOW, the better processors look at the line(s) directly above and below, the line(s) above and below in the preceeding and succeeding fields and/or frame(s), and the pixel(s) to the left and right. Possibly even DIAGONALLY.
When you fold over the field, it loses some of its ability to do that (can no longer use nearest alternate field, since it's been moved over to the side), making it's "guess" that much less accurate.
AutoGK crop the borders but that's it. If you try to encode the result to mpeg 2 16:9 then you have to make that anamorphic. So you add lines.
nick1977, unfortunatelly, it doesn't work this way. Crop/unfold/resize/fold, make the interlace lines more fat that they have to be so to show correct to the display.
There is no way to convert those 432 lines to 576 again if you don't make your source progressive. I know it sucks, but this is it.
@Kevin.Dorrell:Grab your VHS tapes at 720x576. Then crop the boarders left and right. You gonna end with a file which is 544x576. Now resize this to 720x576 or 352x576. Then encode to mpeg 2.
For the 16:9 stuff into the 4:3 transmission that is on the 16:9 canvan (complicated eh?), do the same, but resize only on 720x576. Then follow my guide.
Don't expect much regarding quality with that source of course...
I posted a Revisited version today. Better explanation, more filters, few new ideas.
Hi! Just wondering if someone has done a neutral review of the Video Enhancer product. On the surface it seems great, but I am wary of a product where only the owner shows great improvement..
Perhaps a comparison of Video Enhancer with Digital Anarchy Resizer or Instant HD ?
@zcream: Video enhancer is a great product, but only if you have a quad core CPU. Compared other upscalers, it does a better job IMO. It's a matter of taste. Batch encoding in the last version also is a great plus!
@ Chris K: For progressive source we don't have any problem to resize any cropped image to 720 x 576. With the interlace source is the problem...
Also, adding lines is not that easy: Interpolate or create new lines from the combination of the others? Video Enhancer create new lines from the rest ones (they use the rest as reference) while what is the method I point here, just interpolate a progressive image to a new dimension.
Actually, I have to update the guide, 'cause now I know more things and tricks to do this convertion (and some new filters too). I'll do so, in the near future, when I'll have some extra time.
There is even a way to keep interlacing, with a 80% success! But when you fail, you fail misery (and you can't see the bad result until they show on TV)
SatStorm wrote:Video Enhancer create new lines from the rest ones (they use the rest as reference) while what is the method I point here, just interpolate a progressive image to a new dimension....
Actually, I have to update the guide, 'cause now I know more things and tricks to do this convertion (and some new filters too). I'll do so, in the near future, when I'll have some extra time.
Thanks for this great guide.
I tried following it and I did not understand this portion.
Just remember: You have to crop 144 total vertical lines - No more, no less! It can be more only if you also crop a bit on the side so to apply a slight "zoom". Since this step is done after the de-interlacing and before the resizing (upscaling), it doesn't harm the aspect, if you are careful!
When I finished my project and made it to dvd, I noticed their faces looked long so I messed up somewhere but the picture still look really nice, if I can just figure out how to keep the aspect ratio correct then I'm good.
Let say that the original picture is 720X576
You crop 72 lines from the top and 72 lines from the bottom. A total of 144 vertical lines.
After the cropping the result is a 720X432 picture. Now resize the result to 720X576. What end up, is an anamorphic picture. It looks wierd on PC, but it's going to look OK on your TV later.
Why I say 144 total vertical lines? Because it can be (for example) 92 lines from the top and 52 from the bottom. Or the opposite. In al cases, it is 144 vertical lines in total.
When you encode, you set the source as 16:9 and not 4:3 or whatever. In TMPGenc, also keep out from the 16:9 (PAL) or 16:9 (NTSC) choices. TMPGenc do an additional resize, that way from 720X576 to 704X576 in 720X576 canvas!
You have to set manually that the source you created is 16:9.
Overall, my guide is very outdated: I even manage to find a way so to keep an interlace source, as interlace! The success of my new methods is very high (about 90%). But, I don't have time to write again the guide.
(Currently, I'm searching for a new job). I hope I'll do it in the close future.
thanks for taking the time to explain that, now I see what I did wrong, thanks again, I look forward to your new guide whenever you decide to write it.
Great advice. However, in my first experiment attempt/project, I came up with this problem/result:
Description of the problem in words: “Diagonal distorsion of output image. “
I followed the instructions of thread http://forum.videohelp.com/topic316902.html
For doing so, I installed VirtualDub with the necessary filters.
So I ran the project:
Imported my letterboxed NTSC .avi source file.
Marked project begin/end points on time-line.
Applied the DeInterlace and msu_smart deblocking filters, plus set the cropping offset parameters. I want to get rid of all static black area, but retain all available substance, so my offsets are not just a total of 144 vertical lines, but something odd like x1=+3 ; x2=+7 ; y1=-55 ; y2=+62
Compression: Since the projected output would @lossless be 18GB from a 200MB input, I selected "Xvid MPEG-4 Codec" compression, with then an output of 300MB, quality very similar to DVD source IMO.
I entered "Save as avi" to avi file and watched the progress of the dubbing in the preview window - completely satisfactory, as I wanted it (with frames not "combed/quilted/...", but everything mostly nicely crisp & smeared).
The review of the generated .avi file is different from what the preview showed:
The distorsion with that diagonal cut was of course not intended, neither shown in the preview.
Else, it looks great.
Do you have any advice? Thank you,
Do you have any advice?
Even numbers - looks like this is it! Worked! Thanks a lot! This advice might apply to all conversions.
i have a video 4:3 480x360 how to resolution 16:9..?
Thank you to everyone for this invaluable thread.
I hope to replace my 4:3 camera heads - a Sony DXC-537AP and a DXC-327AP with 16:9 versions, but until then I've been using TMPGEnc 3.0 or 4.0 XPress to get really good results. The 16:9 footage from DVCAM or hard drive matches up well with my DSR-500WSP and HVR-A1E footage.
The process for converting is as follows:-
1. Import the 4:3 clip into TMPGEnc and choose Image 16:9 in the Clip Info Aspect Ratio box.
2. Select Filters and Crop. Take any amount from top and bottom as long as Size After Cropping equals 720x432. Also select Resize filter and ensure that Keep Aspect Ratio is unticked.
3. Select Set Output and AVI Output. In the Size Settings make sure it is 720x576, and magically the drop down menu in the Output Codec Setting will include your specific avi codec (in my case - MainConcept DV Codec).
4. Select Encode, choose the path, and voila you will have a near perfect 16:9, perfectly proportioned version of your 4:3 footage (after considerable rendering time!).
I recently bought a used DVDO iScan HD Video Scaling Engine with the hope that I could cut out the long rendering time. I set it up for 720x576 and letterboxed 16:9 with source locked 50Hz (there was no existing preset for this, as it is mainly for HD scaling), and outputted through the analogue RGB to a Grandtec Hand View VGA to S-video converter. Unfortunately, the resulting picture, although good, had a slight colour misalignement. I tried adjusting the Y/C delay on the DVDO, but it made no difference. I also tried using a Vine Micros Multigen 2 converter, but there is the same problem. Does anyone have any idea what can be causing this?