HD to SD again
 

I shoot with an EX3 for broadcast television, who only accept standard definition. I am, after many months, still struggling with the best way to downconvert the footage - using a MacPro Laptop. Shooting with 1080/50i and using MpegStreamclip I find that the resulting pictures are softer than I think they should be, and do not compare well with Digital Beta, particularly if I use the “better downconvert” setting to remove artefacts . I’ve tried Compressor at various different settings and the result is even worse.

I tried 720/50p recently, and the pictures look much better - I think because there are fewer artefacts coming from the lower resolution. However any pans or moves look horrible, very stuttery. When I look at the pans frame by frame there’s no visual evidence of interlacing, the sort of tearing pattern that you get on edges with a 1080i downconvert. Yet the BBC website recommends 720p as the best format for standard definition because they say each of the 50 frames becomes one of the fifty fields per second in PAL. But it seems to me that fields in my downconverts arise by dropping every other frame and making two fields from the one frame, if you follow my drift. So I acctually end up with only 25 frames in effect, and that's why it's stuttery. I see the same thing when I use the (inferior) Compressor. If I could just get one field from each of those frames I’d be happy.

I’ve read and re-read most of the posts relating to downconversion but I’m still puzzled. So, what are other people’s experiences with downconverting from 720p? Would hardware like the Matrox MXO improve things?

I use both MX02 and AJA for down converting in real time as I am editing, different suites.

I have a JVC DT-20 HD reference monitor with 1080 i or p feed over HD-SDI, showing the HD version.

And I have a Sony PWM-20MPU SD reference monitor for SD, and I have a few consumer Plasma and LCD HD monitors as well 720 and 1080. I can input SD or HD to those.

I pay real close attention to how the Sony interlaced monitor looks if final product is SD. I also have Tektronix scopes running full time while editing. And it helps to watch SD on the consumer monitors as well.

When everything is done I output to BETA SP through the AJA or Matrox for the SD version. I often make both SD and HD versions of my projects.

Using this method I know exactly what I am getting full time while editing and I spot any trouble with interlacing or illegal output right away.

I find the images from the EX cams to be excellent converted to SD this way. I have a lot of legacy projects in SD I add HD content to. Then I use M100 to convert the HD to SD and those results are excellent as well. I cant tell the difference from the old and new footage.

Richard, this is an old chestnut. But this is my method and it's the best one I have used. In fact I use this very method on an almost daily basis.

I shoot 730P50. Bring the footage into FCP.Once you have done your edit, make a new SD sequence. DVPAL5:4, Edit timebase 50 frames, 16:9, Compressor DV-PAL, click on advanced, Motion frames per second =50, 16:9, interlaced ( or not up to you) interlaced to DVD.

Drop sequence 1 ( the HD sequence) onto sequence 2 ( the SD sequence. FCP will ask if you want top change th sequence, say no. You will then need to render out the sequence 2.

Then send this new SD sequence to Compressor. Choose Frame Controls - better. If to DVD choose upper field first. if not progressive.

Do a 30 second clip test. Its the best results I have had. Hope that helps.

Are you downconverting the final edit master, or the rushes (before you edit)

Steve's workflow is good for post-edit downconverts, but it ties up final cut pro during the send to compressor? AFAIK.

If you want to downconvert rushes, compressor does is very well indeed, but you need the right settings.

Perhaps you should post your complete workflow, as it sounds like the problems you are having will concern field, frame settings.

Duncan.

Richard,

That's one of the hardest tasks to achieve with good results. Some computers can do it very well and others have trouble, like yourself. One of my associates has an EX1 and is currently stressed out by the same issues you are experiencing and he has a desktop mac. He has followed my workflow only to disappointed with the results. I actually offered to run his FCP project through my computer to see what can be achieved.

I, on the other hand, have always achieved excellent results from my 50i HD to SD conversions from V1, EX3 & PMW cameras. I edit with Media 100 and sometimes FCP on my main computer. I use Compressor & Bitvice. On my second computer I run FCP with compressor & DVDSP4. My laptop runs FCP, Compressor & DVDSP4. Sorry I cannot comment on 720 as I have never used it.

Best wishes

Hi Richard - I had the same problem about a year ago on my first project with the EX3. Have found that downconverting through the camera or the AJA Kona LHi works really well and saves a lot of faffing around with different compressors. The quality is much superior too (in my opinion) as it is a hardware downconvert. You can either downconvert all your rushes if you don't want an HD master and then edit in an SD timeline. Alternatively you can just downconvert the finished HD timeline through the Kona / MXO2. I downconvert to my 2nd suite (old G5 based system), but you can downconvert to a deck or whatever.

Ok it may take time to digitise everything, but the amount of headache trying to 'finish' it saves far outweighs this time.

Cheers, Martin.

Video and DVD Production, Corporate Video, Editor, Video Services, Business Video Production, Freelance Cameraman, DVD Authoring

Have found that downconverting through the camera .....................................

How are you doing this specifically?

Thanks,

John

Thanks everyone for your interesting and varied responses. I know the HD to SD problem is an old chestnut, but they’re the hardest to crack - if I remember correctly from my childhood days with conkers in England!

Duncan, yes, I have to convert all my rushes to SD. My workflow is ... Log and Transfer into Final cut pro onto one drive, then downconvert to DVCPro50 (required for delivery) using Mpeg Streamclip onto another drive. I generally leave the “better downscaling” unchecked to preserve better definition. I do all this at night after shooting, so the procedure has to be simple. I copy the SD files onto a third drive and this is what the editors use. If some shots have distracting artefacts I can replace the files later with the softer, but cleaner files by running them thru Mpeg Streamclip again with better downscaling.

I’ve tried all sorts of settings with Compressor, different Frame controls etc, but Mpeg Streamclip looks so much better.

My real query is with 720p which seems to give the sharpest, cleanest images on frames without motion. But what I really want to know is - what sort of hardware or software converts each one of the 50 frames into 50 fields? Both Mpeg streamclip and compressor seem to drop every other frame, then convert each of what’s left into two identical fields. Result, jerky pans.

I think I will try hardware conversion after reading Olof’s and Martins’s posts. There are Kona cards where I work, and see if the result looks better using the SDI output of the camera. I’ll post my impressions when I’ve done the tests. May be a while.

Are you looking at this footage via ClipBrowser or on a timeline? Try creating a DVD from the footage and view it on a TV, I think you should find it very smooth. I have used the 720p for many DVDs without any problem.

Just for extra information. I shoot everything using 1920x1080 23.98, 25 or 30p and then convert the footage down to an AVi file using a Matrox RTX2 card, this process is accelerated, generally twice as fast as real time. The resulting quality is as good, if not better than shooting SD on my Canon XHA1, in fact I do mix footage from the two cameras in a single production - I may just purchase another EX camera at a later stage.

Go into the camera Video Set menu, then YPbPr/SDI Out - select SD
.... then under Down Converter option select Squeeze (I think!)to get picture into Widescreen.

This should give you realtime high quality downconvert through the SDI or Component outputs.

720p 50 is an excellent format for downconversions and for cross converting to 1080i. This is also confirmed by a close friend and colleague who works on HD delivery and infrastructure at BBC TC.

Cheers, Martin.

Richard,

I online a huge number of commercials at DV50 (around 500-600 a year), I drop a whole range of material direct onto an FCP timeline and let the software deal with it there. This includes old SD DV, HDV1080i, 720p and 1080p. Converting to an intermediate format for that kind of work has no real benefit as the timelines have so many layers of text, graphics and footage that it would never run realtime anyway. And having HD material on an SD timeline lets me reframe shots if needed.

But I do some work where I down-convert first, to SD DV or sometimes DV50. I use a compressor droplet for these situations and the results are good enough for me (your results may vary). I can email you the droplets if you want to compare them with yours.

Are you required to deliver interlaced video in the final format? With field movement? ie video look?

If my delivery was interlaced I would shoot in 1080i and use the droplet to downconvert to DV (25or50)

I haven't tried the 720p50 to 625i approach personally.

Duncan

OK,

I use that method to output my camera to a monitor while I am shooting, but you are saying that also outputting to a deck will give better results than down converting in post. I'll have to try that.

Thanks,

John

Richard

I convert 50 frames to 50 fields all the time and have no problems, using compressor, however the following must be set correctly.

I use Prores as your video codec many of the other codecs will only take one frame and convert it to two fields, each being identical. If this happens you can easily tell because you don't get a sawtooth pattern on edges (obvious where there is movement) when using the Quicktime Player. I think jpeg 2000 also will allow you to do a proper conversion, but the mpeg 4 codecs will not work. You then must go to the codec preset settings,set frame rate to 25 and select the correct interlace setting for your purpose. Then go to the frame controls, and choose the correct field output setting to match the codec settings that you have just set. According to your needs choose the resize settings, the better settings give a slightly better result but at a significant render time penalty. Deinterlace setting is set to fast. Again adaptive details is set to on or off. I have found that setting the adaptive details can result in a dramatic increase in render times, with little improvement in quality. Geometry is set to SD size output using the SD 601 preset. There are two filter settings which can affect your end result. The first setting is under the color tab where you set the output to the CCIR 601 color space preset. (HD uses quite a different color space therefore needs to be converted to a SD color space). I usually add a tadd of edge sharpening too.

You are now ready to interlace and downscale.

regards

Ian Skurrie