Looks like no one’s replied in a while. To start the conversation again, simply ask a new question.

Bonsai With Pro Res

Lately, I've been shooting HDV 1080i 60 footage with my V1U, and capturing it into FCP 6.0.2 using Pro Res. I then edit and render my timeline, and export the timeline directly into Compressor, using the Best Quality 2-pass DVD settings for whatever time limit I need. These files are imported into DVDSP, and I then produce SD DVDs.

This is all standard procedure, mentioned often in this and other forums. My SD DVDs have been looking OK, but not great. Especially not great on my Sony 60" SXRD TV, which shows every flaw. But, my customers have been happy, so maybe I'm picky.

Today, I decided to retry the now 2-year old Bonsai method (without the blur), with some modifications.

After editing (but NOT rendering) my Pro Res timeline, I created a new sequence. I used the (Matrox) MXO DV50 NTSC 48khz Anamorphic sequence preset, DVC PRO50 NTSC compression setting, lower field dominance, and used the Bonsai-recommended 'faster(linear)' motion filtering setting under the sequence settings video processing tab.

I copied my completed Pro Res sequence into this new timeline, and then exported it directly into Compressor as I've usually done. Used the stock 2-pass preset. After building and burning a DVD with the resulting files in DVDSP, I popped it into my DVD player, and watched it on the SXRD.

There was a huge difference; much improved in every way. Far less pixelation, jagged lines, motion artifacting, etc. I also played it on a Sony 36" CRT television, and the improvement was dramatic there, as well. I can't believe I hadn't tried this workflow before; it's a significant enough difference that I will be using this method from now on, or at least until something better comes along. standard DV footage.

I noticed that after Compressor has finished its encoding, the timeline will begin rendering, probably because Auto-Render is checked in preferences; this rendering seems to take a very long time, but it can be cancelled, since the compression has finished.

I'm looking forward to redoing a few earlier projects to see what improvement there will be.

One question, though, for those that have made it this far: I'm not familiar with DV50; what is the difference between it and the regular DV NTSC anamorphic settings? I'm planning on trying some samples using that setting to see if there is any difference in quality.

Mac Pro 2.66/12GB/X1900XT/Sonnet E4P/30"+23"ACDs, Mac OS X (10.5.2), 2 Sonnet Fusion 500Ps/14 WDRE2 500GB/Matrox MXO

Posted on May 2, 2008 8:34 PM

Best reply

Brian Conner

Level 6

16,030 points

Posted on May 3, 2008 4:48 AM

+One question, though, for those that have made it this far: I'm not familiar with DV50; what is the difference between it and the regular DV NTSC anamorphic settings?+

DV50 is twice the bandwidth and it's a 4:2:2 codec. That means red & blue (chroma) channels have twice the samples, resulting in smoother edges.

View in context

35 replies

May 3, 2008 10:07 AM in response to hanumang

I'll second that, and add this: regardless of what Apple chooses to call it, the correct term is Field Order. It isn't Field Dominance, it's Field Order.

If you were doing a linear edit, rather than non-linear, then Field Dominance would be an issue, but it would be handled by the edit controller, so no worries.

Field Order concerns which of the two fields of each frame of video is recorded and displayed first. If the order of the fields is reversed (from what was recorded) when displayed, then you will see stuttering in areas of movement during playback.

May 3, 2008 10:12 AM in response to hanumang

Thanks again for the replies.

I'll be gone for the rest of the day and have no time to test a short segment, so I'm running this whole timeline through Compressor again, 1 hr 45 min. worth.

Original sequence, ProRes (from HDV 1080i 60 footage); going to Compressor directly from the timeline, starting with the DVD Best Quality 120 minute setting.

Adjustments I've now made are to change dialog normalization to -31, and to go in to frame controls and manually turn them on; leave all defaults except bump deinterlacing up from good to better.

The more I think about it, the deinterlacing might have been the original problem; this footage is of a stage play, and there is a staircase with a 45 degree railing that is in the back center of the stage; of course, it doesn't move, but as people move past it and around it, I sometimes see jagged edges on this railing; it's almost like the railing becomes a miniature set of stair steps by itself. If this is noticeably improved, I'll know the deinterlace setting change worked.

This will probably take awhile; the audio has just finished and I'm seeing a remaining time of 36 hours for the video, so I'll post back when I get a chance to see how this looks.

May 3, 2008 10:20 AM in response to Red Jacket Mike

Michael,

Usually, when we're testing different settings, we'll mark a 15 to 30 second portion of the Timeline, the most difficult section, and compress only that, rather than the whole thing.

FWIW.

May 3, 2008 10:33 AM in response to Brian Conner

I usually do that as well, but since I'll be gone for the day, and tomorrow's Sunday, I did the whole thing. If it works well, then I won't have to redo. Thanks for the help.

May 3, 2008 10:46 AM in response to hanumang

If I'm using RJM's workflow in FCP: HDV 1080i60 transcoded to ProRes 422, working in a ProRes time line with ProRes rendering, is there a reason for "not rendering" and sending video directly form FCP to Compressor, rather than rendering, exporting a reference movie (which points to the render files), then using compressor to deal with those?

My experience with sending directly from FCP to compressor is that it is much slower than rendering, exporting (reference), then compressing from that. Apparently with two processes running, each small group of frames requires a process switch which slows everything down more than 2x.

Now, I've not done this in a while, so maybe FCS 2 (FCP 6, compressor 3) are better at it? But once everything is transcoded to ProRes prior to editing, the only "rendering" needed should be on transitions and where you've altered color, opacity, etc. or applied other effects.

I'm likewise going to SD DVDs, for longish productions of stage shows (let's hear it for bad lighting for video!). I'm considering going to 1080p30 for the next one, but would like to have more knowledge of best practices.

Eddie O

hanumang

Level 6

9,250 points

May 3, 2008 11:50 AM in response to Edward A. Oates

In case you're unaware, exporting unrendered timelines via File > Export > Using Compressor means Compressor receives raw video frames, along with Filter and Motion tab settings, that it (Compressor) then directly processes into the MPEG-2 file.

Before Final Cut Studio 2 brought us Color and ProRes, many folks were left to perform serious color corrections using FCP's 3-Way Color Corrector and exporting unrendered timelines allowed us to eliminate the rendering artifacts inherent to compressed codecs like DV. Also very beneficial for titling and compositing, but I saw the greatest benefit with DV color-correction myself.

Yes, it takes twice as long when compared to a rendered timeline (sent directly to Compressor), since the rendering is, in effect, performed by Compressor. Never mind that it is infinitely slower than sending a QT movie. Also, and this is a huge downside for those of use with 4+ core machines, you can't use virtual clustering.

But as I said, this was a bigger deal prior to FC Studio 2. Now that folks have Color in their toolset, fewer folks use FCP for those types of hardcore adjustments. But, still, if your work is filter-heavy and/or you really need to crunch your bitrate down (lower than 4.0 Mbps average/constant), you'll see some benefit to this technique.

Hard to say if it's worth the extra time though. As always, your mileage may vary.

May 3, 2008 1:24 PM in response to hanumang

hanumang wrote:

+stuff removed+

Yes, it takes twice as long when compared to a rendered timeline (sent directly to Compressor), since the rendering is, in effect, performed by Compressor. Never mind that it is infinitely slower than sending a QT movie. Also, and this is a huge downside for those of use with 4+ core machines, you can't use virtual clustering.

But as I said, this was a bigger deal prior to FC Studio 2. Now that folks have Color in their toolset, fewer folks use FCP for those types of hardcore adjustments. But, still, if your work is filter-heavy and/or you really need to crunch your bitrate down (lower than 4.0 Mbps average/constant), you'll see some benefit to this technique.

Hard to say if it's worth the extra time though. As always, your mileage may vary.

So are you saying that those of us on older machines might be better off doing what many of us have been doing in 4 & 5, which is exporting the timeline as a ref or self-contained mov? I have had the best success with self-contained mov's myself. I don't do effects but may at some point.

Thanks,
Robert A. Ober

May 3, 2008 2:01 PM in response to Red Jacket Mike

Well, I'm back; event cancelled due to a . . .SNOWSTORM! Amazing weather we have up here.

Anyway, I did a test with a 2 minute segment using the settings I mentioned above. The resulting SD DVD looks the sharpest of any method I've tried, due to the 'better' deinterlacing setting in frame controls. Most of the jaggies are gone, too. It looks very good.

The downside is, I am seeing a small amount of jitter during any kind of movement; it's not bad, but it's there. Just a slight jerkiness occasionally. Could this be a field order issue? I guess I'll have to try another test and try switching to lower.

Any other suggestions for eliminating this? I don't know that I'd want to try the 'best' deinterlace setting; encoding would take forever. i don't think this type of jitter is due to deinterlacing, anyway, but I'm not sure.

hanumang

Level 6

9,250 points

May 3, 2008 2:16 PM in response to Robert Ober

Robert Ober wrote:
So are you saying that those of us on older machines might be better off doing what many of us have been doing in 4 & 5, which is exporting the timeline as a ref or self-contained mov? I have had the best success with self-contained mov's myself. I don't do effects but may at some point.

Not sure if I can frame it accurately as a hardware thing. In my eyes, it's more about the codec involved and whether rendering out to that codec injects an unnecessary drop-off in quality (which is a yes for DV and native HDV, for instance) depending on what you're doing.

If your work is not effects heavy, I can't see exporting unrendered timelines as being worthwhile.

hanumang

Level 6

9,250 points

May 4, 2008 11:57 AM in response to Red Jacket Mike

Red Jacket Mike wrote:
The downside is, I am seeing a small amount of jitter during any kind of movement; it's not bad, but it's there. Just a slight jerkiness occasionally. Could this be a field order issue? I guess I'll have to try another test and try switching to lower.

Just so we're not caught up in semantics, you're not confusing jitter (an interlacing artifact that affects even still shots) with +frame judder+ (a compression artifact that makes camera moves seem choppy), right?

If it's frame judder, it could simply be the case that there's too much motion and detail in the frames in question for MPEG-2 to handle at the given encoding settings. You might try upping the bit rate, but you can only go so high before you bump into playback issues. Instead, you might try altering the encode to Open GOP, though that works best for material that has longer shots. (If you go this route, I'd attempt to test an Open GOP encode with an entire take/shot, rather than a 2 minute snippet.)

Of course, if it's truly jitter, then yeah, you could try forcing the material to Lower First to see if that improves anything.

May 4, 2008 1:11 PM in response to hanumang

Yes, I guess frame judder would correctly describe it, although it's not apparent so much on camera moves, but when the camera is locked down and people are moving within the frame. It's very slight.

Anyway, what I've done since yesterday is to change the field order in Compressor's Frame Controls area to bottom first. This has all but eliminated the judder.

There are actually two places in the Inspector window where the field order can be changed; in the encoder, and under frame controls. Changing it to lower/bottom in the encoder makes it worse; is this because these are the settings of the original pre-encoded file being brought in to Compressor?
So I left it as upper there, and changed it to lower in frame controls, and I'm now getting much improved results. The best I've seen yet.

I've also managed to get QMaster working with virtual clusters (I have set 2 instances on my Mac Pro 2.66 quad) and the ridiculous processing times have been reduced quite a bit.

Another small question, concerning the adaptive details checkbox in frame controls; this is checked by default, but I've read elsewhere that this can be unchecked for the HDV to SD DVD type of compression I'm doing, to speed up processing without losing any quality...is this correct?

Thanks again for all your help; it's much appreciated--and as I continue to fine tune this process, the snow outside is melting . . .

SSteele

Level 3

515 points

May 4, 2008 1:55 PM in response to hanumang

Tried the 5.5 bit rate CBR and no difference. Found it plays OK on 2 Sony DVD players and not OK on 2 other Sony DVD players, not necessarily the newest ones working and the oldest ones not. Have been reading Mike's variations to see if I think will solve my problem and don't think so, so will start a new thread if needed.
Thanks Mike, this thread has been helpful.

hanumang

Level 6

9,250 points

May 4, 2008 2:42 PM in response to Red Jacket Mike

The Encoder tab uses QuickTime's methods (for stuff like altering sizing, timing, field order, etc). I don't think I'm breaking new ground when I say that Frame Controls (because they are based on much more advanced technology) are a far superior way of doing things. So, yes, your experience here is correct: if you have to make an adjustment like that, do so using Frame Controls.

And since you're downconverting, yes, you can uncheck +Adaptive Details.+ In my experience, you'd only want that box checked (and the options adjusted) for SD-to-HD upconverts.

HDV is so weird to work with - I've never come across another instance where altering the Field Dominance - or is it supposed to be Field Order? does anyone know? 😉 - actually produces better encodes for DVD. Go figure...

hanumang

Level 6

9,250 points

May 4, 2008 2:43 PM in response to SSteele

Does it always choke at/around the same spot(s)?

May 4, 2008 2:56 PM in response to hanumang

I had read an article a few months back on switching field order (according to earlier posts in this thread it is order, and not dominance) in Compressor to improve the image, but it didn't work for me, since I hadn't discovered the frame controls yet, and just switched it in the encoder. The result was worse; but now, switching it in frame controls has made a significant improvement.

I held up on marking this solved, since Sharon was still trying to solve her problem, but she may now start a new thread, and I'm off to the QMaster forum to look for advice on speeding this process up even more, so solved it is. . . thanks.

Bonsai With Pro Res