I have read all the posts I can find on HDV workflow and it sounds like my method of choice will be Jeron's - capture downconverted to DV into a 1080i project switched to 30iNTSC; edit; switch to HDV and recapture tapes in HDV, then relink.

I started an HDV project several months ago - we have only been acquiring footage up till now and have done no editing. I have been capturing the footage in downconverted letterbox format, planning to edit in SD and then recapture in HDV. But I just realized that all my titles and graphics will not be in the right aspect ratio to match the letterbox. In other words, the right format to capture in SD using this method would be squeezed (aka "anamorphic"?), correct? And view throuigh Mojo while editing, on an SD 16x9 monitor?

2nd question. When the edit is finished and it is time to relink... How does my project know to relink to the new HDV footage and not just use the old SD media files?

3rd Question. Instead of using this (downconverting) method, is there any advantage to capturing in HDV, transcoding to DNxHD and editing it all in 1080i format? I am looking for the editing format that will be fastest and most responsive. It seems that with hours and hours of original tape, transcoding all the original footage to DNxHD might be a little much. And I suspect rendering times will still be higher than SD.

 

bump

I had success editing downconverted DV and then creating an HDV project, bringing in the sequence and batch capturing. I waited to add titles until it was in the HDV version but they were only head and tails. It was a 1 hour multicam of a dance performance and the batch seemed to be frame-accurate...

rfmeredith:

I have read all the posts I can find on HDV workflow and it sounds like my method of choice will be Jeron's - capture downconverted to DV into a 1080i project switched to 30iNTSC; edit; switch to HDV and recapture tapes in HDV, then relink.

Rick,

good questions. I'm in the process of onlining an HDV documentary which followed a workflow similar to what you describe in the first paragraph. The HDV was captured in a 30i NTSC project by downconverting to DV25 in deck and offlined. The project was then re-batched into 1080i HDV and the material transcoded to DNX 220. Several comments to note on this approach:

1) I didn't check the captured DV for TC accuracy (the offline was done by someone else) but I'm somewhat confident about its accuracy. It is possible that TC and media were not in sync due to the offline digitizing process (I don't know how this was done).

2) The HDV re-capture was done by decomposing the sequence and batching in a NEW HDV project. This is preferable to batching a timeline for a number of reasons. The outcome of this process was that several HDV shots needed to be slipped to match the DV picture. Infact, while the timecode's of DV and HDV matched, the picture didn't and the HDV showed audio/video off sync indicating that the HDV picture TC reported was incorrect.  My experience with this process is that your milage varies according to the type of HDV shot, the camera/deck used and the project complexity.

3) The transcode to DNX seemed to "re-align" A/V sync resulting in synced DNX picture while retaining A/V offsync in the HDV footage. All this has led me to suspect that Avid TC is suspect when dealing with transport stream MPEG2 (i.e., non I-frame MPEG2 or intrafrane compression codecs). While I'm not familiar with the interior workings of DNX (I suspect it uses wavelet based compression), it appears to definitely be robust not only as a codec but also on a TC/A/V frame integrity basis.

The present workflow has re-affirmed my faith (at least for the time being) in an alternative workflow which you allude below in your third question. I'll continue inline below.

rfmeredith:

I started an HDV project several months ago - we have only been acquiring footage up till now and have done no editing. I have been capturing the footage in downconverted letterbox format, planning to edit in SD and then recapture in HDV. But I just realized that all my titles and graphics will not be in the right aspect ratio to match the letterbox. In other words, the right format to capture in SD using this method would be squeezed (aka "anamorphic"?), correct? And view throuigh Mojo while editing, on an SD 16x9 monitor?

That's right. The correct way to capture downconverted HD material is as anamorphic SD (assuming a later HD conform). This gives you a 1:1 AR when you upconvert to HD. I'd recommend dumping your letterbox footage and recapturing the HDV as "squeezed" (this sounds like Z1 nomenclature); assuming this is the workflow you want to follow.

rfmeredith:

2nd question. When the edit is finished and it is time to relink... How does my project know to relink to the new HDV footage and not just use the old SD media files?

Do this by switching your 30i NTSC project to 1080i HDV, then duplicate your 30i sequence, modify the sequence format to 1080i HDV and decompose it. The decomposed offline clips should reveal they are 1080i HDV clips. You can then batch them. Your duplicated sequence will automagically link to the the new media. One way to keep SD and HD media separate is by either using separate drives or by using multiple Avid MediaFiles folders. Rename your old SD folder and Avid will create a new one when you batch the HD footage. Any sequence can only relink to media it "sees"; that is, media in standard MXF Avid folder.

rfmeredith:

3rd Question. Instead of using this (downconverting) method, is there any advantage to capturing in HDV, transcoding to DNxHD and editing it all in 1080i format? I am looking for the editing format that will be fastest and most responsive. It seems that with hours and hours of original tape, transcoding all the original footage to DNxHD might be a little much. And I suspect rendering times will still be higher than SD.

Yup! This is the way you can verify TC/A/V sync and dispense with tape once captured in a first HDV pass. It also has the side benefit of avoiding multiple passes on your source tapes. The only caveat is that it is presently preferable to transcode to both DV25 and DNX (145 or 220 for 1080i/60). The former is useful for offline and allows you to edit in a standard DV25 world including proper client monitoring. The latter is used for online. Of course, DNX is of little use on AXP if you plan to output to tape. It is very useful for online pre-conforms (as I call them) to then take the project with DNX media to an Adrenalin HD or a Symphony Nitris. These two allow you to perform color correction while monitoring DNX through HD-SDI or DVI as well as master HDCAM, D5, D7 etc. Color correction and I/O is really where the buck stops when it comes to DNX HD on AXP.

Note that you only need to transcode your source footage from HDV to DV25 for offline and then just transcode the sequence clips from HDV to DNX in a conform process. This saves considerable HDD space as DNX 220 uses about 9x more space than DV25/HDV2 (both are ~25 Mbps); about 115 GB/hr.

DNX is very responsive compared to HDV (no GOP decoding in CPU required!) but, yes, keep in mind that rendering anything with 2 MP frames takes considerably longer than 0.35 MP frames (NTSC). Note also that DNX is full raster square (1920x1080) 4:2:2 while HDV is thin raster anamorphic (1440x1080) 4:2:0.

Hope this helps. David.

We are doing the later.  When we shoot HDV, we use the 1080i component HD output of the BR-50 (JVC HDV Deck) and run that into a Component HD to SDI HD converter and run that around through our patch panels to the HD suites.  We are running with 1080i upconverted from 720p as our final format to edit with.

 

Sean

It helps a lot! Great answers, just the kind of detail I was looking for. thanks.

doxilia:

Yup! This is the way you can verify TC/A/V sync and dispense with tape once captured in a first HDV pass. It also has the side benefit of avoiding multiple passes on your source tapes. The only caveat is that it is presently preferable to transcode to both DV25 and DNX (145 or 220 for 1080i/60).

BTW,

AXP v5.5 should provide us with a way to monitor full field HD on a second OpenGL DVI output. In other words, we will be able to monitor HD and not only downconverted SD through a Mojo or transcoder. In principle this allows us to work exclusively with DNX and avoids having to offline and online the footage. You could skip the transcode to DV25, transcode once only to DNX, edit (offline/online) and then take your online pre-conform sequences (with consolidated DNX media) to an adrenalin or nitris for packaging.

I guess the benefit of this approach is that you can actually watch HD as you edit (nice!) at the cost of storage space and throughput. AXP v5.5 should be similar to FCP, Vegas monitoring; I'm hoping better. I'm hoping there is some code in there that actually will give us a bcast signal on the DVI monitor; in other words, a "well" scaled full raster HD frame with frame rate accuracy. Some LCD's can be set to 24, 50 and 60 Hz providing a synced refresh/frame rate.

DNX 220 (the highest DNX bit rate) has the bandwidth of uncompressed SD (220/8 = 27.5 MBps). For snappy timeline response I currently use a 5 x SATA HDD array in a port multiplying (PM) external enclosure. Depending on your controller and interface (PCI, PCI-X, PCI-e), you can obtain as much as ~200 MBps of sustained throughput; more than ample for multiple DNX streams.

Note that while PM's reduce your peak bandwidth compared to direct attached SATA (DAS) (when the drives are empty), they deliver constant average bandwith. You end up with "sinusoid shaped" throughput for the entire array instead of "downward slope" throughput for DAS.

I found that simple dual drive software stripes didn't quite deliver for DNX HD post.

DeepBlueEditor:

We are doing the later.  When we shoot HDV, we use the 1080i component HD output of the BR-50 (JVC HDV Deck) and run that into a Component HD to SDI HD converter and run that around through our patch panels to the HD suites.  We are running with 1080i upconverted from 720p as our final format to edit with.

 

Sean

Sean,

I've done some experiments with 1394 to HD SDI converters (e.g., Miranda, Convergent Design); in principle with the advantage that these transcoders embed the audio in the HD SDI stream and provide RS-422 control over the 1394 link to the HDV device. The Miranda device didn't work very well on a DS Nitris but didn't have the opportunity to try it on a Symphony Nitris or Adrenalin HD.

What if any A/V sync issues have you had with the YPbPr -> HD-SDI apprach? I assume that you feed the RCA audio outputs from the BR-50 to the patch panels, right? Or does the converter you use embed the analog audio into the HD-SDI stream? What do you do for machine control; LANC to RS-422? How does that work? Or perhaps you use 1394 for deck control; I know this is possible in principle on an Adrenalin HD.

I'm inclined to favouring the 1394 to HD SDI converters as I suspect that the baseband decompression is higher quality than what is done by the HDV decks/cams. It'd be interesting to have someone (like Graeme Nattress) compare the two HD signals from the DDC (my approach) and the ADC (your approach).

cheers, David.

Bah I'm too late. David said it all

I'm usually at work when I'm on the forums and don't have my Avid machine to look at the specifics. I'm not experienced enough to remember all of the terms either, so forgive me for not having the correct terms (or use the wrong one).

With the end delivery format being SD DVD, I capture the downconverted tapes, then do all of the work. After burning the DVDs, I'll do the HDV masters.

My projects are typically about 8 - 10 tapes of live performances where I cut from camera to camera. So my segments using a few seconds from one tape (camera a), then a few seconds from another (camera b), then a few seconds from the first tape (camera a), and so on. That means, given two tapes, I pretty much use use the whole tapes "time-linearly-speaking".

When I'm finished with my DV/SD version and ready to work with the HDV I run a process that I can't remember the name of, I think it is "consolidate or unlink". Whatever it is, I end up with copies of my sequences with media that is off-line.

I copy them to a new bin and recapture the tapes in HDV to that bin. I then use the feature/process that "relinks" the sequences to those new HDV clips. Is it "batch capture"? There is one that captures from the tape and there is one that you can relink to media already captured. Since I manually recaptured each tape in HDV, I use the one that lets you relink to media already captured.

... I know, I'm sorry for not having the correct terms...

I've had to tweak some titles and I've even had to tweak some effects and color correction as well once I've relinked to the HDV. I've never found any rhyme or reason; e.g. sometimes I'll do an effect in DV and think for sure I'll have to "touch it up" in HD, only to find that it looks fine.

I use the Sony HDV cameras and deck and I've never had any timecode problems. I think some people have reported problems. If the relink to HDV had timecode problems that would be a total nightmare. The only timecode problems I've had were ones that I expected, like a known bad spot on the tape, which crops up before I even start editing on the initial captures.

I also haven't really cared much about the HDV master, since I don't need to deliver HD. If I did, I would do things a little differently.

First and foremost, I would never, ever work with the native HDV. I would always work with the HD in DNxHD format. The DNxHD format looks great (I can't tell a difference with my eyes) and any boost in performance will be greatly appreciated.

I would try this workflow if I had to deliver HD...

1. Capture HDV

2. Transcode to DNxHD

3. Transcode to DV

4. Do as much off-lining in DV as I can as far as effects, transitions and color correction. This is purely to save time rendering as you work. 80% of my timelines have some sort of effect and/or color correction (since I'm the world's worst cameraman), so rendering kills me when working with HDV.

5. When you've gone as far as you can with the DV, relink to the DNxHD

6. Work with the DNxHD format to clean up in HD any titles or color correction, etc.

7. I think finishing in DNxHD should be fine, but you can always relink to the original HDV to finish.

Now I know 1, 2 and 3 are a huge time-sink as well as media-sink, since you'd need a lot of disk space, so it may not be practical. I know if I had to deliver an HD project, I would give this workflow a try. Step 2 and 3 can be done at different times as well. After doing step 1, you could do step 3 to get DV to work with, then when done, trash the DV, transcode to DNxHD and relink to clean up everything and finish.

Right now I only capture twice because it takes 3x real-time for my system to transcode to DV, so it only takes me 2 hours of capturing to get a 1 hour tape into both HDV and DV format. I think the transcoding would be the best bet as opposed to how I do it.

Have fun rendering! (aka get a good book... I'm enjoying the Song of Fire and Ice series by George R.R Martin ATM) 

Jeron

JeronCoolman:

Bah I'm too late. David said it all

I'm usually at work when I'm on the forums and don't have my Avid machine to look at the specifics. I'm not experienced enough to remember all of the terms either, so forgive me for not having the correct terms (or use the wrong one).

With the end delivery format being SD DVD, I capture the downconverted tapes, then do all of the work. After burning the DVDs, I'll do the HDV masters.

Jeron

Hey Jeron,

I seem to recall discussing things in another thread. The one thing I don't understand about your HDV - > DVD workflow is that you seem to shoot HDV, then capture as downconverted DV and proceed to compress your MPEG2 from this footage. If this is the case, you'll be surprised by the difference in quality if you compress from HD sources. While you benefit somewhat from the oversampling of HD when capturing DV, the image is grossly different from compressing SD MPEG2 from HD. I also don't understand what you mean by "After burning the DVDs, I'll do the HDV masters."

Whether you go to HD tape or SD DVD you'll find the DNX workflow to be advantageous. Capture your footage as HDV, transcode to DNX (145 is sufficient for DVD), edit, finish and export your sequences as QT reference. This will create full raster (1920x1080) Quicktime files linked to the Avid MXF media. Drop these QT files into whatever you use to encode DVD MPEG2 and compress. I like Main Concept as it is top quality and extremely fast. Compressing from DV NTSC typically takes 0.7 RT on a modern CPU (2.8 GHz or faster). This is of course faster than a hardware encoder as these are realtime. Encoding from an HD source will take longer but the images are pristine. I can't say the same is true from in camera/deck DV downconverts which by the way are limited to Sony HDV since JVC doesn't provide this capability (don't know about Canon).

If you haven't tried this, give it a shot, you'll be pleasantly surprised.

BTW, DNX 145 uses about 65 GB/hr or 650 GB for 10 tapes which you mentioned. If this is too much storage consumption for your projects, you can transcode to DV for offline and just transcode the final sequence footage to DNX (as mentioned in my earlier post). Between adding storage and transcoding to DV I would opt for the former; in the end one's time is more expensive than hardware.

I pooled 5 x 120 GB hard drives I had used in different machines, converted them to SATA and threw them into a port multiplying enclosure. I have two Avid machines; one with a standard PCI 33 bus and one with a PCI-X 133 bus. On the former I get ~70 MB/s throughput on the latter, ~180 MB/s. This is with some old ATA 100 drives. Throw SATA II drives into the mix and one's well over 200 MB/s. These things are on my mind as of late as I'm interested in multiple stream DNX or UC SD capability in light of the new Media Composer announcements.

Anyway, on a final note; your point 7 is something to definitely avoid. Decompressing HDV to DNX is a good thing. The opposite should be avoided if possible. DNX is a post codec; HDV is an acquisition codec and moreso than any other acquisition codec I know of. If mastering to HDV is absolutely necessary, I would be inclined to mixdown the DNX video to an HDV stream rather than relink. Having said that, relinking will probably be faster in v5.5 as the update should allow us to rebuild only the GOP's that have been affected. Of course if the sequence has effects (such as CC) on shots throughout, one's back to square one.

On a note of rendering and mixdowns, note that it's possible to import AAF projects into Vegas 6 and Premiere Pro 2 (I believe, don't use this one much). In Vegas, you can render the DNX to either Sony YUV or HD uncompressed and master this directly via HD SDI with a Blackmagic card (Vegas, PP) or with something like a Matrox Axio LE. The whole thing is rather tedious compared to an Adrenaline or a Nitris but it's a "home backed" solution. Matrox gives us HD SDI for $4000, Blackmagic gives us HD SDI for $1000, Avid gives us SD SDI for $2500 but HD SDI for $10000! I really wonder how an Adrenalin box differs from an Axio LE box when it comes to performance (I haven't tried the latter). They essentially provide the same I/O capability. I guess Avid charges $4000 for hardware and $6000 for the brand!

But I veer off topic.

Cheers, David.

<I pooled 5 x 120 GB hard drives I had used in different machines, converted them to SATA and threw them into a port multiplying enclosure>

David - What do you mean "converted them to SATA?" Is there a way to take an old ATA drive and convert it to SATA?

Check me on this...

My current project has about 20 hours of source material. I am capturing it as downconverted DV25 anamorphic and will edit at DV25. My final product will be a 16x9 SD DVD. I want the best image quality of course.

It sounds like you are saying there is a better way than just using the final DV25 output for the DVD. ("You'll be surprised by the difference in quality if you compress from HD sources")

I think I really need to edit my project at DV25 (consider it an offline format in this case) because it seems clear that will use the least disk space and be the most responsive. But when I'm finished, to get the best quality DVD, I assume I should batch capture my timeline from HDV, relink, then transcode just those clips to DNX 145 (then relink again?), export as QT ref, transcode to m2v (I use TMPGEnc) and make my DVD.

Am I getting this right? Do I have the all the steps.. and in the right order? I'm a little unsure about the batch capture from the timeline, relinking, then transcoding to DNX and relinking again.

When I transcode to DNx, since I will be doing little editing from that point on, will I still need RAID, or can I use a single drive?

Does transcoding to DNX also give me 4:2:2 colorspace? According to Convergent Design's "HDV Editing Made Simple", converting to DNxHD does.

 

David,

I think we are in agreement 100% on all points. You are just articulating it far better than me.

doxilia:

I seem to recall discussing things in another thread. The one thing I don't understand about your HDV - > DVD workflow is that you seem to shoot HDV, then capture as downconverted DV and proceed to compress your MPEG2 from this footage. If this is the case, you'll be surprised by the difference in quality if you compress from HD sources. While you benefit somewhat from the oversampling of HD when capturing DV, the image is grossly different from compressing SD MPEG2 from HD.

Keep a couple of things in mind about my workflow. I don't get paid for my work and I deliver SD DVDs My biggest problem with HDV is the render times, so I substitute some quality for a smoother workflow. To put that in perspective, a 5 minute segment with a simple PIP w/ two streams of video takes me approx. 10 minutes to render in DV. The same segment, once relinked in HDV, takes 13 hours.

In fact rendering in HDV is so incredibly, painfully slow, I keep thinking I'm doing something wrong. I can't really get a much faster machine than my dual Xeon 3.6.

With that said, I agree with you. The compressing native HDV or DNxHD into the footage for the DVD is a better quality than the downconverted DV.

doxilia:

I also don't understand what you mean by "After burning the DVDs, I'll do the HDV masters."

What I mean by "do the HDV masters", is all of the recapturing, relinking and re-rendering of the segments in HDV, then exporting the output to an HDV device. I just do that for the projects that I'm really happy with that I want a nice HDV version of on tape, so probably half of the work...

doxilia:

Whether you go to HD tape or SD DVD you'll find the DNX workflow to be advantageous. Capture your footage as HDV, transcode to DNX (145 is sufficient for DVD), edit, finish and export your sequences as QT reference. This will create full raster (1920x1080) Quicktime files linked to the Avid MXF media. Drop these QT files into whatever you use to encode DVD MPEG2 and compress...

...If you haven't tried this, give it a shot, you'll be pleasantly surprised.

I haven't played enough with DNXHD codec to do some timings on rendering. I am planning on using it for my next HDV project that I edit with Xpress Pro. If I'm looking at a 13 hour render for every five minutes of a segment, I would be disappointed. I expect it to render far faster than native HDV, but not as fast as the SD + Mojo combination.

The question is, "Is it going to be fast enough to not interfere with my workflow?"

This is why I mentioned my second workflow. I think I'll capture HDV, transcode to DNxHD and throw away the native HDV. Try using it. If it is too slow, transcode to DV and use it. Then I'll just relink to the DNxHD when it comes to "Export to HDV Device" after the DVDs are made. (I'll only transcode to DV to play with timings. So far I've found recapturing downconverted DV works better for my needs.)

I just wish I could capture into DNxHD

Every single NLE I've worked with, Xpress Pro, FCP, Vegas and Premiere Pro all say "Here you can edit native HDV... but you really don't want to, you want to capture or transcode into an intermediate codec and work with that instead."

I think editing native HDV is just some "marketing feature" where it is more important to support it because the competition does, but no one would ever want to actually do it because it is so painful.

doxilia:

BTW, DNX 145 uses about 65 GB/hr or 650 GB for 10 tapes which you mentioned. If this is too much storage consumption for your projects, you can transcode to DV for offline and just transcode the final sequence footage to DNX (as mentioned in my earlier post). Between adding storage and transcoding to DV I would opt for the former; in the end one's time is more expensive than hardware.

This is what I was trying to say in my second workflow, just not as well.

doxilia:

Anyway, on a final note; your point 7 is something to definitely avoid. Decompressing HDV to DNX is a good thing. The opposite should be avoided if possible. DNX is a post codec; HDV is an acquisition codec and moreso than any other acquisition codec I know of. If mastering to HDV is absolutely necessary, I would be inclined to mixdown the DNX video to an HDV stream rather than relink. Having said that, relinking will probably be faster in v5.5 as the update should allow us to rebuild only the GOP's that have been affected. Of course if the sequence has effects (such as CC) on shots throughout, one's back to square one.

I agree that #7 is only going to be something to try if the v5.5 update proves to work nicely. I think in most cases, your best bet is just going to output from the DNxHD codec. We are just talking about a "fire and forget" process at the end anyway. It isn't like creating some effects and having to wait hours for it to render so you can play the clip to see how it looks and flows with the rest of the segment.

doxilia:

On a note of rendering and mixdowns, note that it's possible to import AAF projects into Vegas 6 and Premiere Pro 2 (I believe, don't use this one much). In Vegas, you can render the DNX to either Sony YUV or HD uncompressed and master this directly via HD SDI with a Blackmagic card (Vegas, PP) or with something like a Matrox Axio LE. The whole thing is rather tedious compared to an Adrenaline or a Nitris but it's a "home backed" solution. Matrox gives us HD SDI for $4000, Blackmagic gives us HD SDI for $1000, Avid gives us SD SDI for $2500 but HD SDI for $10000! I really wonder how an Adrenalin box differs from an Axio LE box when it comes to performance (I haven't tried the latter). They essentially provide the same I/O capability. I guess Avid charges $4000 for hardware and $6000 for the brand!

<sigh> Unfortunately I agree. Which is I have Vegas and why I just took delivery of Adobe's Video Bundle and am running a tutorial DVD each night to learn it

Bottom line...

If you have to work with HDV, do it transcoded in DNxHD format. If your workflow/style of working involves a lot of effects and rendering, off-line it in DV/SD and a Mojo, then relink it to the DNxHD media to output the HD deliverables.

The best quality you will get for the DV/SD will be to transcode the HDV or DNxHD to it rather than capture downconverted, but if your goal is to deliver HD, then this wouldn't matter as much.

Rick,

The best image quality I've seen was capture HDV, edit HDV, then output a windows media file or qt movie to ProCoder/Sorenson for encoding. This quality is better than outputting the DV25 to the encoder for DVD.

But working with native HDV isn't fun because of how long it takes.

I think working with downconverted DV25 (like you said) will make the editing experience a lot better. Then when it comes to getting the best picture quality, you should relink it to the HDV or DNxHD version before outputting to the encoder will give you the best quality and editing experience. I don't do this because the DV25 quality is "good enough" for me.

Depending on how much you have to work with the footage once it is relinked to the HDV or DNxHD version, you may not need a DNxHD flavor of it. When I look at the HDV versus DNxHD, I don't see a difference, but I have very "untrained" eyes.

I think you have the right idea. If I were you, I would try a couple tests with a 10 or 15 minute segment!

Jeron

P.S. If you decide to relink to the HDV before outputting to your encoder for the DVD, you might want to consider capturing HDV and transcoding it to DV to do the editing, then you can just relink to the original HDV before outputting to the encoder.

I am currently finishing a 7 minute short shot on HDV.  I captured circled dailies to HDV, then transcoded my selects to DNxHD220.  I then edited the DNxHD footage and it worked very well.  All I have is a 2 SATA drive stripe and it worked fine for single layer editing (didn't try multilayer work).

Here is a question though.  I color corrected, magic bulleted and all that, then exported a Quicktime Ref. for Sorenson Squeeze.

For those of you using this workflow, what field dominance do you guys select for mpeg-2 compressions (top or bottom)?

I am fairly certain DNxHD is upper/top.  I have noticed that the footage, once on dvd, looks as if it has been turned somewhat progressive or de-interlaced (no jagged edges though).  And this is counting the fact that I actively turn off de-interlacing in Squeeze (knowing that it has it turned on by default).  Regardless, I acually like the fact that the footage looks less "videoy", but I am still concerned about where in the process this happens.