Can you mix DNxHD 145 and 220 in the same HD project. And also, which is a better format to use. Thanks for your help!!

 You can, but when you do a digital cut you will be asked to transcode everything to one or the other, and in my case after I go through the transcode process something fails, so I manually transcode some clips and it still fails, so I have to do a crash record without timecode matching the job when I print to tape. So from my perspective I stay with 220 8-bit all the time.

220 and 145 refer to the bit rate so it depends what you're shooting on (and what's an approved workflow, if any, from you client).

We shoot dvcproHD and HD cam which have bit rates of 100 and 140 respectively. This means we gain nothing by choosing 220 over 145 and just sacrifice disk space.

I've never tried mixing the two in the same timeline and laying off to tape but if you do transcode the sequence and it still wont work its probably because you have a title or matte that MC wont transcode. You'll have to reimport any mattes and recreate title media.

Red Ochre:

We shoot dvcproHD and HD cam which have bit rates of 100 and 140 respectively. This means we gain nothing by choosing 220 over 145 and just sacrifice disk space.

Not true, unfortunately.  Because DNxHD is a square pixel codec and DVCproHD is anamorphic, you can't directly compare them.  With the anamorphic stream you're compressing 1382400 pixels 30 times a second, with square pixels (DNxHD) you're compressing 2073600.  For that reason I would go for 220 rather than 145.  You will still see some slight resolution loss because of the pixel aspect ratio conversion, but it should be less obvious using 220 than 145.

In the 50 Hz world I use 185 10-bit for the same reason.  The difference is in my case that it's a comparison between 1555200 25 times a second to 2073600.  In theory to acheive the same compression rate as 220 in the 60 Hz world we should use 183.33, so I guess 185 is better!

jwrl:

Because DNxHD is a square pixel codec and DVCproHD is anamorphic, you can't directly compare them.  With the anamorphic stream you're compressing 1382400 pixels 30 times a second, with square pixels (DNxHD) you're compressing 2073600

jwrl can you expand on this? Since DNX is full raster and DVCPRO is not wouldn't 145 be sufficient? How can pixel aspect make that much difference?

I realize it gets tricky directly comparing just one aspect like bit rate (ie. dv and HDV are both 25mps) but why would avid engineer a codec for HD delivery that was inferior to DVCPRO HD? If this is the case why is 145 accepted as delivery by most cable networks?

Red Ochre:

jwrl can you expand on this? Since DNX is full raster and DVCPRO is not wouldn't 145 be sufficient? How can pixel aspect make that much difference?

Pixel aspect ratio makes a difference because it governs the number of pixels that need to be handled.  Other things such as the algorithm used come into play too, which makes it hard to directly match two different codecs as well.

Red Ochre:

why would avid engineer a codec for HD delivery that was inferior to DVCPRO HD?

They haven't.  Arguably DNxHD is considerably superior to DVCpro HD especially if you use the 10-bit versions.  However when dealing with DNxHD as a means of handling transcoded media it will inevitably be better to go with lower comression rates (higher numbers) to minmise artifacts and resolution loss.

Red Ochre:

If this is the case why is 145 accepted as delivery by most cable networks?

What cable networks require can hardly be regarded as the benchmark standard for image quality.  As far as the average cable operator is concerned, the more channels you can squeeze into a given space the better.

Looking just at the numbers is a surefire way to make poor decisions, as bit rate in no way takes into account the efficencies of the compression schemes.

DVCPRO HD is an interesting example - slated at 100Mb/s, but I believe it seldom is (at least not in the Varicam scenario). It is 100Mb/s at 60fps, but at 30fps it's only 50Mb/s and at 25 it's only about 42Mb/s. Also, it has a thin raster. XDCAM EX is another one, it's only 35Mb/s but achieves that with a high-efficency Long-GOP codec, meaning it achieves results that can't be compared to another methodolgy at the same bitrate.

HDCAM is another one - 1440x1080 HD at 144Mb/s, but with 3:1:1 sampling. As the transfer from the tape to Avid is not native data, but instead 'baseband' 4:2:2 HD-SDI then capturing at 145Mb/s, which is a marginally higher bitrate but is potentially more lossy, as you are using roughly the same bitrate to compress a lot more information.

In fact anything that's coming in over HD-SDI has been unpacked to a native 4:2:2 uncompressed stream. That is 440Mb/s if memory serves. That signal is then being recompressed with a different compression method.

Overall, unless you have some compelling reason not to, I'd be inclined to use the highest qualtiy (lowest compression) that is practically available to you. Diskspace is cheap, and you may as well be making the best pictures you can, right?

jwrl:

What cable networks require can hardly be regarded as the benchmark standard for image quality.  As far as the average cable operator is concerned, the more channels you can squeeze into a given space the better.

What cable networks expect from a production company and what they expect from themselves are two different things.

Also, I found this on the Avid website. If this is true I hope Avid will speak up for themselves here.

"Avid DNxHD: For high-quality mastering when using 8-bit lower data rate sources such as HDCAM and DVCPRO. 145Mbps is the data rate for 1920 x 1080 30fps interlace sources (60fields). Progressive sources at 24fps will be 115Mbps and at 25fps will be 120Mbps"

Fazz Powell:

... so I have to do a crash record without timecode matching the job when I print to tape.

There is no need for loosing timecode match. You still can connect your DNA-TC-Out to your record machine TC-In. Crash recording WILL have now timecode match. I do this all the time when I have to play out HD to SD.

pixel

 Yeah Pixel - Someone on here gave me that work around a while back, but for some weird reason I can't get time code out of the Adrenaline BOB and I can't figure out why. Drives me CRAZZZZY. I 

If your using LTC you must enable Generate LTC on Playback under general settings.

 THATS were they hid that setting. Holy cr_p Batman, I'll FedEx you a beer. Thank you SOOO MUCH!

Chimay red will be fine.

Red Ochre - Yeah seeing you're in SF, wine makes plenty of sense. In the rant catagory, it's hard to believe printing a HD program to tape is such a pain with MC. I have no idea how to convert the title keys to DNxHD 220 (and Avid's "assurance" support doesn't either). We use Inscriber and Boris Red for Titling. I wonder if in 3.0 you can do an HD Digital Cut without jumping through a zillion hoops.

Chimay is an over priced belgian beer which I can't normally splurge on, but since your buying!

We'll be getting 3 new edit suites (up from Xpress meridien!) in about a month. Guess I'll be able to tell you then.

For us the transcode problem usually stems from switching our formats back and forth. Something with alpha inevitably gets imported while in 30i and we don't notice 'till an hour before fedex. Very stress inducing.