Why does my HDV to DVD look like crap???

Randy Brown wrote on 11/21/2010, 8:21 AM

Thanks Opampman but that is something I tried with no results.
Thanks again,
Randy

fldave wrote on 11/21/2010, 8:40 AM

Two things that I see immediately:

MPG Average bit rate of 6MP. Too low for the better quality, even 2 pass doesn't look as good as 8MB constant, in my opinion.

Deinterlace Method: Interpolate throws away every other field. Try Blend, or the Lagorith deinterlace method in Virtual Dub.

That said, how are you comparing the two? I tried a 704x480 widescreen property setting, and it didn't look too bad, for 6MB mpg. If you are trying to compare the two on a HDV timeline, of course the 6MB mpg will look like crap upconverted to the display. You are compressing, then upconverting. Compare at the SD 704 setting.

PerroneFord wrote on 11/21/2010, 9:17 AM

I don't shoot HDV with the EX1. But on material I've shot on the EX1, edited, and then put through Squeeze... it's very very nice.

Laurence wrote on 11/21/2010, 9:42 AM

One more thing: pretty much all the CRTs that are still around are 4:3. There are exceptions. For instance, I have a 4:3 CRT HD TV that has a true 16:9 interlaced "squeeze" mode, and occasionally you'll see a 16:9 CRT HD TV (not too many though because they were really expensive: I'm talking several thousand dollars at the time).

Anyway, what this means is that when you do a 30p 16:9 render and play it back on a 4:3 TV, it is going to have every fourth line dropped and go through a hardware deinterlace. This is going to happen whether you like it or not.

Anyway, the point is, that practically speaking, this should get rid of any line twitter. Even if it is there you shouldn't see it. This would be in line with my experience, where I remember seeing line twitter on my 16:9 CRT, but as a general rule, it never bothers me.

It seems to me like any steps to avoid line twitter may have been relevant last year even, but shouldn't be a problem anymore.

Laurence wrote on 11/21/2010, 11:13 AM

>Why should it make a difference whether the Shapen FX is on the event level rather than the track level? I'm not questioning this wisdom - just trying to understand it.

If you put it on the track, you'll be sharpening your transitions as well. Sharpening through crossfade transitions is not something you want to do.

This leads me to a related question:

When you put it on the events, do you put it pre or post pan/crop (In Vegas 10 speak)?

In other words, do you want the sharpening to happen at the HD resolution, or after it is downrezzed?

Also, can you leave it on during HD renders? I know that you would want to bypass the zero sharpen during a smart-render, but does it hurt anything for a regular HD render otherwise?

amendegw wrote on 11/21/2010, 1:06 PM

"If you put it on the track, you'll be sharpening your transitions as well. Sharpening through crossfade transitions is not something you want to do."Okay, I understand that, but if Sharpen is set to zero, will it have any noticable effect on the transitions?

...Jerry

Edit: I did a test. Took a couple of 1920x1080 60i AVCHD clips, added a transition and resized it to 720x480 Widescreen MainConcept template. Included is a transition. Left side has Sharpen=zero at the Event level. Right side has Sharpen=zero at the track level. I can't see a nickel's worth of difference. You can right-click and "save target as" here.

PeterDuke wrote on 11/21/2010, 4:02 PM

"In other words, do you want the sharpening to happen at the HD resolution, or after it is downrezzed?"

You should actually blur the HD image slightly before resizing down to reduce the effects of aliasing high frequency components. Then you can apply some sharpening.

John_Cline wrote on 11/21/2010, 4:12 PM

"You should actually blur the HD image slightly before resizing down to reduce the effects of aliasing high frequency components."

I've never found it necessary to do this.

Laurence wrote on 11/21/2010, 4:19 PM

There are always more artifacts and blockiness at the transitions because of how little each frame has in common with the one before at these points. Adding sharpening on the track instead of on individual events is going to exaggerate these artifacts. Sharpening on the events will not.

Edit:[Having said that, I don't see any problem on the crossfade transition in amendegw's clip either. It makes me wonder if maybe you could get by just adding a sharpening filter with a value of zero on the master video bus. Hmmn.

PeterDuke wrote on 11/21/2010, 4:21 PM

You only do it as necessary to reduce twinkling or jaggies.

Laurence wrote on 11/21/2010, 4:38 PM

Is there a script that will put a sharpness filter on every event with a value of zero? It seems like a pain on a big project to do this manually and have to go through and turn it on and off clip by clip for your HD and SD renders.

amendegw wrote on 11/21/2010, 4:49 PM

'You should actuallyI don't have an answer to the before or after pan/crop question, but I used to be a big proponent of using Gaussean blur to reduce the Moire patterns produced when downrezing HD to SD. However, I've been doing some experimenting on this and I now feel that even the minimum blur results in too soft a render.

IMHO, a better solution is to change the clip's properties to "Reduce Interlace Flicker". A test is here. Note the softness of the video versus the flicker on the the herringbone jacket.

...Jerry

Edit: I guess I should have said, the linked video is rendered using a 720x480 MainConcept Widescreen DVD Template. Source is 1920x1080 60i AVCHD Canon .mts

PeterDuke wrote on 11/21/2010, 4:53 PM

Can you apply sharpness to a dummy event and then copy it and paste the attributes to multiply selected events?

PeterDuke wrote on 11/21/2010, 4:56 PM

It shouldn't be just any sort of blur but rather a low pass filter. Did you follow up with sharpening your SD result?

Laurence wrote on 11/21/2010, 4:59 PM

Here is another idea:

Lots of cameras that do 30p will let you do it as 60i. In other words, the video is 30p but it is flagged as 60i.

What this will do in Vegas is that when you resize it (as in an SD downrez render) it will separate the even and odd fields, resize, then fold the fields back together. This might actually be a good thing in eliminating the twitter we are discussing. Maybe the solution is as easy as just shooting 30p flagged as 60i.

I've just decided to start shooting my 30p this way since that is the only way to make it smart-render in Vegas. I had been worried about losing resolution in SD DVD downrezzes, but maybe that's a good thing. Anyone have an opinion on this?

amendegw wrote on 11/21/2010, 5:19 PM

"Lots of cameras that do 30p will let you do it as 60i. In other words, the video is 30p but it is flagged as 60i.I don't know how good a job PF30 will downrez to SD - I'll have to do some testing on this, but I do know that PF30 (pseudo 30p) does a much better job of stabilization via Mercalli V2 - see: http://www.sonycreativesoftware.com/forums/ShowMessage.asp?Forum=4&MessageID=734464 (and scroll down to be bottom).

...Jerry

Laurence wrote on 11/21/2010, 6:39 PM

Frame mode pseudo 30 fps is different than what I'm talking about. Frame mode 30p is basically 60i with one field thrown away and the other field doubled. The Z1 did this. So did the HVR-A1. What the HVR-Z7 does is true full resolution 30p with both odd and even lines shot once every thirtieth of a second. That true 30p can be flagged as either 30p or 60i depending on your settings. Either way it looks equally sharp (and twice as sharp as frame mode 30p).

farss wrote on 11/21/2010, 10:57 PM

The solution is to add GB in the vertical direction only.
It is possible you may need a little in the horizontal direction but not much. V values of 0.001 to 0.005 should work. H value of 0.001 should be enough if you're getting bad aliasing as well as line twitter.

There's no hard and fast rules. You may never see this problem either. Shooting outdoors with say the EX1 at f8 or above the diffraction from the iris is probably softening the image enough to avoid the problem. At f4 indoors....

It is absolutely vital this is done before scaling. If you want to add sharpening because your SD is a bit soft for your taste then this must be done after scaling. Exactly how you achieve this is up to you. The few times I've done it I useda nested HD project inside a SD project. There might be simpler ways.

Bob.

Bob.

Christian de Godzinsky wrote on 11/22/2010, 2:54 AM

A informative and long thread, but also a little confusing...

Could someone pro be so kind and summarize:

1. Is it possible to produce professional (or excellent) quality down-conversion from AVCHD (pal 50i) material to SD (mpeg 2 pal50i) - just using Sony Vegas Pro 10 as the only tool?

2. If yes, what is the work-flow (- are there any special "tricks" required, in addition to the usual ones)?

3. If no, what tool(s) do I have to get in addition to VP10 to get excellent results?

I am a little confused, that's why I ask... I have never been satisfied with the AVCHD (PAL50i) to MPG2 (PAL50i) down-conversion results and would need to find an easy way of achieving better results...

Christian

amendegw wrote on 11/22/2010, 2:56 AM

"The solution is to add GB in the vertical direction only."Not to belabor the point, but middle clip in the render I posted here has the very minimum of blur (.001 vertical only), and there is a noticeable softness when compared to the original. As you can see, the "reduce Interlace flicker" test produces the best balance of flicker reduction and sharpness. Pay close attention to the herringbone jacket.

You should right-click on the link and "save target as" 'cuz this is a .mpg file that is not amenable viewing in a web mediaplayer or progressive download.

I guess, it would be nice to be able to set blur to a lower value than .001, but I don't think that is possible. Maybe a feature request to Sony?

...Jerry

craftech wrote on 11/22/2010, 4:44 AM

I don't shoot HDV with the EX1. But on material I've shot on the EX1, edited, and then put through Squeeze... it's very very nice.
==================================================
I don't either so it sounds like something for me to try. I am always looking for better HD to SD. As many have said, it is the number one challenge for HD shooters especially for CRT television.

Thanks,
John

farss wrote on 11/22/2010, 4:47 AM

That's a very interesting outcome and pretty much the exact opposite of my experience. In my case Reduce Interlace Flicker is just way too much blur, worse then your 0.001 VGB.
I have a horrible feeling something is messed up between PAL and NTSC with Vegas in this areas because more than once over the years I've found us PAL users are getting very different outcomes compared to our friends in NTSC land.

I'll try to dig into this further. One thing I would prefer though is to use tools like res charts. Sometimes its hard to tell where an issue is coming from.

Christian above is asking about AVCHD > SD PAL and to me this is a very difficult and different issue. I've noticed the compression artifacts that AVCHD produces are hard on the mpeg-2 codec. Now that's another issue to cloud these already muddied waters. I'm not ignoring his question just wishing I had more time to devote to researching these issue because they are not trivial but to get the right answers takes a lot of care and time.

I'll also say this is something SCS should be doing more about to assist all of us. Doesn't matter how good the product is, if the users cannot figure out how to use it correctly, it's lost.

Bob.

Kimberly wrote on 11/22/2010, 6:15 AM

Hello All:

To tag onto Christian's post, can someone summarize a "best practices" for shooting HDV in the hi-def mode, editing in Pro 10.0, and rendering for distribution on a (standard def) DVD made in DVD-A Pro 5.0?

I work on a dive boat and produce a weekly trip video. My customers may view their DVD on a flat screen TV, tube TV, or computer . . . but my guess is that viewers on Ye Olde Tube TV are the minority. Most customers are NTSC. Maybe 10% are PAL.

I'm looking for a workflow that gives the best one-size-fits-most approach. The answer is probably somewhere in this thread, but I'm a newbie in this regard so it's confusing to me.

Regards,

Kimberly

Laurence wrote on 11/22/2010, 6:35 AM

Not to head off into space, but isn't the issue with interlaced SD 16:9 being played back on a 4:3 TV more of a field order issue than anything else?

Imagine the following:

1/ 24p movie shot. Encoded onto DVD with 3:2 pulldown. DVD player is setup for a 4:3 CRT so it drops every 4th line: letterboxed widescreen that looks fine.

2/ 60i (or 50i) HD or HDV movie shot. Downrezzed with Vegas into 16:9 60i (or (50i). DVD player was designed for 24p movies. DVD player designers didn't care about 16:9 interlaced formats because they are not used by major studios. With that in mind, DVD player drops every 4th line just like it would with 24p. In the new letterboxed image, line one is from field one, line two is from field two, line three is from field one, line four is also from field one because original fourth line was dropped... in other words the field order is screwed up. To avoid this, the DVD player adds a quick blend fields deinterlace to the output when it sees an interlace comb. Now what you have is a slightly blurry image which still twitters any time too much detail sneaks past the deinterlace.

3/ Movie is shot in 30p hd. No big deal when every fourth line is dropped to make the letterboxed 4:3 CRT image.