Sony Vegas: Print to Tape boosting contrast?

[orcus]

I used Sony Vegas's Print to Tape function to output my video to a miniDV tape in a camcorder over firewire. I then recaptured the video from the camera (again over Firewire) using Adobe OnLocation to compare it to the original to make sure everything went okay.

The new DV print has noticeably higher contrast than the source. Light areas are blown out, dark areas have become darker and lost detail.

What's going on?

[jagabo]

Adobe is probably using the ITU-R BT.601 (AKA rec.601) matrix to convert to RGB for display whereas Vegas uses the PC.601 matrix. YUV video is supposed to have luma components that vary from 16 to 235. 16 is black, 235 is white. But on computer monitors 0 is black and 255 is white. Vegas defaults (?) to leaving the luma in the 16-235 range, but most other programs default to expanding the luma range from 16-235 to 0-255.

In all likelihood both files are the same, Adobe is just displaying it differently. Open the file Adobe captured in Vegas and compare.

[JohnnyMalaria]

Just to add to jagabo's wise words, use Vegas' scopes to compare them - it will help reduce the subjectivity.

[orcus]

Thanks for the replies. I'm still confused, though.

My source is a lagarith-encoded AVI (made from an edited Vegas project). I made a new Vegas project and rendered the AVI as DV.

If I import the DV into Vegas alongside the source AVI and flip between them, they look the same (both look correct). Scopes show the same as well.

But in any player (WMP, Media Player Classic, Quicktime, or AvsP's preview window), the DV seems to have clipped levels. Lots of detail is lost in the highs and lows. Doing screengrab comparisons from VLC, MPC, or AvsP shows this inarguably.

I even tried rendering the source AVI to lagarith AVI again, and even this had the clipping problem, so it doesn't seem to be an issue just with the DV codec (I had thought that might be the case, since I saw on another site "Some codecs such as Vegas' DV codec will want to see studio R'G'B' levels").

I need to get this DV to a mastering house so I need to make sure they will see what they are supposed to see, and if only Vegas displays it correctly, I don't think the final result will be what I intended.

Any more ideas?

[edDV]

Can you post some representative stills of the files that clip? And the original?

You either have a player problem or a RGB to DV transcoding error.

It would also help if you summarize the process path. Are these DV original camcorder files or something else. What are the steps to the final DV file.

[orcus]

One more wrinkle:

I applied the Color Corrector plugin to the source AVI track in Vegas, using the "computer to studio RGB" preset, then rendered to DV. This time, the DV looks identical to the source AVI when flipping between the two in AvsP, BUT now it doesn't match the source in Vegas. And, somehow, the DV is still clipped compared to the AVI in VLC, WMP, MPC, etc.

I guess my question is, which playback display do I trust and which method do I use to ensure that the tape I get back from the mastering place is identical to the source?

[edDV]

I recommend you always test your processing path with the Belle Nuit color bar. It quickly reveals where problems originate. Load the Tiff and run it through the process. Use the DV version of the chart.


http://www.belle-nuit.com/testchart.html

[orcus]

Hi edDV --

The video was shot mostly with miniDV cam and composited in After Effects. Scenes rendered from After Effects to lagarith-encoded AVIs.

Scenes were edited together in Vegas and the final video rendered out as a lagarith-encoded AVI (this AVI is what I am calling the "source" here). I am now trying to print that source to tape, but when rendering to DV the levels appear to be clipping.

Here is an AvsP framegrab from the source:



Here is a grab from the DV render of the source avi:



And here is a grab from a DV render made with the Color Correct plug applied to source avi:

[edDV]

I'm remote on a laptop today so won't be able to test with Vegas until this evening. Noon here now.

[jagabo]

If I import the DV into Vegas alongside the source AVI and flip between them, they look the same (both look correct). Scopes show the same as well.

But in any player (WMP, Media Player Classic, Quicktime, or AvsP's preview window), the DV seems to have clipped levels.

As noted earlier, Vegas is unusual in that it doesn't expand the luminance when it converts YUV to RGB for display whereas most other programs do.

[edDV]

Clipped levels only happen to DV video over level 235* when transcoding maps 16 to zero and 235 to 255 for RGB. There are techniques to compensate for this once the process flow is known.

This Process flow works (DV format Vegas project).

1. Cap DV to Vegas or via WinDV
2. Export from Vegas to AVI uncompressed RGB.
Vegas maps zero to zero and 255 to 255 thus passing overshoots.
3. Process externally as RGB
4. Import to Vegas DV project as RGB
5. Verify levels with scopes.
You may need to adjust black or white with levels filter. Maybe gamma. VirtualdubMod needed no adjustment.

* Consumer camcorders essentially record 16-255 to fake low light preformance. They are very bad for highlights and usually clip white at 255. Pro camcorders record 16-235 with only highlight detail and spikes above 235.

[orcus]

The scopes show the two files are the same, both spanning 0 to 255. But if they are actually the same, why do they display differently in every other player? Shouldn't whatever is happening (or not happening) in those players happen to both files so that they display the same, since apparently both files have the same range of values?

Should I just assume they'll transfer correctly when the mastering place converts it to BetaSP?

[edDV]

The scopes show the two files are the same, both spanning 0 to 255. But if they are actually the same, why do they display differently in every other player? Shouldn't whatever is happening (or not happening) in those players happen to both files so that they display the same, since apparently both files have the same range of values?

Should I just assume they'll transfer correctly when the mastering place converts it to BetaSP?

These should have most "weight" in the 16-235 levels (0-100%) with only peaks above. Broadcast players assume white at 235 and treat above 235 as bright detail. If you are pushing white to 255, these will be non-standard and not play well side by side with a commercial DVD.

Any levels below 16 (0% on the scope*) will be crushed (not visible) on a properly adjusted HDTV.


* For DV scope should be set to
7.5 IRE setup = off
Studio RGB (16 to 235) = on

[orcus]

Man, what a nightmarish business the world of video is

At this point, since I neither understand any of this nor how to correct it (if in fact anything needs correcting), I'll just give them the tape and hope for the best.

Thanks to all for your efforts.

[edDV]

Man, what a nightmarish business the world of video is

At this point, since I neither understand any of this nor how to correct it (if in fact anything needs correcting), I'll just give them the tape and hope for the best.

Thanks to all for your efforts.

If they are good, they will know how to fix it but will probably charge extra and you may lose a generation.

[2Bdecided]

Your final problem - looking the same in Vegas, but different in everything else - is probably down to how Vegas calls the decoders. Some decoders output RGB only, some can do YUV or RGB depending on what they are asked for.

If Vegas asks for YUV, while a media player asks for RGB - and if the decoder maps YUV>RGB 16-235>0-255, while Vegas doesn't expand the range, then the Vegas version will look more "dull" - the media player is showing the correct (but clipped) playback.

If the other file is a codec which only does RGB (e.g. a lossless file containing RGB!) then you get what's in the file - apparently the "dull" unclipping version, playing the same in everything (on a PC).

If you don't like what your media players are showing you, then you have a problem - because what they are showing you is correct!

As edDV says, experienced people can correct this for you, bringing the levels back down again without expanding (clipping) them on the way. Inexperienced people will just clip at and say your source is clipped!

Cheers,
David.

[jagabo]

Most media players will ask for YUV (typically YUY2 or YV12) and display it using hardware video overlay. Then the graphics card will perform the YUV to RGB conversion to send the video to the monitor. You can adjust the video overlay proc amp settings with the graphics card's setup applet.

If a decoder can't supply YUV the media player will ask for RGB instead. In that case the decoder will perform the YUV to RGB conversion. Some decoders let you specify how the conversion is done. The Cedocida DV codec for example.

[2Bdecided]

Most media players will ask for YUV (typically YUY2 or YV12) and display it using hardware video overlay.

I didn't know that. DirectX seems to give me RGB more often than not, whatever that means (I'm still learning!).

Still, the reasoning stands: if file A is YUV 16-235 range (i.e. normal) and file B is RGB 16-235 range (i.e. non standard), then most software will expand file A but not file B. Vegas won't expand either. Hence, the files look the same in Vegas, but file A looks brighter (actually more contrasty) than B in everything else.

Cheers,
David.

[jagabo]

Agreed.

[orcus]

If my source video shows an RGB of 0-255 for the source in Vegas's scopes, should I use the Vegas Color Correction plug with Computer to Studio RGB preset before rendering to DV? That would seem to be the case based on what 2BDecided said, that the media players are showing the "correct" playback.

[jagabo]

If my source video shows an RGB of 0-255 for the source in Vegas's scopes, should I use the Vegas Color Correction plug with Computer to Studio RGB preset before rendering to DV?

Yes.

[edDV]

Difficult to tell form your stills if the color space was stretched in the frame capture program or in your processing filters.

For a DV project the standard Vegas SMPTE color bar (see media generator) places black at zero % (digital level 16) and white at 100 % (digital level 235). Only the pluge extends below black.



The Belle Nuit color bar extends markers into the 0-15 and 236-255 levels. It is used to test the full overshoot ranges and quickly indicates where clipping is happening.



Your clips extend below black but have little extension into the white overshoot range. Below black levels will be "crushed" or not visible on a properly adjusted monitor.

source.png


dv.png


dv-cc1.png

[edDV]

As said above, if one passes the belle nuit color bar through the process, one can see where luminance expansion is happening. It can also happen using the Vegas levels and color correction filters with extreme adjustment so one must watch the scopes.

The levels or color correction filters can easily fix black for the stills shown.

Here I just adjusted black up to level 16 and set white around 235 with a few overshoots above 235.



Here I pushed white up to 255 to show clipping at ~108%.



These should be monitored from the IEEE-1394 port on a properly adjusted TV to judge colors.

[jagabo]

If it will help anyone out, here's a 1 second long DV AVI of the Belle Nuit chart with luma below 16 and above 235.

belle%20nuit%20ntsc%20dv.zip