I've always used the Avid test patterns for monitor calibration and file checking, but today I discovered that they are unreliable :(

I imported the HD_1080 SMPTE bars, and checked the values using the match colour eye dropper in CC mode. In the PLUGE area "superblack" is R5 G5 B5, and black is R15 G15 B15. This is in a 1080/25P project with YCbCr709 colour space. I ran the same checks in another edit suite with same MC version V6.5.2 and got different values, including  a G value 2 lower than R and B for all teh PLUGE area!

When I try the same tests, using the same eyedropper tool with the www.belle-nuit.com test chart, every value in the chart checks out correctly in both edit suites.

I feel daft for having trusted the Avid supplied test patterns without checking. Any idea why they check out so badly?

Thanks for any info.

Nigel

How did you import the bars?  They should be imported with 601/709 selected.  Sounds as if you brought them in with RGB selected.

Also, put them on the timeline, open the Color Corrector and check out levels with the waveform scope.  How do they look there?

Jef

cheers Jef.

They were imported correctly as 709, so that Avid would not remap the values.

The waveform scope shows the same values as the dropper - superblack sits above the 0% line.

And the belle-nuit test chart checks out EXACTLY as it should!

I just checked out the values of the PLUGE in Gimp, and got similar results - Superblack is 6,5,6, Black is 16,15,16.

You can check it yourself using Avid or Photoshop or any other programme.

The file is at Applications/Avid Media Composer/SupportingFiles/Test_Patterns/HD_1080/SMPTE_Bars.tif

Interesting.

Out of curiousity I opened the SMPTE Bars.tif in Photoshop and you are correct, the levels are not perfect.  But hey, these are just test chart that we are supposed deliver masters to the networks with.  This should be good enough.

Jef

This is pretty shabby of Avid IMO - including faulty test patterns with their software!

I will use the Belle Nuit one from now on.

While they may not be perfect, I would still use them for a deliverable over the Belle Nuit chart.  All the numbers on it, while incredibly useful for diagnostics, would risk a rejection.

Remember those charts were developed in the analog days to be used with analog scopes.  The tolerances were not down to the level you have noted.

I do agree that it could be better.  But, then what can't be better?

Jef

I'm curious which chart files you both tested. I checked this a couple of months ago, using v6.0.x and the charts were exactly on. Checked using Photoshop and Avid. Every component of the chart I used was where it was supposed to be.

Mac HD>Appications>Avid Media Composer > Supporting Files > Test_Patterns > HD_1080 >  SMPTE_Bars.tif

Nigel, I could be wrong, but I thought the superblack in the PLUGE just had to be reasonably below reference black.

Are you saying it should be 1, 1, 1?

"Super black" should be 0,0,0, is my understanding, and "super white" should be 255,255, 255.

What really bugged me was that there was a colour imablance - G was 2 points down on R and B in every part of the pattern including colour bars, when I checked them in CC mode.

Maybe Jef is right and I'm being too harsh, but I would expect Avid to fix stuff like this so it IS accurate.

If I were a betting man, I would say these test patterns were captured from some test generator, not created from scratch.  And a whole, long line of small inaccuracies and deviations are present.

But as the fine folks at Belle Nuit have shown, it is possible to make a high quality chart.

Jef

My understanding is that super black and white are recordable ranges outside of reference black and white that can be used for signal under or over shoot. Which would make their most extreme values 1 and 254, as 0 and 255 are strictly illegal for broadcast or in Rec 709.

But I do agree that the chart should be accurate and that it should not have a discrepency between R, G and B values.

Hi all, this is RandyF from Avid Engineering.  Thanks for bringing this to our attention.  We'll take a look and keep you posted.  

My initial read is that this is due to the fact that the test patterns were initially created in YCbCr for the Video Output Tool.  This possibly is due to the conversion to RGB.  

Are you noticing noticing any problems with the test patterns in the Video Output Tool when connected to a scope?

-RandyF

I stand corrected

According to Wikipedia you are right Bobby - "Interface codes 0 and 255 are used for synchronization, and are prohibited from video data". [http://en.wikipedia.org/wiki/Rec._709]

Further research suggests that values for the SMPTE pluge should be:

3.5 IRE (Super Black) = 7,7,7 [2% RGB]
7.5 IRE (Video Black) = 16,16,16 [6% RGB]
11.5 IRE = 24,24,24 [10% RGB]

[http://accad.osu.edu/~aprice/courses/752/blacklevel.html]

The levels I get from the Avid SMPTE test pattern are

Superblack 6,5,6,

Video Black 16,15,16,

Grey 25,24,25.

As Jef says the errors may be a result of digitising from a signal generator.

Let's hope this thread helps resolve any confusion?

Cheers

Nigel

Edit: Thanks for chiming in RandyF. Not been able to test with external scopes yet, but I think you need someone with Avid hardware to check that? Any problems with me "could" be down to my MX02 mini-max, anyway.

YCbCr to RGB conversion aliasing/errors sound feasible.......

Hi RandyF, thanks for coming in on this.  I was about to post something really wordy but I'll just summarise instead and hopefully you can comment on my findings.

Trying to be as succinct as possible, my findings from last year with 5.5 are:

The Avid charts are poor quality and wrong, especially when pixel peeped at in Photoshop.  Should go without saying that once taken into Avid, with the correct 601/709 setting, the values should still be wrong - wouldn't expect or want Avid to mysteriously fix them - and they are indeed wrong.  Interestingly the Photoshop readings are different to Avid readings.

With perfect 75% EBU, 100% EBU and SMTPE but not Belle Nuit charts (not tested at all), self created in Photoshop, once taken into Avid the values still aren't correct.  Considering that RGB graphics need to be converted to YCrCb video there is bound to be some rounding of values going on at some point so if we are still looking at RGB values the colour picker will need to convert back so it should be no surprise that the readings in Photoshop and Avid differ slightly.  If you factor in that Avid could be using a higher bit depth internally, 10bit, 12bit, or even 32bit floating point, then going back to 8bit for the colour picker reading then something is highly likely to change if the exact formula used isn't flipped to go back the other way.  My charts with absolute (16, 16, 16) and (235, 235, 235) values would change, within in Avid, to read from something like 15 - 18 and 233 - 236.

Assuming Avid imports, with the 601/709 as the import setting, are transparent and as good as possible (depending on which codec is used) this should be acceptable.  However, if you roundtrip a chart from Photoshop into Avid via import (into most available codecs - not tested exhaustively) and export it back out as a still image and pixel peep at it in Photoshop the process isn't "transparent".  I know pixel peeping at "video" isn't fair but what were once perfect solid blocks of the same colour are now different.  They are vertically striped.  This suggest to me that as well as having undergone a slight change in value via the RGB to YCrCb process there is also something like deliberate noise being added.  This to me is/was unknown and somewhat of a surprise.  Horizontal striping I can understand, as it could be an interlacing issue, but vertical stripes are a weird one - especially because the stripes are perfect and not broken, therefore not random like noise would normally be.

That all said, Avid's charts and my own behave as expected in the scopes and looking particularly at the vectorscopes the points land on pretty much the right areas.  They just don't behave when trying to look at them as 0-235 values.