Ok this thread is half a technical inquiry and half a project thread similar to my Upscale Project and 35 Project.
There is a long standing theory here on the forum that I always wanted to test. It was noticed some time ago that many video masters made from roughly late 90's to mid 2000s have a magenta (or red or pink) look to them when viewed on modern TVs. You can see it often once you know what to look for. White clouds that have a pink hue, faces that are red with high blood pressure, blues that looks more purple then blue, etc.
The thought as to why this happened is that the mastering monitors used to make these videos were CRTs. Those CRT monitors work a lot different then our modern screens and they tend to push very green and yellow. This likely happens because unlike modern panels which have color standards to achieve (Rec709, Rec2020), a lot of CRTs only had color temperature to work with (D65 or D93). In order to compensate the masters were pushed toward red in order to keep the white balance correct. That worked fine when we were in the world of CRTs but once the world migrated to flat panels these old masters took on a very magenta or reddish look.
The industry moved on to LCD in mid 2000s and then OLED panels. With that move the red bias went away. But the work was already done. In the run up and adoption of HD in the 2000s many studios quickly mastered their old movies and those masters still are being trotted out. You see it a lot with 20th Century Fox, Disney but I see it the most by MGM, who remastered in HD most of their film library in the early 2000s.
So this thread's goal is to try to prove if this theory is correct or not and if it is, to adjust some masters to reflect how they look on a CRT so you have the proper look on modern flat panels. So, step one is to get a CRT. Not just any CRT but one used in the mastering process. A Sony BVM, PVM or some such. That is a difficult proposition because those monitors are expensive since retrogamers like to play on them and I simply don't have the money to afford one. Luckily, I have a friend that had the perfect answer. An HDCRT mastering monitor in the Sony HDVS line that is useless to retrogamers as it will not do 240p or 480i. But it will do the HD resolutions that I need at D65. I borrowed the monitor, hooked it up to my laptop with its calibrated OLED screen and played a movie on both simultaneously. Now the iPhone I used to take the pics is skewing the colors a little (especially towards the red unfortunately) but these rough, rough pictures should let you know what I'm seeing till I break my DSLR out and replace the pics later with better examples.
Top is a calibrated OLED screen/Bottom is HDCRT
Star Trek 2 (1st gen BD), Office Space, Jurassic Park, Temple Of Doom, The Dark Knight (1st gen BD)
And what I'm seeing is it looks like there might be some truth to this theory. It does appear that the white balance is better for some of these masters, the skin tones are better and the pervasive magenta/red tint is lessened or gone. Overall the colors are much better in my opinion. That is not to say not more theatrical accurate. There is no way for me to know that but hopefully more accurate to what the person grading the film was going for.
So the next step is to do more experimentation and start trying to color grade these masters to what I am seeing
That's pretty cool. I have heard good things about CRTs in other areas too, in terms of black level for example, if they are properly set up that is. They also seem to be good for watching movies because something about their refresh rate is closer to how a cinema projector works supposedly.
Have you calibrated the CRT as well? Otherwise it might not be a fair comparison. I would be surprised if a calibrated CRT had significantly different colors to any other display, but I'm happy to learn something new.
I can definitely imagine a CRT having a different look without any calibration though.
Edit: If you could take some pictures in .raw and send along with the original video frame, I could try running it through a tool of mine to see if I can come up with a correction matrix to simulate the CRT look. In fact, the best might be if you find a good reference file of a Color Checker or IT-8 target and let it run over the CRT, then it might be even easier, since I could calculate a matrix to "correct" the CRT, but then just reverse the matrix (online calculators everywhere) and that might give the opposite - a CRT look simulator matrix.
Edit 3: Not sure where I found it but I attached one that already comes in sRGB and has been measured from 30 different charts and averaged or something, so it should be pretty close.
2020-06-12, 04:33 PM (This post was last modified: 2020-06-12, 04:34 PM by BusterD.)
I'm lucky enough to have a 20" PVM-20L5 (got it for about $400 in 2016 which was a bit more than the going rate at the time, but I'd probably have to pay more than $1000 for one now, if I could even find one).
It's not professionally calibrated or anything but I can do some tests if you'd like, never tried playing any BDs on it but it does do up to 1080i and I have an HDFury2 I can use for HDMI sources.
A couple years ago I visually compared my The Lost World: Jurassic Park widescreen VHS to the HD master... The VHS was definitely more on the yellowish side than the HD. HD colors were more vibrant while I would say the VHS colors seemed a little more neutral. Just working off of my memory, but I do remember noticing a big difference. Who knows which one is correct...
(2020-06-11, 10:39 PM)TomArrow Wrote: That's pretty cool. I have heard good things about CRTs in other areas too, in terms of black level for example, if they are properly set up that is. They also seem to be good for watching movies because something about their refresh rate is closer to how a cinema projector works supposedly.
Have you calibrated the CRT as well? Otherwise it might not be a fair comparison. I would be surprised if a calibrated CRT had significantly different colors to any other display, but I'm happy to learn something new.
A surprise to me was the black levels aren't all that amazing compared to OLED. At least on this machine which was manufactured in 2000. That 2000 part is key since its the time a lot of the 1st gen HD masters were made. I did noticed my brain missed the smooth look of phosphors even running in 1080i/60Hz. Digital does have a sharper but harsher look.
I should point out the calibration isn't really something simple with this. Like it would be with a flat panel using an ISF calibrator. The control points are sort of the standard CRT controls. Bright, Contrast, Chroma (aka saturation) and Aperture (aka sharpness). All of those don't change the colors of the monitor just mostly change the backing luma. Color wise in the menu the only settings are the "white point" of D65 and D93. D93 (very blue) being for the Japanese and 65 for the rest of the world. Running a color bar through the machine (I have one and there is an internal one plus I’ll be trying yours) and checking the white seems to be on point for 65 (slightly yellow, warm daylight white) but I need to get some pics to make sure.
In the service menu you can change the intensity of red, green and blue but seems to be to offset the aging of the phosphors and checking the menu it seems that the hours are low and the tech who worked on it already adjusted it
(2020-06-11, 10:39 PM)TomArrow Wrote: I can definitely imagine a CRT having a different look without any calibration though.
I think that is the point I'm trying to go for or what I'm looking to find out. That the color and therefore color grading of a CRT is different out of the box. A tech was not expected to calibrate a pro monitor. You expected it ready to go from Sony. Plug in and start grading. I don’t want it to be a fair and level playing field. I’m looking to see if the differences make a..difference. I want to see if those differences make these masters look "normal".
But who knows I could be full of crap.
Moving it closer to say Rec709 if possible might be a nice secondary experiment but me figuring out how to get better pics is the key to that experiment. In the meantime I might try a few small test grade to show what I am seeing until I can get better pics. Maybe develop a “CRT LUT”.
An example is I watched parts of the 4 Alien films last night. Both Alien and Aliens looked off as I expected. The assumption is since both got remastered for BD that they would of been calibrated on a LCD PVM which was the default at the time. So I expected those to look different then what they would look like on a flat panel.
3 and 4 were not remastered and use the same older master used on the dvds. That means they are old enough to be mastered on a HD CRT. So I expect them to look redish, which they are and for that red to disappear on a CRT, which it does. 3 has the skin tones more orange and the overall tone more golden. 4 becomes more yellow and green but even more telling the red flashes at the end when the newborn is on the Betty become white.
(2020-06-12, 04:33 PM)BusterD Wrote: I'm lucky enough to have a 20" PVM-20L5 (got it for about $400 in 2016 which was a bit more than the going rate at the time, but I'd probably have to pay more than $1000 for one now, if I could even find one).
It's not professionally calibrated or anything but I can do some tests if you'd like, never tried playing any BDs on it but it does do up to 1080i and I have an HDFury2 I can use for HDMI sources.
That is a nice machine, you could definitely get a lot more for that these days.
I would love another pair of eyes to test with. After all two monitors are better then one. And it lessens my confirmation bias.
Ironically I have the fury2 also. Bought it a long time ago.
Well, that whole thing about the black levels caused me to spend a few days researching this once. Basically, as these things age and the elements in them degrade, they need to be recalibrated. The black level in particular is affected by this. So out of the box it will probably just look "meh". I think this setting is called "Bias" and can be set for each channel. The normal monitor menus typically don't allow you access. It gets a little tricky there. There's a software whose name I forgot that you can use together with a special USB device, which lets you hook up 4 little cables to a little board inside the monitor (it's typically hidden behind some small plastic cap, but it's possible that for some monitors you have to remove the entire back, idk). From that software and together with a measurement device, you can do a calibration. But more importantly, this software lets you set those bias levels.
Basically, the perfect black level is possible because no signal is simply no signal. The phosphor won't glow unless hit by the electron beam. But the "bias" basically does the same thing as in color grading I presume, subtracts or adds a signal. So if there is too much positive bias, even the zero value drifts into a positive place and thus your blacks become grey.
I once tried to fix a Dell unit with the method I described, but I think it must have been damaged or something because I could never get it to work with that software and USB thing. But if you google around a little, it's a common advice to do this on CRT enthusiast forums, so it's not completely exotic or anything.
With all that said, you can calibrate it software-side without doing any of what I just said, like you would a normal display, with something like the ... uhm ... Colorspdier? Spidermonkey? something like that lol. It's a little measurement thing you put in front of the display and it does some measurements and it calculates an ICC (?) profile, which it then sets in Windows, so everything that Windows outputs goes through that profile and compensates whatever color shifts or whatever the display may have.
Thing with white balance is, it doesn't tell the whole story. You can have the same image in two different color spaces, both perfectly white balanced, yet looking different. This is because each color space (sRGB is one) has its own primaries. Primaries are the coordinates of the R, G and B colors on an X-Y diagram which is based on human vision. I can't explain what it means exactly but I can tell you roughly - it basically says what intensity/saturation/purity the individual colors/subpixels have. Imagine a very saturated red subpixel, and then a subpixel that is just slightly red in comparison, but a bit pale. Send the same signal through both and you will get a different look. That's why you basically need a color profile for the monitor, which in the very least has a color matrix, which "translates" values from one set of primaries to another, roughly speaking.
For example if you go into the color settings in Photoshop and you go set a "Custom RGB...", it will ask you for the color primaries of R, G and B and the white point. The white point is kind of the equivalent of the white balance, but for the colors the R, G and B primaries are infinitely more important.
Long story short, for a perfect workflow this process has to be done for every component in the signal chain - for the camera/scanner and then for the display/printer.
Calibration targets like the one I put in my post are used for this. They are basically perfect reference values spread out somewhat evenly through the color space and if you display them or print them or whatever, and take a calibrated photo of that (let's just assume a DSLR is reasonably well calibrated for our demands), then you can take the resulting values of those colors, and then do some mathematical equations basically saying "We got A, but our perfect reference value is B. Let's find a formula to get from A to B", but of course 3-dimensionally for R, G and B. And that is the calibration matrix. Mathematically speaking it translates 3-dimensional values from one coordinate system into another. In this case the coordinate systems are color spaces, kinda.
Hence, a picture of that color chart on the CRT might allow us to create a matrix "correcting" the CRT to what it "should" look like, basically trying to arrive at the reference values. And that matrix then can be reversed and you should be able to just apply it to any image or video (via Channel Mixer in linear gamma) to get exactly the CRT look, at least color wise. That's what I meant earlier. You can of course fiddle around by hand, but I think for that matrix with 9 values, depending on the precision you're going for, you have trillions (or higher illions) of possible combinations, so you might be sitting there for a while.
Sorry for the wall of text lol, it's just something I've been a little obsessed about recently. And I still don't understand it nearly as much as I'd like to, heh.
Would be great to find an average setting that, more or less, will "fix" old masters, so we can get "right" colors on HD displays (mainly OLED and LCD, better FALD) similar to old CRTs. Curious to know what plasmas look like, if more towards CRTs or LCDs.
I guess it's a lot of trial and errors, but I think the thing to do is:
- calibrate the CRT
- calibrate the OLED/LCD
- feed the CRT with various old masters - possibly from laserdisc, better pre "digitally mastered", perfect would be to use a master struck from release print, should not be very hard
- correct the RGB (and eventually brightness and contrast) in the OLED/LCD to look like the CRT
- when found an average setting, feed the OLED/LCD with a new master
- this time, use "reversed" previous setting in CRT, so it should look like OLED/LCD
(I know, easy to tell, but the theory *should* be right!)
My cherished XBR960 HDCRT has the option to defeat NTSC red push in the menu and this was something Sony developed and then put on all of their late model high end prosumer units. It's under advanced video options and you have the option of monitor mode being set to standard NTSC with red push or Monitor mode which defeats it and results in a more professional image. I'll try to get some photos up to show the difference later on.
![[Image: K0XvfKB.png]](https://i.imgur.com/K0XvfKB.png)
