There are indeed. Quoting myself:
1 Like
It can go beyond personal preference which I made earlier: use the setting on your TV that allows for good linearity and gamma. If your TV can handle D65 well, use that, if it can handle D93, then use that. If both give good results, then I guess it does come down to preference. Ideally watch content in a dim environment where the TV can be much brighter than the surround to allow chromatic adaptation to work its magic. Then your eyes will compensate for the white point naturally.
2 Likes
Transformation via bradford is simple to do, but adequate LUT could give better results.
The only problem is that some LUT tools use “bradford” for their calculations. 
1 Like
Does any of this correspond to reality or is ChatGPT drunk again? Maybe @anikom15 knows?
9300 K “Sony look” is mostly a marketing / display mode label
- In the late 1980s–1990s, many Sony Trinitrons had a user-selectable color temperature setting labeled “9300 K” (or “cool” mode) in addition to “6500 K” (“normal”) and sometimes “5500 K” (“warm”).
- 9300 K mode was never truly 9300 K in absolute colorimetry ; it was a visually cool setting intended to appear brighter and sharper in stores or under bright ambient lighting.
- On most consumer Trinitrons, measured white point for 9300 K mode was more like 7000–7500 K , not a literal 9300 K.
Factory Defaults Were Warmer
- Sony TVs typically shipped with a default “Standard” or “Normal” mode that was actually warmer than D65 (6500 K) .
- This default mode is where the red channel boost (+3–6 %) comes in, producing a white point closer to 6000–6200 K .
- This is what gives the classic “Trinitron glow” — slightly warm whites, slightly enhanced reds and yellows.
Why People Associate Sony with 9300 K
- The “9300 K” label was on the remote / menu , so hobbyists often equated it with all Sony TVs.
- Marketing materials for CRTs sometimes touted “9300 K for crisp store-quality whites,” giving the impression that Sony TVs ran extremely cool by default.
- In reality, most consumers never used the 9300 K mode , because the picture looked too blue / cold. Default mode = warmer than D65.
What it Means for Red Channel
- Red push (~+5 % gain) existed in default / standard mode (~6200 K) .
- Switching to the 9300 K mode :
- Red bias is reduced (to make the picture look cooler)
- Blue channel is boosted
- Whites look “bluish white” rather than warm
- So the red boost and the “9300 K” label are actually opposites : red is stronger in default mode, weaker in the 9300 K mode.
So when people say “Sony TVs are 9300 K,” they’re usually referring to a menu setting — the real factory default mode was warmer (~6200 K) with a red boost , which is why NTSC composite and arcade RGB looked vivid.
Factory Defaults Were Warmer
Why People Associate Sony with 9300 K
What it Means for Red Channel
1 Like
You could try asking it for actual references, I have a hard time believing that user-selectable color temperature was a common thing already in the late 80s when users usually still didn’t have access to a service menu via remote.
Edit: There is already a feature called Trinitone found in user manuals of the early 90s EXR series
Also, there is a notch filter setting:
Seems these were really premium features at that time though. You have to go forward for color temperature settings on standard TVs, check out some manuals from the 2000s.
2 Likes
I’m not sure about that stuff, but it is correct that Sony made one of the earliest TVs where you could change the temperature. Maybe the earliest. Some time in the 90s they introduced a ‘game mode’ that made the temperature warm as well.
2 Likes
Right out the gate, i am dead certain that is incorrect as a generalization.
As i mentioned earlier in the thread, the 1990 “W0006311M.pdf” manual specifically says that the cooler/bluer setting is the factory default for Trinitone on those displays, even in the US.
As @anikom15 said, Sony introduced adjustable color temperature fairly early. (1984 is the earliest reference to Trinitone that i’ve found.) Given W0006311M.pdf and other manuals, odds are that setting factory defaulted to the coolest option until either 1992 around the time the early HD MUSE displays launched (supposedly with a D65 factory default), or c. 1996 when Sony started including those “game” modes that automatically switched to D65.
2 Likes
Sony TVs are still blue out of the box.
Sony sets factory defaults for sales floor use, but their products have always been prosumer- and enthusiast-oriented. People were meant to play with the knobs and settings. They were never really an ‘out of the box’ company. I wouldn’t give factory defaults too much weight.
2 Likes
True, and this also kind of gets into questions along the lines of “When did nerdier people outside of the film and television industry start becoming more aware that D65/warmer color temperatures were the more “correct” settings?”.
From what i’ve seen of old Usenet posts, that happened somewhere around 1992-1993. Which i suspect may have had something to do with articles about the MUSE/HDTV mentioning color temperature, bringing it into the broader conversation.
1 Like
It seems to coincide with LaserDisc and marketing driven by George Lucas and THX. Some of those LaserDisc masterings had more thought and care put into them than modern Blu-rays do.
1 Like
This is such a weird time for film restoration and home video. Cameron and Lucas have become completely infatuated with ruinous processing, and the major studios have practically stopped caring about quality control.
Meanwhile boutique labels are putting out utterly flawless 4K restorations of things like Guyver and Chronicles of Riddick for something like a 4 or 5 digit number of videophiles xD
1 Like
While the “red push” may have been added a bit later, the CXA2025AS decoder itself seems to do a massive push to red based on some quick experiments done in Nestopia. Does this seem correct based on the data sheets?
This seems to corroborate
This thread has so many different topics being discussed at the same time.
That old post from me that you’ve linked is outdated. A lot of the older things in my thread are messed up to some degree. The reason why I made my posts sound so confident was because no one else was sharing presets that simulate these red-oversaturating decoder chips.
and yes, the data sheet for the CXA2025AS has R-Y’s gain relative to B-Y much higher than the standard. Most other NTSC chips have R-Y’s relative gain extra high too, as you can see in the big long list below. These are all from consumer CRTs though, so we don’t know if developers’ PVMs did the same thing. We have no data for what developers’ PVMs did; they could be doing plain, unmodified SMPTE C decoding for all we know, without these oversaturated reds.
Big long list
const float demodulatorinfo[] [2][3] = {
// Index 0
// Dummy -- No color correction!
// Use this for content in the PAL or SMPTE-C that did not use color correction.
{
{90, 236, 0}, // angles (degrees)
{0.56, 0.34, 1.0} // gains
},
// Index 1
// CXA1464AS (JP)
// Used in Sony Trinitron ~1993 - ~1995
{
{98, 243, 0}, // angles (degrees)
{0.78, 0.31, 1.0} // gains
},
// Index 2
// CXA1465AS (US)
// Used in Sony Trinitron ~1993 - ~1995
{
{114, 255, 0}, // angles (degrees)
{0.78, 0.31, 1.0} // gains
},
// Index 3
// CXA1870S (JP mode)
// Used in Sony Trinitron ~1996
{
{96, 240, 0}, // angles (degrees)
{0.8, 0.3, 1.0} // gains
},
// Index 4
// CXA1870S (US mode)
// Used in Sony Trinitron ~1996
{
{105, 252, 0}, // angles (degrees)
{0.8, 0.3, 1.0} // gains
},
// CXA2061S (JP and US modes)
// Used in Sony Trinitron ~1997 - ~1999
// But datasheet doesn't state axis info! Possibly similar to CXA2060BS below.
// CX20192
// Used in Sony Trinitron "late 80's/early 90's"
// Can't find datasheet
// CXA1013AS
// Used in Sony Trinitron ~1993
// Can't find datasheet
// CXA1865S
// Used in Sony Trinitron
// Apparently an upgrade to CXA1465AS
// Can't find datasheet
// Demodulators above this point were *probably* all designed with the same phosphors in mind
// They can all be found in a cluster of Trinitron models with overlapping tubes and/or demodulators.
// And those phosphors are *probably* the ones in the Trinitron P22 gamutpoints constants above.
// Demodulators below this point *might* be designed for use with the same phosphors,
// or Sony might have changed phosphors at some point -- I don't know.
// Index 5
// CXA2060BS (JP mode)
// Used in Sony Trinitron ~??? (probably around 1997)
{
{95, 236, 0}, // angles (degrees)
{0.78, 0.33, 1.0} // gains
},
// Index 6
// CXA2060BS (US mode)
// Used in Sony Trinitron ~??? (probably around 1997)
{
{102, 236, 0}, // angles (degrees)
{0.78, 0.3, 1.0} // gains
},
// Index 7
// CXA2060BS (PAL mode)
// Used in Sony Trinitron ~??? (probably around 1997)
{
{90, 227, 0}, // angles (degrees)
{0.492111/0.877283, 0.34, 1.0} // gains
},
// Index 8
// CXA2025AS (JP mode)
// Used in Sony Trinitron ~1997
{
{95, 240, 0}, // angles (degrees)
{0.78, 0.3, 1.0} // gains
},
// Index 9
// CXA2025AS (US mode)
// Used in Sony Trinitron ~1997
{
{112, 252, 0}, // angles (degrees)
{0.83, 0.3, 1.0} // gains
},
// Index 10
// CXA1213AS
// Used in Sony Trinitron(?) ~1992
// Does not appear to have distinct JP and US modes
// It's possible this chip is either JP or US and there exists another chip number for the other.
// Theoretially, blue at a non-zero angle should mean that gains need renormalized,
// but I suspect they were not in practice.
// (0.77, 0.3, 1.0 is more similar to other chips than 0.74, 0.28, 0.96
{
{99, 240, 11}, // angles (degrees)
{0.77, 0.3, 1.0} // gains
},
// Index 11
// TDA8362
// Used in Hitachi CMT2187/2196/2198/2199
// Very likely match to the Hitachi P22 phosphor constants above.
// This chip does not have distinct JP and US modes, so one color correction matrix used for both apparently.
// It's also not clear if televisions sold in the US and Japan shared one whitepoint.
// These values are pretty wild (especially red gain). Not sure if gains should be renormalized for blue at non-zero angle. (Probably should since not doing it looks pretty bad.)
{
{100, 235, -10}, // angles (degrees)
{1.14, 0.3, 1.14} // gains
},
// Index 12
// Unknown chip used in 1989 RCA ColorTrak Remote E13169GM
// US model
// calculated from measurements by Patchy68k
// https://github.com/ChthonVII/gamutthingy/issues/1#issuecomment-2672961597
{
{94.5488524399, 255.376054365, 0}, // angles (degrees)
{0.806809988011, 0.295471057738, 1.0} // gains
},
// Index 13
// TA7644BP
// Used in 1985 Toshiba Blackstripe CF2005
// https://github.com/ChthonVII/gamutthingy/issues/1#issuecomment-2679838025
{
{107, 240, 0}, // angles (degrees)
{0.95, 0.31, 1.0} // gains
},
// Index 14
// TA7644BP Measured
// Computed from measurements of 1985 Toshiba Blackstripe CF2005
// https://github.com/ChthonVII/gamutthingy/issues/1#issuecomment-2700255641
// Unit was very deteriorated.
// Discrepancies from datasheet may be due to poor condition, or datasheet being inaccurate in the first place
{
{108.9, 243.3, 0}, // angles (degrees)
{0.996, 0.351, 1.0} // gains
},
// Index 15
// Panasonic CT-36D30B measured by Patchy68k
{
{101.580982936, 232.937328893, 0},
{0.91062624791989, 0.388131199173125, 1.0}
},
// Index 16
// TA8867AN
// From Toshiba CE-20D10, launched in 1994, manufactured in 1995
// I can't confirm, but the middle letter of Toshiba CE/CF CRTs seems to correspond to the launch year--C is 1993, D is 1994, E is 1995, etc.
// Might possibly pair with "EBU-ish" phosphors from a Toshiba patent in 1992.
// The chip has a PAL mode too with different settings.
{
{112, 237, 0},
{0.84, 0.33, 1.0}
},
// Index 17
// TA8801AN
// From Toshiba CF3272B, manufactured in 1993, and probably launched in 1993
// Might possibly pair with "EBU-ish" phosphors from a Toshiba patent in 1992
// The chip has a PAL mode too with different settings.
{
{95, 240, 0},
{0.84, 0.31, 1.0}
},
// Index 18
// TA8867BN
// Unknown what CRT this appears in, but similar chip to TA8867AN
// Might possibly pair with "EBU-ish" phosphors from a Toshiba patent in 1992
// The chip has a PAL mode too with different settings
{
{104, 240, 0},
{0.91, 0.31, 1.0}
},
// Index 19
// TA7698AP
// Datasheet paper has the year 1988
// Appears in JVC TM-9U(CV), also rebranded as the Sensormatic RM409
// I could've sworn that there was a Toshiba CRT from 1988 on crtdatabase that had this chip.
// Might pair with the same phosphors as the CF2005, but don't know.
// Also might pair with the "EBU-ish" phosphors from a 1992 Toshiba patent.
// The chip has a PAL mode too with different settings.
{
{105, 235, 0},
{1.0, 0.38, 1.0}
}
};
What Nestopia and my older presets are missing is the correct primaries, reference white, and default tint/color settings. As I’ve posted a million times by now, I am very confident that the correct primaries are somewhere near this data that Chthon came across, that the reference white is x=0.281, y=0.311 (illuminant 9300k + 27MPCD), and that the default tint and color settings are set to make the red/yellow/green region approximate 1953 primaries with illuminant C (or, for the JP axis mode, illuminant D93 / 9300K + 8MPCD, x=0.283, y=0.298), preferring correct chromaticity over correct luminance, which is very similar to what this ancient paper describes. Although there’s no decisive proof anywhere, I can’t figure out any other way to make this data make any sense, no matter what else I try. Again, we don’t have this kind of data for professional video displays, so this is no help for determining the artists’ intents. (Computer monitors are a completely different story.)
Tritone was discussed earlier. Other Sony TVs have a feature called “Dynamic Color” which can be turned off in the service menu. According to data sheets like the CXA2025AS and CXA1465AS, Dynamic Color also alters the white balance, but the data sheets don’t give any clues as to what else Dynamic Color is doing. My best guess is that it switches between a 9300k+27MPCD decoding based on that ancient paper I mentioned and doing some other decoding.
2 Likes
PVMs came in two flavors: SMPTE and EBU phosphors, both based on P22, just doped differently to give optimized persistence for each format.
The greatest source of uncertainty (in regards to color) for PVMs is white point. There are limitations with how they may be adjusted. Some can only be calibrated to D93, some to D65, and some can do both, and adjusting the white point from ‘native’ can impact linearity, so care and compromise had to be taken. Calibration was done with a pattern generator and colorimeter (basic adjustments can be done by eye).
Are PVMs/BVMs with 1953 to SMPTE C circuitry known to exist?
4 Likes
It is but I don’t think you’re giving the brain enough of the credit it deserves here.
1 Like
A slight oversight, but that would be going too far. lol
1 Like
@PlainOldPants do these screenshots look at all accurate to you in regards to color conversion circuitry? I tried Parker’s method. It seems his definition of ‘flesh tone’ must be based on whatever film or camera he was working with at the time. The simulated display here was set to D65 and I used his C illuminant matrix. The 9300 matrix with D93 gave bad color response. A distortion modeler and tone mapping was used to handle colors beyond 100 IRE. Is Bailey’s method much improved or is there some secret sauce I’m missing?
Giant image dump incoming. I implemented Bailey’s method with his distance parameter of 80%. I used NTSC 1953 Bradford transformed to D93 as the reference and the NTSC-J phosphors from the paper @DariusG posted.A distortion modeler limits color values to 110 IRE and a tone-mapper scales this down into [0, 1] range. Phoshpor gamut was compressed (actual compression; not clipped) with a perceptual technique to fit into sRGB. I then took screenshots of this output. I scaled the video level by -3 dB to ‘calibrate’ the Capcom color bars and took additional screenshots. Finally I took screenshots of without the NTSC conversion matrix but with all the other processing in place as a comparison to what an NTSC-J monitor without a conversion matrix would represent. The order of the images is uncalibrated left, calibrated right, and no conversion underneath.
Commentary
The method seems acceptable for natural colors, like Donna’s skin or the water and sky in Green Hill Zone. However, we can see it break down for the F-Zero title screen. We can see characteristic red crush on the Capcom color bar test. While it was common for consumer TVs to have this crush in the red, they also had this crush for blue and sometimes even green, and calibration could not always remove it. The biggeset concern I have is with red. It is even less pure red after the conversion, and with calibration to show full bars it comes out as nearly orange. We also see a hue shift towards red for magentas in the F-Zero title screen, but I don’t know if that’s closer or further to what consumer TVs were doing (anecdotally, I played on Port Town a lot and always thought of it as being dark; the sky after conversion seems too bright).
If anyone is interested in playing around with this, I can provide a simplified shader stack for it.
3 Likes
In my experience undersaturated red gets closer and closer to burnt orange. It feels almost like you have to clip the red to get it to actually look like pure red.
This was worse on my OLED TV than on my LCD TV.
Similar behaviour occured when Paper White Luminance was lowered. On my OLED TV when set below 450 nits, red started looking burnt orange. However when set to 630 and above it looked red but clipped, it seemed a bit harsher and there was less fine detail in the contrast.
This is purely anectdotal and may have nothing to do with the issue you’re experiencing here.
2 Likes
I think it is partially display limitation. The panel I develop on is sRGB but has a slightly wider gamut. On my TV (which is DCI native, BT.2020 compatible) I tried it out by mapping to BT.2020 instead of sRGB. It comes out slightly less orange.
I don’t have any OLEDs, can’t comment on that.
1 Like