The retro gaming community sorely needs gamma-correction profiles

[TrantaLocked]

There is an aspect of retro gaming on modern TVs that has been neglected and that I believe the retro community should feel more fire to start working on. The issue is of how the imperfect brightness curve (not just level, but shape!) of the typical CRT TV is not translated well to modern HDTVs, which manifests as a flatter than ideal picture. Even when both the CRT and modern TV are set to the same gamma level, they will look very different. The effect is that games mastered on CRT monitors and tested and played on consumer CRT TVs in the 80s, 90s and early 2000s will look too bright in the lower end of the signal and too dark in the upper end, therefore appearing flat on modern TVs. Look at the differences here: https://imgur.com/a/sEOsvMm (from https://www.atari-forum.com/viewtopic.php?t=37519) Notice how bright the CRT looks in the upper end of the signal, which is completely missed when the same game is viewed on a modern TV using a perfect power-law gamma curve.

My main gripe is that this problem has been seemingly ignored by most of the retro gaming community. Even when we see examples like above and like in this LTT video (https://www.youtube.com/shorts/8o1iUy8x7Yw, https://i.imgur.com/tzqv3Rw.png) comparing Sonic on a CRT vs an OLED, which clearly shows a flatter than expected picture on the OLED as a result of a mis-matched gamma curve shape that is technically correctable, it’s brushed off as just “CRT better.” We are ignoring – or are unaware of – the core problem which is that the brightness curve shape is severely mismatched between retro CRTs and modern HDTVs and it cannot be corrected by a simple gamma level adjustment or a contrast adjustment.

I want to make it clear that changing settings like contrast or brightness level cannot account for this mismatch, only act as a band-aid. You can try and calibrate your own gamma curve if your TV supports multi-point brightness correction, but that’s asking a ton from retro gamers to do this job accurately. It would also be a giant waste of energy when we can solve this the right way.

The solution that would best benefit the retro gaming community would really pay off in the long term. It involves implementing either a single averaged – or if necessary, multiple generation-based – CRT offset-gamma profiles in retro gaming scalers and emulators. This is my suggested gameplan:

1. Roundup various retro gaming software and hardware developers from around the globe to start measuring the brightness levels and gamma curves of various popular CRT TVs from 1980 through 2005 at their default settings.

2. Develop an average gamma-curve profile, or if necessary, multiple profiles based on each console generation by averaging the most commonly used CRTs from each generation’s period.

The assumption is an offset from a perfect power-law curve, which most modern HDTVs follow relatively closely. There would be a setting for gamma level of your HDTV in conjunction with the correction profile itself. For example, the user would select “5th-Gen Gamma Profile” (based on an average of common CRTs manufactured from 1992-1999), then the user selects from a sub-tree including 2.0, 2.2 and 2.4. Gamma level 2.2 would likely be the default selection.

I’m not entirely sure what the common gamma level at 50% signal level was for CRTs, whether 2.2 or possibly higher, but step #1 would give us that data. In the end, we want to make sure the profile selector knows what gamma level the user’s TV is set at, which is typically 2.2 or in rare cases 2.4, before the CRT curve shape offset is applied. The offset will naturally decrease lower signal levels and increase upper signal levels by the respective amounts.

3. Implement said profiles as an optional selection in scalers like the OSSC and Retrotink and emulators like Retroarch.

It may not be necessary to do generation-based profiles and having even just one single averaged profile would already be a huge upgrade over what we have now – which is essentially nothing.

[Author]

Posts