Crestron HD-Scaler RGB 0-255 issues with Retrotink 2x Pro

[GrimShins]

I have a Crestron HD-Scaler as part of my signal chain to prevent sync losses during resolution switches on the Retrotink 2x Pro. For this purpose, it works very well, but there is one issue: the HD-Scaler does not play nicely with the 2x Pro's 480p RGB 0-255 output, thus clipping everything below 16 and above 235. Near as I can tell, the HD-Scaler interprets all CEA-861 video resolutions as though they're 16-235, regardless of how the quantization flag is set. If I set the 2x Pro to pass-thru, the HD-Scaler displays 240p 0-255 perfectly, but 480i 0-255 has the same clipping issues as 480p. This is unsurprising, as according to CEA-861 RGB signals should use the 16-235 quantization range unless they are 640x480 VGA.

My question: Is there any way to override this behavior and display 480p 0-255 correctly? The HD-Scaler's current firmware is v1.2124.00017; not sure if there's a newer firmware which resolves this issue, since I don't have nor can I obtain a Crestron account.

Alternatively, does the Extron DSC-301 HD properly interpret the RGB 0-255 signal from the 2x Pro?

Of course this whole issue could be resolved if future Retrotink firmware updates provide user-selectable quantization range, but if there's a solution to be had now I'd like to implement it.

[Fudoh]

Alternatively, does the Extron DSC-301 HD properly interpret the RGB 0-255 signal from the 2x Pro?

yes, it does, but I wouldn't recommend the combination. The Retrotink's output just isn't good enough to pair it with an ultrasharp scaler as the Extron, that provides a full 4:4:4 processing path.

[Fudoh]

Can you please quote the part of the CEA-861 specs, where it says to use limited range for RGB signals?

[Unseen]

Can you please quote the part of the CEA-861 specs, where it says to use limited range for RGB signals?

Multiple places, for example in section 5.1, which talks about default encoding parameters: "If a source determines that a sink is incapable of receiving AVI InfoFrames or is incapable of receiving YCbCr pixel data, then it shall, by default, encode the CE video format pixel data in RGB color space using limited quantization range levels."


CE Video Format is every format defined in those specs except 640x480p (VIC 1)

[Fudoh]

So, in the case of a fall back (after trying to negoiate YCbCr) it should default to limited range RGB, but not when trying to send RGB in the first place?

I need a hate figure for my crusade against limited range RGB. CEA-861 might be a good start.

[Unseen]

So, in the case of a fall back (after trying to negoiate YCbCr) it should default to limited range RGB, but not when trying to send RGB in the first place?

I quoted only parts of the section, but my reading is that CE formats should always default to limited range when using RGB and IT formats should always default to full range when using RGB.

I need a hate figure for my crusade against limited range RGB.

I like limited range RGB. It needs one bit less for the YCbCr->RGB conversion matrix coefficients and no extra pipeline step for offset compensation. ;)

[GrimShins]

Alternatively, does the Extron DSC-301 HD properly interpret the RGB 0-255 signal from the 2x Pro?

yes, it does, but I wouldn't recommend the combination. The Retrotink's output just isn't good enough to pair it with an ultrasharp scaler as the Extron, that provides a full 4:4:4 processing path.

That's good to know. Image quality doesn't need to be optimal for my use case, just need to eliminate signal dropouts during 240p->480i transitions.

In the meantime, I have ordered a simple HDMI to VGA converter; since VGA from Dreamcast is quantized correctly by the Crestron, my hope is that all analog RGBHV sources will work properly. Could be a foolish hope, though, since it uses the wrong horizontal sampling rate since it interprets The DC's resolution as VGA instead of DTV.

[GrimShins]

An update: the AmazonBasics HDMI-to-VGA adapter clips below 16 and above 235 when receiving 480p, so there's another device that expects limited range quantization with CE resolutions. The more equipment I test, the more I think that the Retrotink 2x Pro really ought to default to Y'CbCr or RGB 16-235 unless it detects the device is DVI-only. At the very least, 16-235 shouldn't clip on any HDMI device, whereas it seems much more hit-and-miss with 0-255. Selectable quantization would be the ideal of course, but I'm not sure how viable implementing it would be given that there are only two buttons on the unit.

I suppose I could try the DSC 301, but aside from the quantization issues I have been extremely happy with the performance of the Crestron. Hopefully a future 2x firmware update will render the issue moot.

[captaineos]

"What he said". I don't have written references to quote but I was a major home theatre nerd when DVD came out. Stacey Spears used to write a blog that deep-dived on scaling and I followed things like that for years. Full range RGB is only for VGA timings, with the baseline range of 16-235 as the default. It's plagued videophiles for years. I have a Pioneer laserdisc player that allows and IRE setting as my Japan Star Wars discs use an IRE of 0 which is like 0-255 RGB, yet most discs are IRE of 7.5 which is like 16-235.

[energizerfellow‌]

An update: the AmazonBasics HDMI-to-VGA adapter clips below 16 and above 235 when receiving 480p, so there's another device that expects limited range quantization with CE resolutions. The more equipment I test, the more I think that the Retrotink 2x Pro really ought to default to Y'CbCr or RGB 16-235 unless it detects the device is DVI-only. At the very least, 16-235 shouldn't clip on any HDMI device, whereas it seems much more hit-and-miss with 0-255. Selectable quantization would be the ideal of course, but I'm not sure how viable implementing it would be given that there are only two buttons on the unit.

I agree that the Retrotink devices need a range option on both the input and output sides. IRE 0.0 vs 7.5 on the input side and 16-235 vs 0-255 on the output side.

[GrimShins]

An update: the AmazonBasics HDMI-to-VGA adapter clips below 16 and above 235 when receiving 480p, so there's another device that expects limited range quantization with CE resolutions. The more equipment I test, the more I think that the Retrotink 2x Pro really ought to default to Y'CbCr or RGB 16-235 unless it detects the device is DVI-only. At the very least, 16-235 shouldn't clip on any HDMI device, whereas it seems much more hit-and-miss with 0-255. Selectable quantization would be the ideal of course, but I'm not sure how viable implementing it would be given that there are only two buttons on the unit.

I agree that the Retrotink devices need a range option on both the input and output sides. IRE 0.0 vs 7.5 on the input side and 16-235 vs 0-255 on the output side.

I'm not sure a 7.5 IRE option on the input side of things would be that useful, since I can't think of a retro console that actually has pedestal on CVBS or Y/C. And Y'PbPr always has black at 0 IRE. It could be useful for VHS and Laserdisc I suppose. Although come to think of it, the ADV7280 might automatically detect whether an input is 0 IRE or 7.5 IRE black level.

[mikechi2]

Interesting, I may give limited range RGB a shot.

I will say limited range YCbCr, despite it being the "standard" was a complete shit show with constant complaints of "washed" out colors. Full range RGB and setting the IT content flag has been almost trouble free with the exception of these various converter boxes and dongles.

[GrimShins]

I will say limited range YCbCr, despite it being the "standard" was a complete shit show with constant complaints of "washed" out colors.

I wonder how many of those complaints were from people who like to set their TVs to "Vivid" and max out the sharpness control. I wish I were being facetious, but apparently a lot of people out there really enjoy the image "pop" that comes from crushed shadows and blown-out highlights, to the point where Microsoft actually intentionally cocked up the gamma curve on the Xbox 360. That led to a lot of complaints about the PS3 versions of multiplats looking "washed out" even though it was using the proper curve.

I have encountered computer monitors that will not display 16 as black under any circumstances, though, which I suppose could have resulted in some legitimate complaints. FWIW the PS4, Xbox One, and Switch all use RGB 16-235 as their default settings. Personally, I feel like the designations "limited range" and "full range" have done irreparable harm to video fidelity, since the masses seem to think that "full" means they're getting extra detail, or that "limited" means they're losing detail. The importance of connotation is apparently lost video engineers. Naming the ranges "Video" and "PC" probably could've avoided a lot of headache. Hell, on the Xbox One the settings are called "Standard" and "PC RGB," and PC RGB comes with a warning that basically says "don't use this, it's probably going to mess things up."

Nothing's ever easy, is it?

[strayan]

Naming the ranges "Video" and "PC" probably could've avoided a lot of headache.

This is the naming convention my trusty old plasma from 2004 uses.

[Joelepain]

Naming the ranges "Video" and "PC" probably could've avoided a lot of headache.

I would argue the contrary. At least with "full" and "limited" you have a glimpse of an idea of what's going on behind. With "PC" and "Video", that means nothing. Even more in the videogame console world, where the whole pipeline of generating imagery is exactly the same as a PC.

the masses seem to think that "full" means they're getting extra detail, or that "limited" means they're losing detail.

Technically they are not totally wrong. Sure the most important aspect is to match your settings between the source and the display. But full range will always carry more color gradients than limited range, No matter what.


Limited range at the end user stage, and that's even more true with videogames, is completely useless. And I don't care about what is the standard, the legacy, etc... it's just f*****g plain stupid.
I'm sorry to repeat myself in a lot of topic about this subject, but I'm in a little crusade too, like Fudoh.
What bothers me is that there is some kind of general feeling where every one is saying "argh this stupid full range thing is a non-sense, let's get rid of it", when in reality people should fight against limited range instead.

[Fudoh]

just reiterating what I said many times before: I'm not against a limited range format, I just think limited range should coincide with YCbCr encoding and not be used with RGB.

I also don't agree with the CEA-861 recommendation: if a source can't negotiate a YCbCr transmission, chances are that it's a DVI display that only works with RGB and doesn't know shit about limited range.

I can't think of any device that's can only accept RGB, but not YCbCr, but can differentiate between full and limited range RGB levels. Who would build such a thing?

[GrimShins]

The argument that full range offers better gradient resolution is pedantic at best. Even a Granger rainbow shows minimal visible differences between limited and full quantization levels, and only at the very darkest end of the spectrum. And since retro game consoles don't even produce anywhere near 8bpp color, it's an entirely moot point in the discussion of video processors aimed at retro content. Saying that CEA got it wrong is screaming into the abyss at this point.

[GrimShins]

An update in the ongoing saga of why I hate video processors:

Today I acquired an Extron DSC HD-HD, and it does in fact handle the Retrotink's RGB full output...kind of. When switching resolutions back and forth on the Genesis 240p Test Suite, it gets confused and ends up locking into interpreting it as RGB limited, causing black crush. Additionally, I get black screens during resolution switches which are at least as bad as what I would get with no intermediary scaler at all. Oh, and the aspect ratio is only correct if I set the Extron's output to 480p, which completely defeats the purpose of having an additional scaler. So overall it's utterly useless and I will be sticking with the Crestron HD-Scaler, which does everything correctly except the RGB range, and lets me spam the resolution switch button without ever losing the image. At least the Extron was three times the price and took three times longer than FedEx said it would to get here.

Also, it appears that RGB limited output from the Extron is not properly flagged as limited, as my Asus VG248QE displays it as if it were RGB full, resulting in a washed out mess of an image. Rather annoying, since the limited output of the Xbox One and PS4 are correctly flagged and the Asus displays those properly. Y'CbCr output is displayed correctly FWIW. The Extron also includes Y'CbCr modes which use full range quantization, which is so asinine I'm actually struggling to find words to express my disgust; obviously the resulting image is completely wrong on every single device capable of receiving Y'CbCr.

My main takeaways from all of this banging my head against the wall:
-ALL of my HDMI devices display Y'CbCr 16-235 correctly
-MOST of my HDMI devices display RGB 16-235 correctly IF the content is flagged as limited range
-MANY of my HDMI devices display RGB 0-255 incorrectly because the full range flag is ignored
-I absolutely hate that the engineers for both consumer electronics and pro equipment manufacturers basically ignore CEA spec
I performed testing on monitors from Asus and LG, and televisions from Samsung and Vizio. Certainly this isn't enough data to make sweeping generalizations, but overall it does appear that Y'CbCr 16-235 is the most reliable method of getting an accurate image free of washout or black crush.

So I guess the "tl;dr" of this entire exercise in futility is that it would be really nice if the next Retrotink firmware update included options for selecting colorspace/quantization range.

[Fudoh]

With the HD-HD you unfortunately got the ONLY scaler from Extron's line-up that got totally butchered aspect ratio controls. On all other machines you have free completely free control over that.

In the end it just proves my point: we'd all be much happier if there just was full range RGB and YCbCr was locked to limited range (which it mostly is anyways).

MOST of my HDMI devices display RGB 16-235 correctly IF the content is flagged as limited range

here you're lucky, because if you if lived through the transition from DVI to HDMI, you'd have encountered enough machines that didn't know anything above DVI specs (= full range RGB only).

but overall it does appear that Y'CbCr 16-235 is the most reliable method of getting an accurate image free of washout or black crush.

you're right about that. But at least half of your problem is home made. Why don't you simply run the XBO, PS4 and the Extron at full range RGB output instead of trying to get limited range to work to your monitor ? Or the other way around: if you don't mind range, why not use YCbCr (which you found to be more reliable anyway) ?

[GrimShins]

but overall it does appear that Y'CbCr 16-235 is the most reliable method of getting an accurate image free of washout or black crush.

you're right about that. But at least half of your problem is home made. Why don't you simply run the XBO, PS4 and the Extron at full range RGB output instead of trying to get limited range to work to your monitor ? Or the other way around: if you don't mind range, why not use YCbCr (which you found to be more reliable anyway) ?

You misunderstand the issue, which is probably my fault for not clarifying. I have no issues with the output from the Xbox and PlayStation, I simply use them as an example of modern devices that output correctly flagged RGB limited. They work perfectly with every display I'm using, as well as every scaler I've tried. They're irrelevant to what I'm talking about.

My single issue with the Crestron HD-Scaler is that it interprets CE signals (480p/720p/1080p) as limited range regardless of quantization flag, causing black crush. This only affects the Retrotink 2x Pro, since it outputs a CE resolution (480p) but full range RGB, which is a no-no according to CEA-861 spec. Y'CbCr inputs are processed correctly. The Crestron outputs either RGB limited (properly flagged) or Y'CbCr; this works perfectly on every display device I have. The output colorspace does not affect how inputs are treated.

RGB full range from the Tink works fine with my Asus monitor, but is interpreted as limited range by my Vizio television, which ignores the quantization flag for CE resolutions, same as the Crestron. PC resolutions such as 640x480 are displayed with correct quantization. Y'CbCr is always displayed correctly on both the Asus and the Vizio.

This entire exercise has been about trying to build a setup where both my Asus monitor and Vizio television can display the received signals correctly at the same time. On the output side of things, the Crestron does this perfectly. The only roadblock is the Retrotink 2x Pro using a combination of 480p and RGB full range.

To sum up:
-Crestron HD-Scaler interprets RGB full from PC resolutions correctly, RGB limited from CE resolutions correctly, RGB full from CE resolutions incorrectly. Always interprets Y'CbCr correctly.
-Asus monitor interprets RGB limited/full from CE/PC resolutions correctly if quantization range is flagged correctly. Interprets RGB limited incorrectly if not flagged. Always interprets Y'CbCr correctly.
-Vizio television interprets RGB full from PC resolutions correctly, RGB limited from CE resolutions correctly, RGB full from CE resolutions incorrectly. Always interprets Y'CbCr correctly.

And not using the Crestron as an intermediary between the Retrotink 2x Pro and my displays is not an option, since I speedrun games that have resolution switches which causes sync loss for several seconds unless I'm using something to decouple input/output refresh rates. The Extron failed miserably at this task, which is why it's going in the closet.

Bottom line: in this setup, the Retrotink 2x Pro is the only device that is not working properly, and that issue could be corrected by changing its output to either RGB limited or Y'CbCr. Y'CbCr seems to be the option that's most likely to work correctly with the greatest number of setups. Mike says that Y'CbCr resulted in many complaints about color fidelity; given that Y'CbCr works perfectly with my test sample of more than half a dozen HDMI devices with manufacturing dates ranging from 2010 to 2018, I surmise these complaints are due to users who aren't image fidelity-driven preferring the "pop" of the high-contrast image caused by black crush. I'm a proponent of giving users choice, even if they use that choice to make wrong decisions, which is why I would prefer selectable colorspace/quantization. Alternatively, a download of firmware version 1.0 of the 2x Pro, which used Y'CbCr, would be a temporary solution for my personal use case.

If you made it this far, thanks for indulging me! I realize a lot of this probably seems like much ado about nothing to the outside observer, but it is very important to someone like me who strives to have all of their devices calibrated as close to SMPTE/ITU standards as possible (you should see my HCFR calibration folder!). I was an early adopter of the Retrotink 2x, and I've been incredibly satisfied with the Retrotink 2x Pro in all regards except for this one easily addressable area. With the option of either RGB limited or Y'CbCr, my setup would be perfect.

[strayan]

I surmise these complaints are due to users who aren't image fidelity-driven preferring the "pop" of the high-contrast image caused by black crush.

I would bet my house on this.

[Fudoh]

Why don't you just point Mike at this thread? If the chip supports YCbCr output, I see no problem why he wouldn't offer an alternative FW version for those who're happier with that. If I were him, I wouldn't add a switch it to the general firmware though as it would confuse more than it would help.

[strayan]

Why don't you just point Mike at this thread?

Mike responded earlier.

I will say limited range YCbCr, despite it being the "standard" was a complete shit show with constant complaints of "washed" out colors.

Maybe you have to tell them they’re wrong. It might start World War 3 though.

[strayan]

I speedrun games that have resolution switches which causes sync loss for several seconds unless I'm using something to decouple input/output refresh rates. The Extron failed miserably at this task, which is why it's going in the closet.

I love my Cresteon HD scaler (It’s the only device I’ve found that outputs reduced blanking and supports 1:1 pixel mapping) but how do you cope with the 2-3 frames of lag for speedrunning?

[GrimShins]

I love my Cresteon HD scaler (It’s the only device I’ve found that outputs reduced blanking and supports 1:1 pixel mapping) but how do you cope with the 2-3 frames of lag for speedrunning?

I don't have a Time Sleuth or anything, but I have done high speed camera testing with a CRT as a comparison and on my Asus monitor and there's only 1 to 1.5 frames of added delay vs a CRT, which actually puts it roughly identical to the display's internal 480p scaling. Gotta love the model 1 Genesis and its option of outputting RGB and RF at the same time! But in terms of lag, it hasn't affected my ability to hit frame perfect tricks at all. The screen blanking for at least 3 seconds during 240p/480i or 256px/320px transitions was much less tolerable, so the Crestron has become essential to my signal chain.

[mikechi2]

Why don't you just point Mike at this thread?

Mike responded earlier.

I will say limited range YCbCr, despite it being the "standard" was a complete shit show with constant complaints of "washed" out colors.

Maybe you have to tell them they’re wrong. It might start World War 3 though.

Send me a DM or email retrotink2 AT gmail DOT com if you wanna try a firmware with limited range YCbCr... that's a easy change for me.

[GrimShins]

Email sent! Looking forward to testing it out!