Analogue announces the Super Nt for $189 (SNES)

[jayde6]

Thats very strange because all devices should detect Full/Limited range automatically (unless you use an LG display).

The Super NT currently already outputs Full Range because it outputs blacks lower than 16 so if you don't see any crushed blacks now then you also wont when the colors are fixed.

Seems like you and I are destined to do this forever, ha. The OSSC settings you gave me on videogameperfection are still working great, thanks again.

I wonder if the Super Nt has HDMI Flags that force my LG to treat it properly no matter what it is set to.

A 1-chip through the ossc (both with your settings and with the tv full range, ossc default settings) have about the same black levels for me. Then again I never noticed any crushed blacks on my Nt mini either (with the TV set to limited). Some devices like the PS4 have an automatic setting, could the mini and super nt have something similar by default?

[FBX]

Another full range RGB output is the UltraHDMI mod for the N64. It was the first time I realized there was such a thing on my TV.

[Fudoh]

if it's full range after it's fixed would using something like gamma boost put it back into limited range territory you think?

OPTIONAL gamma controls are a great idea. We just don't need (or want) it enforced in three different places: a general gamma boost present all the time (see FBX's measurements), another gamma boost enforced when enabling the scanline options and a third one actually labeled gamma boost.

ONE option with a silder. That's enough.

[paulb_nl]

Seems like you and I are destined to do this forever, ha. The OSSC settings you gave me on videogameperfection are still working great, thanks again.

I wonder if the Super Nt has HDMI Flags that force my LG to treat it properly no matter what it is set to.

A 1-chip through the ossc (both with your settings and with the tv full range, ossc default settings) have about the same black levels for me. Then again I never noticed any crushed blacks on my Nt mini either (with the TV set to limited). Some devices like the PS4 have an automatic setting, could the mini and super nt have something similar by default?

Oh right that was you lol. It could be possible that the Super NT chooses a range automatically but so far it seems that every device incorrectly goes into Limited Range mode with it.

[jayde6]

OPTIONAL gamma controls are a great idea. We just don't need (or want) it enforced in three different places: a general gamma boost present all the time (see FBX's measurements), another gamma boost enforced when enabling the scanline options and a third one actually labeled gamma boost.

ONE option with a silder. That's enough.

100% agree, with all that going on it's no wonder things are off from where they should be.
Was just wondering if using whatever user controlled gamma adjustment they decide to implement would work to get limited range values even though it's suppose to output at full range.

[FBX]

I don't know if it's feasible, but I think an actual option in the Super Nt to set limited or full range might help a lot too once the colors get fixed.

[Fudoh]

Was just wondering if using whatever user controlled gamma adjustment they decide to implement would work to get limited range values even though it's suppose to output at full range.

Gamma controls usually work between minimum black level and the maximum white level, in other words: by adjusting gamma you leave black where it belongs, but you increase brightness in NEAR-black areas. For a true full to limited range conversion you need to move the minimum black level as well.

[FBX]

Was just wondering if using whatever user controlled gamma adjustment they decide to implement would work to get limited range values even though it's suppose to output at full range.

Gamma controls usually work between minimum black level and the maximum white level, in other words: by adjusting gamma you leave black where it belongs, but you increase brightness in NEAR-black areas. For a true full to limited range conversion you need to move the minimum black level as well.

Yeah isn't it like 16 instead of 0 for black and 235 instead of 255 for white?

[thirdkind]

The Super NT currently already outputs Full Range because it outputs blacks lower than 16 so if you don't see any crushed blacks now then you also wont when the colors are fixed.

A properly configured device sending limited RGB will send data below 16 and above 235 to the display; it's up to the source material (in this case, the game) to determine if anything below or above those values is present. The display then determines what to do with those values. This goes back to a lot of complicated analog stuff that involved somewhat floaty black/white levels on CRTs that was carried forward to digital displays for compatibility (and confusion, unfortunately).

Full RGB tells the display that it should show the entire range. Black is at the bottom and white is at the top. It's much more straightforward and provides for a wider range of colors.

The question that I haven't seen addressed in this thread -- and maybe it is elsewhere -- is whether SNES games are programmed with real content in those extended ranges or if they contain information from 16 to 235 that is being remapped to a full RGB range in a conversion by Higan. For example, the game's programmers determined if the whites on the Super Mario World title screen are 235,235,235 or 255,255,255. If it's the former, then their intention was limited RGB. If it's the latter, their intention was full RGB.

Nt Mini is reported as full Range. Playstation 3 and 4 can also output Full Range. OSSC also outputs Full Range only.

Good to know, thanks. I knew about the PS3/PS4, but the recommendation for those has always been limited RGB unless you have a specific reason to use full. My display handles full RGB fine (I have an Nt Mini and OSSC as well), but not everyone's does, and the Super Nt is not the niche product the Mini and OSSC are.

Retro consoles and games were made in a time when everyone was adjusting their display settings to taste rather than to the actual standard. I never saw a calibration VHS tape with test patterns And there were wide variances in how consumer CRTs handled these signals, making much of this subjective. As long as nothing is getting clipped and the colors are being appropriately mapped to either space, it's probably "correct".

[orange808]

I can't see how derailing this with a useless (and completely inductive) argument about "developer's original intent" will help.

I also don't see how the old "Never The Same Color" line (aka analog won't ever be right, so whatever) will help.

I see feedback in this thread and it's 100% valid.

[thirdkind]

I can't see how derailing this with a useless (and completely inductive) argument about "developer's original intent" will help.

I also don't see how the old "Never The Same Color" line (aka analog won't ever be right, so whatever) will help.

I see feedback in this thread and it's 100% valid.

You're right, why solve problems with history and data when we can rely upon a lack of basic knowledge about where all this stuff came from?

Did I stumble into neo-geo.com by mistake? I thought this was the smarter forum.

[panzeroceania]

@FBX

I Know you captured data on higan and super NT, have you been able to capture data with the gamma boost turned on in both to see how that compared to you original 2 sets of data?

[FBX]

@FBX

I Know you captured data on higan and super NT, have you been able to capture data with the gamma boost turned on in both to see how that compared to you original 2 sets of data?

Although I didn't include them, I did make captures with Gamma boost on, and it only made things worse. In view of that, I tossed the comparison pics as not being meaningful to the issues at hand. However, I could simply capture them again, if you're interested in seeing the results.

[panzeroceania]

That's okay I trust you that they weren't useful. Did you try higan with the gamma boost? Or just the Super Nt?

[FBX]

That's okay I trust you that they weren't useful. Did you try higan with the gamma boost? Or just the Super Nt?

Just the Super Nt. Last I checked, Higan had sliders for gamma correction, but I'll have another look

[Xer Xian]

This goes back to a lot of complicated analog stuff that involved somewhat floaty black/white levels on CRTs that was carried forward to digital displays for compatibility (and confusion, unfortunately). [...] The question that I haven't seen addressed in this thread -- and maybe it is elsewhere -- is whether SNES games are programmed with real content in those extended ranges or if they contain information from 16 to 235

I don't think CRTs have anything to do with that. To my understanding (well, I've read this from a book), reserving a 'footroom' (below reference black) and 'headroom' (above reference white) in the video signal range is a common practice in the broadcast industry and is meant to avoid unwanted clipping on both ends that may result from applying filters or general processing of video material (that's the main reason at least). In the official Rec.601 paper - that standardized the 16-235 range for YCbCr - they are called 'working margins' (while a restricted range might have been used with analog YPbPr too, that would've been up to personal preference and not an industry standard). I think there would be no reason at all for this practice to be relevant in the video game or computer industry in general, where the visuals are simply the product of the developers' coding and there's no processing/color scaling/filtering going on (generally).

Since it has been established that the color issue is not due to limited/full range mismatch, this discussion is probably off-topic.

[thirdkind]

I don't think CRTs have anything to do with that. To my understanding (well, I've read this from a book), reserving a 'footroom' (below reference black) and 'headroom' (above reference white) in the video signal range is a common practice in the broadcast industry and is meant to avoid unwanted clipping on both ends that may result from applying filters or general processing of video material (that's the main reason at least). In the official Rec.601 paper - that standardized the 16-235 range for YCbCr - they are called 'working margins' (while a restricted range might have been used with analog YPbPr too, that would've been up to personal preference and not an industry standard). I think there would be no reason at all for this practice to be relevant in the video game or computer industry in general, where the visuals are simply the product of the developers' coding and there's no processing/color scaling/filtering going on (generally).

That's all correct, but the whole idea of overshoot/undershoot began with analog signals and displays; it was just carried over to digital signals and codified as 16-235 for the reasons you mentioned.

You're right though, it's veering off topic so I'll let it drop. I'm just glad we're getting a fix and colors will be correct.

[mufunyo]

"Limited RGB" is a non-standard that only adds to the confusion over the actual standards. ITU-R BT Rec. 601 and 709 respectively state that to digitise an analogue signal, the full analogue YUV (PAL, SECAM) or YIQ (NTSC) waveform should be digitised to 8 bits, which means that the blanking level (which is only used for blanking and not picture) is digitised as 0, and that oversaturated white (which is never legally used, but can be the result of improper amplification) is digitised as 255. This digital data is then clamped to 16-235 (that means 16 becomes 0 and 235 becomes 255) before being converted to RGB. Nowhere is there a standard that states RGB should be encoded in limited range; this rule only applies to VIDEO signals, that is the YUV or YIQ representation of a PAL or NTSC baseband signal (CVBS), Y/C separated signal (S-Video), or component (YPbPr) signal.

How this affects the SNES is as follows: the SNES outputs 0-32 (R), 0-64 (G), 0-32 (B) digital RGB which is converted to analogue RGB, and then converted to a broadcast-legal PAL or NTSC signal. If you were to capture the analogue RGB output directly using a frame grabber card such as a Datapath VisionRGB, you will find values that are 0 to 255. If you were to capture the PAL or NTSC signal, through either the composite (CVBS) output or the S-video output on the multi-AV connector, your capture card should generate YUV values that are within 16-235, as per ITU-R BT Rec. 601. If you select an RGB colour space, even though you are capturing a PAL or NTSC signal, the card should convert the values to 0-255, because that is standard for the RGB colour space.

How this affects the Analogue NT depends on whether it is outputting HDMI in YUV colour space or RGB colour space. Since HDMI is a superset of DVI, devices should always accept RGB (even if only as a fallback compatibility mode). It is therefore up to the display controller in the Analogue NT whether to output YUV or RGB. However, the same rules apply:

If and when it is outputting YUV, the Y values should fall within 16 (black) and 235 (white).
If and when it is outputting RGB, the R, G and B values should fall within 0 (black) and 255 (white).

That is all there's to it.

[paulb_nl]

If and when it is outputting YUV, the Y values should fall within 16 (black) and 235 (white).
If and when it is outputting RGB, the R, G and B values should fall within 0 (black) and 255 (white).

That is all there's to it.

It's that easy! Except that there is a Limited RGB standard in HDMI which outputs 16-235. Thats what causes the confusion. I think the only thing we all can agree on is that this digital Limited RGB garbage should have never existed.

[Artemio]

I'm late to the discussion, but I just received my Super NT.

Tested it with default factory settings, and indeed there is black crush. I have my TV calibrated and use a DVDO Edge to monitor the signal and convert color space if needed. This is of course irrelevant, since theTv calibration and setup was the same when comparing the output of a real SNES via OSSC vs the Super NT, and no AUTO setting was in place. My TV and the DVD are both capable of setting Full Range or Limited, and the Edge can convert if needed or instructed to. This is not the issue either.

Checking with the Suite, the Super NT presents black crush. This is evident when using the PLUGE and the Color bars patterns.

If I enable Gamma Boost, it doesn't fix the issue, but some bars are visible with the 7.5 NTSC PLUGE. Of course we need the 0-255 colors to be displayed, since games don't respect NTSC pedestal at all. (Either because they were developed for NTSC-J or developers worked on the digital domain, regardless games ate in full range RGB).

Just to verify, RGB is sent as RGB 24 bit full range from the Super NT.

[FBX]

Just to verify, RGB is sent as RGB 24 bit full range from the Super NT.

Which means people will have to manually set their displays to full RGB. Auto-detect at least on my Sony wants to go with limited range, but we know that's wrong because of the color crush.

[Seraphic]

Sounds like kevtris should have an update out soon to address RGB color space issue. His latest post: http://atariage.com/forums/topic/242970 ... ?p=3964862

yeah you are right. I think I have it figured out how to make it report properly so it displays as full range. I have this on the list of fixes to add for the next release so hopefully that will fix the problem for people. Funny how we still can't shake the CRT stigma on HDMI with overscan, interlace, and limited range (to account for synch pulses and such AFAIR). The fix will be to report as "computer" which should clear up the range issues.

I am going to add separate rgb gamma sliders so that should quell all the other issues people had with the brightness stuff, and recalculate how the levels are generated. Maybe in the future I will do some testing on various model SNES to get actual voltage levels to see how linear the DACs are (there's four DACs - the three RGB ones and the overall brightness DAC)

[FBX]

Sounds like kevtris should have an update out soon to address RGB color space issue. His latest post: http://atariage.com/forums/topic/242970 ... ?p=3964862

yeah you are right. I think I have it figured out how to make it report properly so it displays as full range. I have this on the list of fixes to add for the next release so hopefully that will fix the problem for people. Funny how we still can't shake the CRT stigma on HDMI with overscan, interlace, and limited range (to account for synch pulses and such AFAIR). The fix will be to report as "computer" which should clear up the range issues.

I am going to add separate rgb gamma sliders so that should quell all the other issues people had with the brightness stuff, and recalculate how the levels are generated. Maybe in the future I will do some testing on various model SNES to get actual voltage levels to see how linear the DACs are (there's four DACs - the three RGB ones and the overall brightness DAC)

Awesome. As long as I can get my linear 0 to 255 RGB output from the Super Nt (like Higan does), I'll be happy.

[GojiFan90]

I have a setup question for my Super NT: what should I set for screen size? I'm playing on a plasma at 720p with scanlines on, gamma boost and vertical interpolation off. Should I set to 1:1. 4, 4.5, 5?

[thirdkind]

I have a setup question for my Super NT: what should I set for screen size? I'm playing on a plasma at 720p with scanlines on, gamma boost and vertical interpolation off. Should I set to 1:1. 4, 4.5, 5?

Is your plasma 720p or 1080p? Sorry, I'm not clear. I don't agree with those who are saying that 720p provides the best scanlines on 1080p displays; I think either of the settings below provides cleaner output. Just make sure your display is set to 1:1 so there's no overscan.

I have a 1080p plasma and I'm happy with the following settings. They'll crop a few pixels at the top and bottom, but I see nothing meaningful lost in any game I've played so far:

Resolution: 1080p60
Width: 1600
Height: 1200
Vertical position: 45
Scaler: No scaler, Disable Vertical Interpolation


The following settings provide pixel-perfect output with zero cropping and no interpolation necessary, but you'll have empty space on all sides because the image won't fill the screen vertically:

Resolution: 1080p60
Width: 5x (1280)
Height: 4x (960)
Scaler: No scaler, Disable both horizontal and vertical interpolation


It also depends what you consider the "correct" aspect ratio for SNES games. My take is that we played the original console on 4:3 TVs, so even though the SNES's internal aspect ratio is 8:7, a CRT would've stretched it to 4:3 anyway, so either of these settings gives you a 4:3 aspect ratio, but feel free to choose what you like best and enable horizontal interpolation if you see any shimmering while scrolling.

[FBX]

It also depends what you consider the "correct" aspect ratio for SNES games. My take is that we played the original console on 4:3 TVs, so even though the SNES's internal aspect ratio is 8:7, a CRT would've stretched it to 4:3 anyway, so either of these settings gives you a 4:3 aspect ratio, but feel free to choose what you like best and enable horizontal interpolation if you see any shimmering while scrolling.

FYI the Super Nt actually labels the correct signal padding formula of 256 * 8/7 for CRT aspect correction. Both the NES and SNES use this same formula for horizontal aspect correction. So for example if you wanted a perfect aspect-corrected 4x scale image: 256 * 4 = 1024, 1024 * 8/7 = 1170. Then you just straight-scale the vertical to 960 (technically 896, but the Super Nt includes overscan).

The same method applies to 5x scale: 256 * 5 = 1280, 1280 * 8/7 = 1462 (rounded down to an even number). And the vertical at 5x is 1200. Thus, aspect corrected 5x scale is 1462x1200. When you set the Super Nt to either 1170x960 or 1462x1200, it labels both as "4:3 aspect correction for 16:9 displays". When using these resolution modes, be sure that "disable interpolation" is checked for vertical and unchecked for horizontal.

I made this proof of concept picture showing how merely pushing the active SNES graphics to 4:3 is incorrect, and the 256 * 8/7 formula gives the perfect aspect correction:



Edit: almost for got to mention the formula for 3x scale: 256 * 3 = 768, 768 * 8/7 = 878 (rounded up). So aspect corrected 3x scale would be 878x720 with vertical overscan. I assume the Super Nt rounds up to 878, but if not, then it would jump down to 876 to avoid odd numbers.

Edit 2: Just checked and Kevtris rounded down to 877. I don't agree with that for two reasons:

1. The final value is much closer to 878 than it is 877.
2. Even numbers allow for perfect centering of the image.

Just my opinion anyway.

[thirdkind]

I've seen both approaches labeled "correct" and I've given up on trying to make a determination

The 4:3 stretch is very minor (undetectable by most without a side-by-side) and since both approaches require horizontal interpolation, it's slightly blurred in both. Since my previous post, I've actually been using 6x horizontal and 5x vertical (1536x1200) with all interpolation disabled. It's only a very slight horizontal stretch compared to 1462x1200, but it's super sharp thanks to no interpolation and 1:1 pixel mapping on my display.

I'm probably going to leave it there for a bit and see how it feels.

[orange808]

I can't see how derailing this with a useless (and completely inductive) argument about "developer's original intent" will help.

I also don't see how the old "Never The Same Color" line (aka analog won't ever be right, so whatever) will help.

I see feedback in this thread and it's 100% valid.

You're right, why solve problems with history and data when we can rely upon a lack of basic knowledge about where all this stuff came from?

What the f#$k does "software developer's original intent" have to do with the colors the hardware should output? I can tell you the 240p Test Suite's developer intent--and the colors are coming out wrong.

Test patterns look wrong. There's your data and history.

What difference does software developer intent in SMW make here? (None.)

Did I stumble into neo-geo.com by mistake? I thought this was the smarter forum.

That's funny, I was thinking the same thing.

Skip on over to the color temperature thread for a dose of neogaf "software developer's original intent" random guessing and speculation.

[thirdkind]

I can't see how derailing this with a useless (and completely inductive) argument about "developer's original intent" will help.

I also don't see how the old "Never The Same Color" line (aka analog won't ever be right, so whatever) will help.

I see feedback in this thread and it's 100% valid.

You're right, why solve problems with history and data when we can rely upon a lack of basic knowledge about where all this stuff came from?

What the f#$k does "software developer's original intent" have to do with the colors the hardware should output? I can tell you the 240p Test Suite's developer intent--and the colors are coming out wrong.

Test patterns look wrong. There's your data and history.

What difference does software developer intent in SMW make here? (None.)

Did I stumble into neo-geo.com by mistake? I thought this was the smarter forum.

That's funny, I was thinking the same thing.

Skip on over to the color temperature thread for a dose of neogaf "software developer's original intent" random guessing and speculation.

Just point to the spot on the doll where he touched you.

[rule_z]

More than a week has passed and the Analogue customer service is still asking me to clean my carts.
I didn't expect this kind of low level support from them, It's like they're trying to exhaust me and left me with my broken unit.