True--I'm just spitballing use cases for LUTs here. I don't think many people who have the NESRGB flash new palettes, and just end up using whatever selection they/their modder chose during install.Guspaz wrote:The problem with using a LUT for NES palettes is that you must have a known fixed source pallet to transform it to the target, at which point you might as well just load the target palette directly onto the NESRGB in the first place.
Another great usage, and this will likely get me in trouble, would be to play 8 and 16bit games in D93 the way the gods intended with a D93 LUT, while viewing other systems in D65. Or being able to instantly jump between the two to play in whatever you prefer.fernan1234 wrote:Yeah, for the PC Engine the RGB -> YUV table is known now, which is why you can get accurate results like with the MiSTer TG16 core. But that's not the case with the NES, which is why there is still such a jungle of RGB palettes for it. The only way to get the "real" colors is by cloning the actual NES PPU, like what Kevtris did with the Nt Mini that can thus output the "real" colors via S-video, which can then be processed by a scaler much more nicely than composite.
Now this is a great idea!HDgaming42 wrote:Another great usage, and this will likely get me in trouble, would be to play 8 and 16bit games in D93 the way the gods intended with a D93 LUT, while viewing other systems in D65. Or being able to instantly jump between the two to play in whatever you prefer.
So here's the PVM-2530 raw D93 read against D65*, followed by a D65 LUT created with 100 patches, and finally a D65 LUT generated via over 1000 test patches.HDgaming42 wrote:I did a color check on my PVM-2530 and it's way out to lunch. So much so that a LUT won't be able to dial it all in--I'll have to open it up to start from a better baseline to get average delta errors under 3. In the meantime I'm running a 12hr patch generation to see how much closer I can get it. Eventually I'll be able to compare a 33 vs 65pt cube.
I'll report back.
Looks like I was correct that the OSSC Pro is using Intel scaler IP in their FPGA. I assume you use some Xilinx IP too? and BTW, how can you buy HDMI and HDCP related ICs? they require 15k of licensing that I am sure you are not paying. Unless you use some IC from China without the need to get license to buy it. I am interested. Maybe you could go for DisplayPort instead of HDMI, then people put a small cheap adapter to HDMI.Not just a bit and if anything it's the other way around. The Morph's custom scaler is more capable than the Intel Scaler IP currently being used in the OSSC Pro, it has Composite/S-Video support, a built-in DAC, and a very capable wifi system processor.
Why are you sure of this?VEGETA wrote:they require 15k of licensing that I am sure you are not paying.
And that's a L.Extrems wrote:Try searching for "Pixel Fx" here.
Just in case it's not clear to others, Pixel Fx is indeed an HDMI adopter listed there, listed since January 2021, as are several other retro companies.Dr. Claw wrote:And that's a L.Extrems wrote:Try searching for "Pixel Fx" here.
I saw HD Retrovision in there too... wonder what they're cooking up.
Pretty interesting that you wrote "even Framemeister" in that list since the 5X is already 100% better than it in every single way, and so was the OSSC since a couple of years ago in every way except compatibility, input type options (no composite/S-video), and deinterlacing (though at least it did it without the lag of the FM).VEGETA wrote:I am interested in what scaler IP you are using to achieve your result, and what killed features you have over 5x or OSSC or even Framemeister.
I don't own a framemeister nor recommend it to people, I recommend OSSC for most of the times. I have an OSSC from Aliexpress that works well.. my only complain is bad 480i handling since it outputs lots of flicker. I play PS2 only on it, and I had to change the PS2 internal resolution to 480p using GSM to be able to enjoy it better on OSSC.fernan1234 wrote:Pretty interesting that you wrote "even Framemeister" in that list since the 5X is already 100% better than it in every single way, and so was the OSSC since a couple of years ago in every way except compatibility, input type options (no composite/S-video), and deinterlacing (though at least it did it without the lag of the FM).VEGETA wrote:I am interested in what scaler IP you are using to achieve your result, and what killed features you have over 5x or OSSC or even Framemeister.
But in any case, have you checked out the roadmap linked on the first post? It lists a lot of killer features that should make you drool. And some of woozle's responses in this topic add even more things to look forward to.
The latency cost also has to be taken into account. The FM pays for its deinterlacing with more latency, and I wonder if those advanced methods could be implemented in FPGA with less latency than the FM.Fudoh wrote:whether you think that a 5X is enough for your deinterlacing needs, is really up to your personal needs, but from an objective view the FM is stil considerably better in that regard. But that's not really surprising given that the FM uses probably THE best video processor IC created during the Full HD era. Any FPGA coming even close is a marvelous achievement already.
The 5X's scaling is pretty good. I'm not sure that many people would profit from a better algorithm. A more flexible engine (with free scaling on both axis) would be welcome of course.
Marqs stated that the "full version" of Intel's deinterlacing algorithm (with diagonal interpolation added to match the Kyoto's capabilities in the FM) is too heavy for the FPGA used and the road map for the Morph also "only" mentions motion-adaptive deinterlacing without going into detail which features will be available.
So at the moment I ASSUME that the 5X, the OSSC Pro and the Morph will all use a rather simple motion-adaptive deinterlacing algorithm and the FM will continue to reign supreme in this particular category. (Important note: Of course I fully agree that the 5X is a much better overall package for almost all users).
I kind of agree with you on this, particularly with 480i sources. The first time I used one of the RetroTinks instead of another off-brand scaler on PS2, I was amazed at how "much like a CRT" it looked for 480i sources. I still like its "bob" use a little more than the OSSC for 480i.Josh128 wrote:I guess de-interlacing is a big thing to some folks, but personally, I actually prefer simple bob over all other techniques as it replicates the look of actual 480i on a CRT and should technically be one of the least latency-additive ones. Every single fancy de-interlacing technique Ive ever seen looks great on still pictures, equal to 480p, but as soon as motion starts they all (every single one that Ive seen) fall on their faces. Its impossible to create image data that doesnt exist, no matter how good you fake it.
I've played with the de-interlacing on the GBS-C and RT5x pro a bit, the motion adaptive GBS-C de-interlacing is very impressive on a VGA CRT, again, until things move. Bob, on the other hand, perfectly replicates 480i on a 15KHz CRT when applied to a 31KHz CRT, and thats my ideal picture, as true to the perfection of technology that I grew up playing these games with as possible.
This. With respect to Fudoh, but honestly who really cares if the FM's deinterlacing merely looks a bit nicer to the eye than the adaptive deinterlacing of the 5X and ostensibly this Morph and the OSSC Pro, when this arguably nicer-looking deinterlaced picture comes at the cost of horrible lag, which I'd say matters for all games, even text-based menu-heavy games. When I select a menu item on an RPG or even text adventure game I want it to feel snappy like on a CRT.anexanhume wrote:The latency cost also has to be taken into account. The FM pays for its deinterlacing with more latency, and I wonder if those advanced methods could be implemented in FPGA with less latency than the FM.
Not many TVs support DP, only monitors.VEGETA wrote:and just a side note: why not use DisplayPort instead of HDMI, then rely on adapters to deliver the signal? this will relief you from licensing and fees.
I think the suggestion is that it's trivial to convert DP to HDMI with a cheap dongle, ostensibly with no effect on picture quality.ldeveraux wrote:Not many TVs support DP, only monitors.
I think you hit the nail on the head here. For movies / video, etc, good de-interlacing replicates what you see in a movie theater / film, and I can understand why some videophiles desire the perfect de-interlacing solution because it helps replicate the way they were originally filmed/presented, but for games, 480i games were always 480i, so anything other than bob-type deinterlacing solutions are presenting them in a historically un-natural / inaccurate way. And of course, the least amount of added input lag is of utmost importance for games, and makes zero difference for video/movies.fernan1234 wrote:
But this is all if you prefer a progressive-looking deinterlaced picture. I agree even more with the above comments about bob+"scanlines" (the bob in the GBS-C, bob+100% scanlines on OSSC Pro, or "CRT simulate" on the TINK products) is the ideal solution, since not only does it look most authentic as interlaced pictures, but is the fastest. We're not talking about video or film here, but games that were made to be played on CRTs. The ideal is for them to look and feel like on a CRT after scaling.
Right, though one detail I forgot to mention is that when you use this approach you darken the picture by half, just like when you use full scanlines for 240p, which can be tricky to compensate for on the scaler or monitor side to restore a bright CRT-like picture without crushing whites and colors or elevating blacks (which is particularly bad if using OLEDs). And this darkening problem is further compounded if you use BFI, which is the only way on modern TVs to approach the original motion clarity that is also essential for a CRT-like experience. Using gamma compensation or LUTs is probably the best approach, which the Morph could use to the benefit of both 240p and 480i games with scanlines. Then I dream further of a scaler that can add a rolling black bar to emulate CRT scanning (of course it would have to be a buffered output, so lag would be a tradeoff).Josh128 wrote:I think you hit the nail on the head here. For movies / video, etc, good de-interlacing replicates what you see in a movie theater / film, and I can understand why some videophiles desire the perfect de-interlacing solution because it helps replicate the way they were originally filmed/presented, but for games, 480i games were always 480i, so anything other than bob-type deinterlacing solutions are presenting them in a historically un-natural / inaccurate way. And of course, the least amount of added input lag is of utmost importance for games, makes zero difference for video/movies.
Are you sure this is the case though?-- When using the GBS-C on my VGA monitor, I dont remember a discernable difference in brightness between passthrough native 480p and 31KHz bob 480i when I tested the other day, not saying you are wrong here, you may indeed be right, but I also know for certain that the Panasonic TC-P50-X60 I used to have used internal bob deinterlacing when fed a 480i signal and it looked every bit as bright as 480p without touching settings if I recall. It looked amazingly CRT like.Right, though one detail I forgot to mention is that when you use this approach you darken the picture by half, just like when you use full scanlines for 240p, which can be tricky to compensate for on the scaler or monitor side to restore a bright CRT-like picture without crushing whites and colors or elevating blacks (which is particularly bad if using OLEDs).
Just to be clear, "bob deinterlacing" alone does not darken the picture. It looks annoyingly flickery precisely because it's not blacking out every other line (or field for each frame). That has to be what your Panasonic does, and what the OSSC does by default. But if you activate alternating scanlines on the OSSC and turn them up to 100% then you get the authentic interlace look, at the cost of halving the brightness. Same with the TINK's "CRT simulate" mode, as well as the GBS-C's bob mode. I couldn't say why you didn't notice the difference when using the GBS-C on your VGA monitor if you switched to bob (adaptive deinterlace is the default), but it has to be the case by necessity--every other line is black! It's like outputting 240p line doubled to 31khz + full scanlines on a VGA CRT, it's half as bright as 240p is at 15Khz on an SD/multiformat CRT, and it's the reason why people do the whole 240p@120Hz on VGA CRTs.Josh128 wrote:Are you sure this is the case though?-- When using the GBS-C on my VGA monitor, I dont remember a discernable difference in brightness between passthrough native 480p and 31KHz bob 480i when I tested the other day, not saying you are wrong here, you may indeed be right, but I also know for certain that the Panasonic TC-P50-X60 I used to have used internal bob deinterlacing when fed a 480i signal and it looked every bit as bright as 480p without touching settings if I recall. It looked amazingly CRT like.
And yes, OLEDs are particularly troublesome because they have terribly low nits for SDR signals, compared to LCDs, as well as more aggressive ABL. The best thing for OLEDs would be to flag the output of the scaler as HDR, which is what OLEDs do best.Josh128 wrote:If it is the case, you are correct about it really only being possibly an OLED issue as modern LCDs get easily bright enough to counter any brightness loss in those modes.
well, you are correct about HDMI and HDCP (this one is more important) fees being sometimes necessary. However, for someone who doesn't have near enough money, it is a dead end. I tried messaging HDCP people by telling them I could pay per unit sold or after selling and getting profits... but they always respond with a no. buying the license and paying the annual 15k is a must before anything else.fernan1234 wrote:I think the suggestion is that it's trivial to convert DP to HDMI with a cheap dongle, ostensibly with no effect on picture quality.ldeveraux wrote:Not many TVs support DP, only monitors.
But the user who made that suggestion may be missing the fact that as a manufacturer of these devices it can make all the sense in the world financially to just swallow the HDMI fees and offer the convenience to increasingly demanding customers.
bobrocks95 wrote:When did everyone get so sensitive to lag? 5-10 years ago 50ms+ LCDs were the norm, and people added scalers on top of that and were happy. Now you've got stuff like that guy in the 5x thread who refused to buy it because it was 4ms slower than the 2x or something close to that.
2 frames of lag on the Framemeister is of course worse on paper, but pair it with a fast display and show me the double blind test that shows anyone can reliably tell the difference. I think the placebo effect is strong here (and, again, sure, the 5X pretty much replaces the FM, yes).
I game primarily on flat panels (LG PC monitor 75hz) and don't feel a thing really, using OSSC and a general component to HDMI product. Yes, input lag is overrated these days, but people think that as long as the ICs are capable of delivering 0 lag then they should.fernan1234 wrote:bobrocks95 wrote:When did everyone get so sensitive to lag? 5-10 years ago 50ms+ LCDs were the norm, and people added scalers on top of that and were happy. Now you've got stuff like that guy in the 5x thread who refused to buy it because it was 4ms slower than the 2x or something close to that.
2 frames of lag on the Framemeister is of course worse on paper, but pair it with a fast display and show me the double blind test that shows anyone can reliably tell the difference. I think the placebo effect is strong here (and, again, sure, the 5X pretty much replaces the FM, yes).
It's not so much a placebo, but rather impressions will depend on what you've been used to. I'm not special, I've simply been using CRTs all my life, and only started really getting serious with scalers and flat panels this year. There's many people who've been on the scaler + flat panel game longer but have also been using CRTs on the side and thus remain alert to the differences. I could immediately tell a "big" (to me) difference between the 5X's triple buffered mode and the framelock mode, whereas other users on that topic were saying it feels the same to them.
Here are Time Sleuth results for 2 different LG 75Hz monitors, and as you can see, they are basically 0 lag. I was thinking my 1080p monitor was a bit slower but I was mistaken-- in these tests its a few tenths of a ms faster, but that could be attributed to the fact that I was not feeding native res (my TS is not currently capable of 1440p) to the QHD monitor. This is within .5 - 1.5 ms of what I measure on an actual CRT with a zero lag HDMI to VGA converter. Im guessing this is the norm for most flat panels these days, up until a few days ago I did not realize that.VEGETA wrote: I game primarily on flat panels (LG PC monitor 75hz) and don't feel a thing really, using OSSC and a general component to HDMI product. Yes, input lag is overrated these days, but people think that as long as the ICs are capable of delivering 0 lag then they should.
I am not familiar with 5x modes, but "framelock" seems like "genlock" right? where input frame v sync is phase-aligned with output frame v sync. right?
