RetroTink 4K All But Officially Announced ....

[BuckoA51]

Yes don't misunderstand me, OSSC Pro and Morph can both scale these resolutions (up and down). Down to 240p or up to their maximum resolutions.

[orange808]

Let's wait and see.

The Retrotink4k isn't hiding the hardware from me.

Want to speculate about the maturity of the tool chain that Mike uses? No problem. I see it's Altera. I doubt there will many hardware defect or upstream "gotchas" later on. Gotchas happen. For instance, the inexpensive GoWin chips fuck up when they fill up. That's a nasty surprise.

Want to talk about the limits of the hardware? We can do that. We know what's in there.

It's not a big secret. It's not a mystery box.

Look for yourself. The numbers and labels aren't sanded off.
https://www.retrotink.com/product-page/retrotink-4k

What's inside the Morph? Heck if I know. ¯⁠\⁠_⁠(⁠ツ⁠)⁠_⁠/⁠¯

[bobrocks95]

Let's wait and see.

The Retrotink4k isn't hiding the hardware from me.

Want to speculate about the maturity of the tool chain that Mike uses? No problem. I see it's Altera. I doubt there will many hardware defect or upstream "gotchas" later on. Gotchas happen. For instance, the inexpensive GoWin chips fuck up when they fill up. That's a nasty surprise.

Want to talk about the limits of the hardware? We can do that. We know what's in there.

It's not a big secret. It's not a mystery box.

Look for yourself. The numbers and labels aren't sanded off.
https://www.retrotink.com/product-page/retrotink-4k

What's inside the Morph? Heck if I know. ¯⁠\⁠_⁠(⁠ツ⁠)⁠_⁠/⁠¯

Morph uses an Efinix Trion T120. Their build chain is set up for Intel, Xilinx, and Efinix chips, so I'm not sure that will end up being much of an issue either, not like they haven't been working with FPGAs for years already.

I don't think PixelFX have intentionally hidden anything, and Mike didn't share the FPGA for the Tink 4K until just now when the store page went up, it was always blocked by a clown emoji in any prior pictures. They're behind on showing more but the hardware isn't a "mystery box".

[tongshadow]

Yes don't misunderstand me, OSSC Pro and Morph can both scale these resolutions (up and down). Down to 240p or up to their maximum resolutions.

Interger scaling 720p/1080p to 2160p is what most people are interested in, so like I said, that's one point in favor of the Tink4K.

[VEGETA]

I noticed the fpga used is this: https://www.digikey.com/en/products/det ... 7N/3879486

it is about 270$ in low quantities, so 750$ is doable imo.

[Guspaz]

Fun fact: the RT4K FPGA is around 36% larger than the one used in the MiSTer. It's a hefty boy for hefty scaling.

[Guspaz]

Interger scaling 720p/1080p to 2160p is what most people are interested in, so like I said, that's one point in favor of the Tink4K.

A long time ago I asked Marqs if the OSSC Pro would be able to scale 720p to 1440p in 4:4:4, and the answer at the time was no. Maybe that's changed since then, but I've used the RT4K to scale my retro macintosh from 720p to 1440p (I have a 4K TV but not a 4K monitor) and it works perfectly, letting me use the mac without blurry scaling for the first time ever. It's been a real treat.

[orange808]

but the hardware isn't a "mystery box".

It a mystery until it isn't. In the meantime, please don't tell me up is down. I know my own directions.

I understand the worries on price, but that depends on the person. Up until very recently, you couldn't get a quality 4k video processor of any kind for $750usd. I have no doubt there's a profit margin in there. Of course there is.

The case looks great. We don't know anything about Mike's overhead costs. I have no idea what kind of yield a person can expect from manufacturing. Building and maintaining the firmware has already been time consuming--and it will continue to be. There's also customer service; that's a another headache--not to mention the hassles of dealing with PayPal.

Gross is not editda. Everything is more expensive that you think.

....

@Guspaz: I see the OSSC Pro performs (essentially) pure frame doubling with a frame lock. Any chance we can get frame double and frame quad options at 1080p? I really want that low lag tate.

[Guspaz]

The RT4K can take a 60 Hz input and double it, sure. Not at 4K, of course, that exceeds the HDMI 2.0 output bandwidth. But it can do it for 1440p120. Or 1080p120 if you want. I think it can do 1080p240, but I've not tried and have no way to test it (I don't have any 240 Hz displays).

Normally the 120 Hz is intended for you to add BFI, but if you want to just let it double the frames, you can do that too.

[orange808]

The RT4K can take a 60 Hz input and double it, sure. Not at 4K, of course, that exceeds the HDMI 2.0 output bandwidth. But it can do it for 1440p120. Or 1080p120 if you want. I think it can do 1080p240, but I've not tried and have no way to test it (I don't have any 240 Hz displays).

Normally the 120 Hz is intended for you to add BFI, but if you want to just let it double the frames, you can do that too.

Great. I'm hoping for 240Hz in the future. Four milliseconds is four milliseconds. I know I sound crazy now, but you'll be surprised how many PCB owners will eventually use it. Looks like there's enough horsepower, unless there's something slow with the rotation. I don't know how Mike handled the memory issue in tate.

[Guspaz]

I think it can already do 1080p240, I just don't have any way to test it.

[Josh128]

I think it can already do 1080p240, I just don't have any way to test it.

I just wish Mike would have done a 240p120 mode for the 5X. The hardware should be perfectlly capable.

[Guspaz]

What's the use case? On a CRT, it introduces massive motion blur and has the same brightness loss as 480p60 with 100% scanlines.

[Josh128]

What's the use case? On a CRT, it introduces massive motion blur and has the same brightness loss as 480p60 with 100% scanlines.

You replying to me Gus? Because if you are, you are completely wrong lol, at least for what Im specifically talking about--feeding 31KHz CRTs. Feed 31KHz 240p120 to a VGA CRT, and you get a much brighter image than 480p60 with 100% scanlines. Should be about twice the lumen output. As far as motion blur, its not blur-- but there is indeed a double strobe illusion on fast moving objects. This could be remedied with frame interpolation without affecting brightness. I tested this a couple years ago. These photos were taken from the same tripod, camera, shutter speed and exposure. The 240p120 is much brighter, it even caused overexposure in the photo-- its even easier to see in person and its not washed out like it looks in the overexposed photo-- the colors have an awesome pop to them in comparison.



Now, if you are talking about feeding 240p120 to OLEDs or LCDs, Im sure you may be correct as I have no idea on those, but no, thats not what Im talking about.

Also, Mike posted some interesting hardware info on a Twitter thread today, see link below.

https://twitter.com/retrotink2/status/1 ... 0471843971

[Guspaz]

You kind of just proved my point though. You get terrible motion issues without BFI with 240p120, and with it, you get an identical image to 480p60 with scanlines. So... just use 480p60 with scanlines.

[Josh128]

You kind of just proved my point though. You get terrible motion issues without BFI with 240p120, and with it, you get an identical image to 480p60 with scanlines. So... just use 480p60 with scanlines.

I dont believe I did, as your brightness claim was false, and your other claim that it caused massive motion blur is not true either- as the issue presents differently from blur on an LCD and its not necessarily bad or even noticeable at all for a lot of content. Games like RPGs, strategy games, and even slower paced action games like Prince of Persia, Star Fox, Pilotwings, etc dont show it at all. It mainly only presents on faster side scrollers and shmups ( projectiles and other fast sprites) and even then the effect that it has on motion resolution is no worse than an average 60Hz LCD display.--the LCD presents fast motion as blur, the CRT as a double strobe illusion-- both are different, but of similar magnitude. A perfect example of what Im talking about is take the classic game Ghouls N Ghosts-- throw the lance: there is a blue jewel near the base of it that you can clearly see when paused. In motion, it remains clear on a 60Hz CRT, but is hard to visuallyl resolve on both the LCD and the 120Hz scanning CRT.

You also casually ignore that there is a potential solution to the issue, one that does not affect brightness, which would be motion interpolation for the 120Hz mode. Marqs has said in the past it would be possible. Regardless, no matter how you try to justify it, its not at all the same as 480p with scanlines.

[orange808]

240p @ 120Hz scanlines look thicker on my monitors. The persistence blur is real, though. I can see blur. It's real blur.

You get thicker scanlines, if you want that.

[Josh128]

240p @ 120Hz scanlines look thicker on my monitors. The persistence blur is real, though. I can see blur. It's real blur.

You get thicker scanlines, if you want that.

By thicker scan lines, you mean blank lines? Thats really strange, as they are thinner on every monitor Ive tried, including the comparison photo above I posted above. The technical reason they are thinner is because the actual lit / scanned lines are double scanned and thus brighter and thicker than they are with 480p + fake scanlines and thus leaves less blank screen real estate for non-scanned lines than 480p + fake scanlines. Theres no technical reason that I can think of for the blank lines to appear thicker in the 240p120 mode. In the absolute worst case they should be the same, but the scanned lines would be still be brighter.

As far as the decreased motion resolution from the double strobe, if you want to call it blur thats fine, even though its technically different than the persistence blur of a sample and hold display. Every single one of us here have experienced it every time we've played a 30/25fps game on a 60/50Hz refresh consumer CRT. Think Streets of Rage, Metal Gear Solid, Super Mario 64 etc-- almost every 3D game of the Saturn/PS1/N64 generation. So, for every single one of those 30 or sub-30 fps games, there is no blur, no double strobing that isnt exactly the same as how we experienced it 25+ years ago. So those games wouldnt require any motion interpolation with a 240p120 mode. Its only 60fps games that would.

Regardless of the semantics I totally agree that it decreases motion resolution vs matched CRT refresh rate / game fps--- but only by roughly the same amount that any 60Hz LCD does. Its most noticeable in bright scenes with fast moving objects. Scene brightness affects its perception quite a bit-- I always thought Doom 64 "appeared" to animate more smoothly than a lot of other 30fps games, the reason for it is now clear-- its because of its mostly dark scenery compared to bright colorful games like SM64.

The potential solution to the issue is simple motion interpolation. How well it would works in practice for 120Hz refresh remains to be seen.

[tongshadow]

Blur-free 240p120hz should be possible for 60fps sources if you double/repeat each of those frames and output it as 120fps.

[Guspaz]

No, it's in fact the repeating of frames that creates the blur in the first place. In actuality, it's not so much a blur as a double image effect you get from the double strobe. This is the same effect you'll see on an OLED enabling BFI on a 30 FPS game.

[tongshadow]

As long as black frames are not inserted, it should look fine. We're not strobing here, the idea is to match frame rates and refresh rates to avoid the double image effect caused by higher refresh/lower frames.

[Guspaz]

CRTs inherently strobe, that's how they work. BFI is just trying to simulate a CRT strobing. So duplicating each frame in a 240p60 signal to produce 240p120 will look like the "Frame Rate at 1/2 of Hz" row in your diagram there. It will strobe twice per input frame, and thus you'll see double.

[Josh128]

As long as black frames are not inserted, it should look fine. We're not strobing here, the idea is to match frame rates and refresh rates to avoid the double image effect caused by higher refresh/lower frames.

Tong, Gus is correct. This has been long proven, I tried it personally years ago and didnt notice at first, but someone here pointed it out and its a real thing. You get a double image effect on fast moving objects when feeding 60fps content frame doubled to a 120Hz CRT. Looks great on 30fps content and slow moving 60fps content, but motion interpolation would be required to eliminate the double image effect without impacting brightness.

[tongshadow]

As long as black frames are not inserted, it should look fine. We're not strobing here, the idea is to match frame rates and refresh rates to avoid the double image effect caused by higher refresh/lower frames.

Tong, Gus is correct. This has been long proven, I tried it personally years ago and didnt notice at first, but someone here pointed it out and its a real thing. You get a double image effect on fast moving objects when feeding 60fps content frame doubled to a 120Hz CRT. Looks great on 30fps content and slow moving 60fps content, but motion interpolation would be required to eliminate the double image effect without impacting brightness.

I'm aware, because that's how I run CPS1 MAME on my CRT monitor: 384x224@120hz. The double image is noticeable because the games are running at 60fps on a 120hz display.
We're discussing hardware doubling 60fps sources into 120fps, like the OSSC Pro, to match the display.

CRTs inherently strobe, that's how they work. BFI is just trying to simulate a CRT strobing. So duplicating each frame in a 240p60 signal to produce 240p120 will look like the "Frame Rate at 1/2 of Hz" row in your diagram there. It will strobe twice per input frame, and thus you'll see double.

You seem pretty knowledgable on the subject, but has someone already tested this? And by this I mean hardware doubling 60fps sources to 120fps into a CRT Monitor@120hz?
I understand the effects of that on a sample and hold display, but havent found much info on strobed displays, it's a really niche application after all.

[Guspaz]

There is no difference between a game engine outputting the same frame twice and a scaler duplicating the frames. It is literally identical. The display doesn't know or care at what point the frame was duplicated in the pipeline, it only knows what it's being asked to draw in the moment. This is not something where we need to test the difference.

[Josh128]

As long as black frames are not inserted, it should look fine. We're not strobing here, the idea is to match frame rates and refresh rates to avoid the double image effect caused by higher refresh/lower frames.

Tong, Gus is correct. This has been long proven, I tried it personally years ago and didnt notice at first, but someone here pointed it out and its a real thing. You get a double image effect on fast moving objects when feeding 60fps content frame doubled to a 120Hz CRT. Looks great on 30fps content and slow moving 60fps content, but motion interpolation would be required to eliminate the double image effect without impacting brightness.

I'm aware, because that's how I run CPS1 MAME on my CRT monitor: 384x224@120hz. The double image is noticeable because the games are running at 60fps on a 120hz display.
We're discussing hardware doubling 60fps sources into 120fps, like the OSSC Pro, to match the display.

CRTs inherently strobe, that's how they work. BFI is just trying to simulate a CRT strobing. So duplicating each frame in a 240p60 signal to produce 240p120 will look like the "Frame Rate at 1/2 of Hz" row in your diagram there. It will strobe twice per input frame, and thus you'll see double.

You seem pretty knowledgable on the subject, but has someone already tested this? And by this I mean hardware doubling 60fps sources to 120fps into a CRT Monitor@120hz?
I understand the effects of that on a sample and hold display, but havent found much info on strobed displays, it's a really niche application after all.

This "hardware doubling" you are talking about of is actually whats called motion interpolation. Motion interpolation generates an "intermediate/estimated" frame between one real frame and the next and that is the only way to potentially eliminate the double strobe effect, and thats why I am asking about / requesting it. Its been long known as the "soap opera effect" on TVs because it took 24 or 30 fps film content and interpolated it to 60fps, making it look like a soap opera, most of which are filmed at 60 fps.

It wont work well for 30 fps video games due to their inherently low frame rate, not to mention it wouldnt look authentic even if it did. 60fps games should be easier to implement, but the real world quality and artifacting would need to be tested. This depends of course, on the quality of the implementation.

[tongshadow]

Motion interpolation, aka fake frames, has too many disadvantages for gaming applications, but it does help in reducing motion blur on sample and hold displays, preferably if it's at least 120hz. Although, BFI is a superior choice for gaming.

There is no difference between a game engine outputting the same frame twice and a scaler duplicating the frames. It is literally identical. The display doesn't know or care at what point the frame was duplicated in the pipeline, it only knows what it's being asked to draw in the moment. This is not something where we need to test the difference.

Seems that's true indeed, but 240p/120hz + BFI may have benefits elsewhere even if it looks identical to 480p/60hz + scanlines.

[Josh128]

^^
Motion interpolation on 240p120 retro gaming content has never been tested, so to say it has too many disadvantages is not really true until then. Its currently being used by AMD (FSR 3/ Fluid Motion Frames) and Nvidia ( DLSS 3) to eliminate stutters and drops below 60fps and overall the consensus is that it works quite well for the most part.

BFI on 120Hz does have benefits-- namely, to be used on LCD screens to greatly reduce blur. Potential benefits to OLEDs as well-- all sample and hold displays capable of 120Hz could benefit from it.

[Guspaz]

It should be cautioned that FSR3 and DLSS framegen are quite different than the motion interpolation done by TVs. Those algorithms work with a significant amount of supplemental information from the game engine that allow them to better predict the intermediate frames. This includes things like per-pixel motion vectors (the game engine knows the location and movement of every object anyway), the depth buffer, and an exclusion mask (to handle things like crosshairs and HUD elements). These basically serve as hints for the GPU's hardware optical flow engine, since things like motion vectors don't cover everything (like particles or reflections). All your TV has to work with is the current and previous one or more video frames, it's missing all those hints.

[fernan1234]

Put simply, motion interpolation of retro games will look like crap full of artifacts. Stick with BFI and pray that good displays can some day get enough full screen brightness capacity to make up for it, CRT masks effects, etc (for OLEDs it will probably never happen).