Analysis Of Scaling To 16:9 Aspect Ratio

[Epicenter]

I did a bit of playing around with screenshots in true resolution from a few consoles, and how they might look placed onto the XGP's originally-planned 16:9 480X272 display. The common belief is, scale the image out all the way to 16:9 from its original 4:3 design and you get a disgusting stretched image, or keep it in the center and have to squint to see it in a sea of emptiness. Well, as it turns out, things look pretty damn good if you just scale the image up, maintaining 4:3 aspect ratio, until the image hits the top and bottom boundaries of the screen. This results in some horizontal letterboxing on the sides, but the vast majority of the screen is still used-- and the end result is an image that's still the same size or larger than how the image would appear on the GP2X.

Take a look at a few example scenarios. Images use bilinear filtering as they likely would scaled by the XGP's GPU, framed in a 480x272 16:9 image .. naturally, blackness indicates unused screen area.

A 320x224 MegaDrive/Genesis frame centered with no scaling on the XGP display. (it's not exactly the same shot because I saved over the original. Whoops. )

A 320x224 MegaDrive/Genesis frame centered, scaled out to the full 16:9. Ugly, isn't it?

A 320x224 MegaDrive/Genesis frame scaled, maintaining 4:3 aspect ratio.

A 256x224 NES frame centered with no scaling on the XGP display.

A 256x224 NES frame centered, scaled out to the full 16:9. Ugly, isn't it?

A 256x224 NES frame scaled, maintaining 4:3 aspect ratio.

Don't get me wrong, I'd still prefer a 4:3 display, by far. But I'm not that worried about GP using a widescreen display anymore..

 

[Vimacs]

the xgp wont be able to do such good filtered scaleing, try it with a normal pixel resize algo and you see the uglynes.

 

[Epicenter]

If you scale an image, you'd do it with the GPU, not with a mediocre dedicated scaler like the one used in the GP2X. That is a standard supported feature and not a very taxing operation-- no good reason it wouldn't be doable in realtime. If I'm not mistaken, the PSP's GPU is what emulator authors use to do their scaling (except all the way out to 16:9, which looks godawful.)

 

[Vimacs]

The gpu will do bilinear filtering at best, yours look like lancos wich is a very cpu intensive filter.

 

[Epicenter]

I used Bicubic, actually. I tried Bilinear just now and it looks a little smoother than this.

 

[DaveC]

I did a bit of playing around with screenshots in true resolution from a few consoles, and how they might look placed onto the XGP's originally-planned 16:9 480X272 display. The common belief is, scale the image out all the way to 16:9 from its original 4:3 design and you get a disgusting stretched image, or keep it in the center and have to squint to see it in a sea of emptiness. Well, as it turns out, things look pretty damn good if you just scale the image up, maintaining 4:3 aspect ratio, until the image hits the top and bottom boundaries of the screen. Don't get me wrong, I'd still prefer a 4:3 display, by far. But I'm not that worried about GP using a widescreen display anymore..

I looks pretty damn blurry to me.

Look at the text " VITALITY". Nice and sharp. Now look at it in your stretched and filtered screen, it is a blurry mess. I zoomed in to show it better. Then there is the fact that after all of that stretching and blurring, when you stretch to fill top to bottom and maintain the aspect the image is STILL just about the same size as the GP2X 3.5" screen at 1:1 only is looks fuzzy.

I don't know why you are on this crusade to prove to everyone, and yourself, that this XGP is a gift from god

I dare you to do a stretched and blurred image of the very first image.

You will see that all of the detail like in the tree graphics in the background will blurr together loosing detail.

 

[shinneri]

That's pretty misleading Dave.

For whatever reason, Epicenter saved the first, non-scaled image as a PNG and the scaled images as JPG. As you probably know, JPG is a disgusting format and the artifacting caused by it makes the scaled shots look worse than they actually are. At the same time the unscaled PNG looks really good.

I see your point, but please adjust your image so that you are not misleading everyone thanks to the crappiness of the JPG format.

 

[Alex.]

The nice thing about retro games' artwork is their detail, every pixel being handcrafted. When you stretch and filter it, you take all of that away and only leave a shell of what the marvel was. Either stretch proportionally and have 1 pixel being 2x2, or deal with the black borders.

- Alex

 

[Epicenter]

Good point, I'll switch them all to PNG. I kept JPEG for some since the PNG files were quite large, due to the increased number of colors brought into play due to the filtering. DaveC-- it looks nowhere near as bad as you make it out to be. Sure, it's not perfect and I agree, emulators should be pixel-perfect if they're to look their best. But emulators are just a simulation of the real thing, and this isn't a showstopping issue for that reason. No one is expecting an emulator to be even close to 'perfect' because by its very nature, it can't be.

Also, anyone who used their classic systems with a TV and RF, Composite, or S-Video connection sees a combined, compressed version of the image, with dot crawl, color inaccuracy and other artifacts that are not present on the true digital frame. Most people don't SEE a perfect image playing the real system, and many games took advantage of it. For example, Ecco the Dolphin 1/2 for MegaDrive/Genesis. I interviewed one of the producers of both games, and he stated they used such blurring effects that NTSC/PAL TVs offered as a FEATURE to blend careful dithering patterns on the sprites together, forming the illusion of hundreds if not thousands of sprites. In a digital frame that is perfect in every way, like on the GP2X's 320x240 screen with no scaling, the image looks pixellated and tacky compared to on an analog TV with composite video!

Another important and overlooked fact is that the NES, SNES and Turbografx-16 SKEWED THEIR IMAGES TO FIT A 4:3 SCREEN when connected to a TV! They had a means of filtering applied as well as the image was stretched by analog means, very similar to how it's done here by digital ones.

If I may quote the aforementioned Ecco producer ..

Never can do a same quality art on a VGA display. I mean the composite video is so different, like an antialiasing module.

That's what I've always called it actually .. 'Poor Man's Antialiasing'. A very large number of games look much better for it, especially on the MD/Genesis with its limited color palette display capabilities.

For evidence ... PNG-24 examples of Ecco 2 at full 16:9 stretched, and with the 4:3 aspect ratio maintained. Using bilinear filtering to resample the image, similar or identical to how the GPU would do it:

16:9 Fully Scaled

4:3 Aspect Ratio Maintained

And here's an example of how the original digital frame looks compared to an image of the same game from a PAL television. Note the illusion of FAR more colors, especially on the color gradients on the Ecco sprite and the rocks:

At any rate, why do you even come into this forum if you think the concept of any sort of XGP machine preposterous? You never make a postive comment ever, and it seems that if it were up to you, in 5 years we should still be using the GP2X, because newer technology is NEVER appropriate, especially when it takes away from the GP2X's development/user base (which it WON'T because NEARLY ALL THE SOFTWARE WILL BE COMPATIBLE, for god's sake.)

p.s. My apologies that this post is in no way 56K friendly.

 

[Captainbubby]

People still use 56k!?

The widescreen certainly doesn't look that bad and I'm sure the emus would have the option to put it back to 4:3 scale

 

[DaveC]

Sure, it's not perfect and I agree, emulators should be pixel-perfect if they're to look their best. But emulators are just a simulation of the real thing, and this isn't a showstopping issue for that reason. No one is expecting an emulator to be even close to 'perfect' because by its very nature, it can't be.

Another important and overlooked fact is that the NES, SNES and Turbografx-16 SKEWED THEIR IMAGES TO FIT A 4:3 SCREEN when connected to a TV! They had a means of filtering applied as well as the image was stretched by analog means, very similar to how it's done here by digital ones.

At any rate, why do you even come into this forum if you think the concept of any sort of XGP machine preposterous? You never make a postive comment ever, and it seems that if it were up to you, in 5 years we should still be using the GP2X, because newer technology is NEVER appropriate, especially when it takes away from the GP2X's development/user base (which it WON'T because NEARLY ALL THE SOFTWARE WILL BE COMPATIBLE, for god's sake.)

So because emulators are not perfect you should have one more thing that is not perfect to add to the list which is blurred out fractional scaling? I don't buy that sorry, bad argument.

Skewed images to fit a TV? Well it wasn't done with crappy fractional scaling and filtering on a fixed pixel screen like a handheld, totally different. All they did was play with the scan rate a bit which produced pixels that were slightly rectangular. The pixels were still even (no 1.5X bullshit) just a bit elongated. Since I had my systems connected with RGB cables that I hacked together I didn't get all of the fuzzing and dot crawl of RF or composite either. 1:1 looks closest to that.

No I never said we should be using the GP2X in five years. Newer technology should give an advantage though or it is not worth switching to. If the screen is blurrier than what I already have there is no point. Then you are doing it just to do it. And yes the XGP will fragment the development /user base, it will happen you will see. I am sure you will like that as you have become a rabid XGP fanboi as of late for some reason. You have even convinced yourself that stretched blurred out graphics are a good thing. If the XGP was going to come out and be monochrome you would be trying to convince us why monochrome is best. To you whatever the XGp will be, will be the greatest and we should all flock to it and abandon the GP2X as fast as possible.

 

[Epicenter]

You only hear what you want to, and you only hear extremes.

#1, I do not hate the GP2X and I don't want people to 'leave' it, that's nonsense. I'd love the GP2X if the launch weren't a disaster, and I could overlook that RIGHT NOW if they'd fix the stick, which threatens to make an otherwise very good system useless for a lot of people, myself included. The stick drives me fucking nuts-- it's like getting the most powerful game console known to man but having to control it by squeezing a crazed squirrel in a certain pattern. And honestly, that's how the stick feels a lot of the time. Like controlling the GP2X with a squirrel.

I'm developing a COMMERCIAL game. I want to be able to give it to people one they buy a GP2X, and enjoy it. When a person picks up a GP2X, and they're struck with problems that require lots of tweaking and adjustment out of the box that should not have been necessary, it hurts the integrity of my work, not just the GP2X. When they encounter flickery screens, headphone jack problems, batteries that last seemingly random periods of time (usually much too little), use Alkaline batteries and get garbage performance because no one told them they had to run out and find special high-amperage batteries the average joe doesn't even know about, and a charger, when they use a stick so poorly designed it feels broken, when they hear parts rattling around inside their unit, that influences the reputability and credibility of my work. We have a publisher who's interested in Stargazer yet upon purchasing an MK2 unit, they thought the device was broken because it he put in Alkaline batteries like a normal user would, the machine crashed moments later with a spectacular explosion of random colors and lines onscreen as though the LCD panel was broken (this is ridiculous behavior for any handheld when batteries are low), after which he couldn't use the machine until he bought an AC adapter. He had no way of knowing about the need for special High-Amperage batteries. The flimsy and hollow feel of the machine and its awkward design faults irritated him, and he thought the machine was more like a child's toy than a real game system. After a significant time investment how do you think that makes a developer feel, to learn the vessel for his works is percieved as worthless?

I look forward to the XGP series because they look like more well-thought-out machines. The build quality and design of the GP32 was far more solid and stable, the launch rollout wasn't such a disaster, and there were nowhere near as many hardware faults, so I have reason to believe GP will produce another fine product in the future.

#2, Because I've discovered there are means by which the system can look good with a 16:9 display doesn't mean I want it to have one. I hope they switch to something 4:3, I've made myself quite vocal about that here and to GP, as have many others. Were the XGP going to be monochrome, I'd say to hell with it. Did you see me lauding the GPKiDS with its mediocre hardware selection and godawful resolution? No, I spoke ill of it quite a bit, if you were paying attention then. If not, there's this forum's post history to prove that. I said the machine shouldn't even EXIST, it was that bad an idea.

Criticize my views all you like, I encourage that. But don't make up exaggerated outlandish claims. It makes rational discussion impossible whne you're having your own words put back in your mouth, inflated a hundred times to the point of distorting what you even meant beyond comprehension. That doesn't benefit anyone.

The Gp32 and GP2x are 'fragmented' perhaps in the sense that it's not always simple to port software. Porting a program between 2 ARM9-Linux devices is a breeze, a child could you do it reading instructions off a flash card. In fact, you may not even need to recompile a lot of programs. You need to stop percieving every new idea or system as a threat. This isn't the first time you've made such a fuss; you did the same over the Gizmondo and there's not many forumgoers who don't remember it, either.

 

[shinneri]

Build quality was iffy on the GP32 as well. The GP2X launch was terrible, and I agree there were FAR too many issues. But it might be premature to compliment Gamepark in that regard.

Often times the shoulder buttons on the GP32 simply wouldn't work. When batteries died, your screen did the same crazy the GP2X does (but GP32 had no indicator light). Dust under the screen was a major issue. They stupidly reversed the stereo speakers on some of the units, and the microswitch control stick had high tendencies to break or become unresponsive.

 

[JaqMs]

Build quality was iffy on the GP32 as well. The GP2X launch was terrible, and I agree there were FAR too many issues. But it might be premature to compliment Gamepark in that regard.

Often times the shoulder buttons on the GP32 simply wouldn't work. When batteries died, your screen did the same crazy the GP2X does (but GP32 had no indicator light). Dust under the screen was a major issue. They stupidly reversed the stereo speakers on some of the units, and the microswitch control stick had high tendencies to break or become unresponsive.

Well, those control problems were easily fixable with non-technical methods (paper mod). I know those fixes shouldn't have been necessary, but at least it didn't come "broken" out of the box and there's no need to take a risk making a d-pad mod.

*5* years ago, Gamepark was smart enough to make sure their joystick and alkaline batteries worked. And most importantly, THEY WERE SMART ENOUGH TO PUT THE HEADPHONE JACK ON THE BOTTOM OF THE UNIT!!! To me, that is one of the GREATEST flaws of the GP2X that just shows pure incompetence. Someone answer this: -why-, I repeat, -why- would someone place the headphone jack on the top? What is the f*cking reason?!

Ok I'm done. Sorry for the rant.

 

[DaveC]

You only hear what you want to, and you only hear extremes.

#1, I do not hate the GP2X and I don't want people to 'leave' it, that's nonsense. I'd love the GP2X if the launch weren't a disaster, and I could overlook that RIGHT NOW if they'd fix the stick, which threatens to make an otherwise very good system useless for a lot of people, myself included. The stick drives me fucking nuts--

I look forward to the XGP series because they look like more well-thought-out machines. The build quality and design of the GP32 was far more solid and stable, the launch rollout wasn't such a disaster, and there were nowhere near as many hardware faults, so I have reason to believe GP will produce another fine product in the future.

#2, Because I've discovered there are means by which the system can look good with a 16:9 display doesn't mean I want it to have one.

You certainly talk as if you hate it. You slam it every chance you get and end your posts with how great you think the XGP will be. As far as "leaving" it you have left it yourself. You have said how the d-pad mod hasn't worked out so you will just toss it aside and wait for XGP and exclusively develop for that. Why even have that banner in your sig. Time to change it to say "comming soon to XGP"

The GP32 had a horrible launch. It was only available "officially" in Korea and had cancelled European launches repeatedly. Gamepark couldn't get out of their own way. The shoulder buttons failed constantly and the stick couldn't do diagonals eventually without sticking a wad of paper in it. Maybe they have changed but if you base it on the GP32 they wouldn't be as great as you think.

You have discovered that you love the XGP and have convinced yourself that stretching looks good. It doesn't look good. It looks worse than 1:1 not better. If the GP2X had a 16:9 screen you would list that as one of the faults.

 

[Epicenter]

As far as "leaving" it you have left it yourself. You have said how the d-pad mod hasn't worked out so you will just toss it aside and wait for XGP and exclusively develop for that.

BULLSHIT. Don't put words in my mouth. The d-pad mod worked out poorly, so I'm taking a different route and combining my GP2X board/LCD/speakers with a game controller so that it is functional and portable, Still. it's easier than working inside the obnoxiously constrictive and poorly-designed case.

As for 'coding exclusively' for the XGP? That's nonsense. I've said it before-- I intend for all software I write to be compatible with the GP2X, XGP and XGP Mini. The only machine not supported will be the GPKiDS.

You have discovered that you love the XGP and have convinced yourself that stretching looks good. It doesn't look good. It looks worse than 1:1 not better. If the GP2X had a 16:9 screen you would list that as one of the faults.

I haven't convinced myself of anything that isn't true, but again you've misconstrued my words. I don't think a stretched image looks better than a 1:1 one. But I don't think it looks that bad at all. I think stretching all the way out to 16:9 looks like shit, and yes, I would list 16:9 as one of the GP2X's faults if it had such a display, just as I'd list it as one of the XGP's if I had to make an objective "Pros/Cons" list about either.

The GP32 has some hardware faults, too. But they're not as severe or show-stopping as the GP2X's, and holding one in your hands, it feels more solid, like a 'real product', not like something that was cooked up in a garage, which is honestly how the GP2X feels.

The GP32 had a horrible launch. It was only available "officially" in Korea and had cancelled European launches repeatedly. Gamepark couldn't get out of their own way. The shoulder buttons failed constantly and the stick couldn't do diagonals eventually without sticking a wad of paper in it.

I don't care about the success of their marketing and distribution scheme-- I care about them not shipping broken hardware. Craig's told me personally that GPH produced the first batch of units from preorder money, and knew the faults they had when they were shipped out. He told me that the prototypes had faults and they couldn't afford to fix them, so they just left the problems there. If you can't afford to produce a machine that you know isn't faulty, you don't rush it to market on borrowed money, you find venture capital and wait until you can produce something of quality. Even now, GPH's unwillingness to fix issues like the stick are claimed to be because of lack of funds. Then where is the R+D coming from to make BoBs and pretty little docking stations and a new line of weirdly-colored units? Where's $5,000 coming from for the contest? This is piss-poor business management.

Sure, GamePark's business plan was not solid at first, they wanted to compete as a real console. Now they have no delusions of this, they want to compete in the niche market or form their own niches; they don't want to compete with Sony or Nintendo's handhelds. They now know the appeal of their product outside of Korea, and they now support open-source and homebrew development/emulation to the best of their ability, not focus on commercial and signed software. GP's original plan is far different from how it was in the earlier days, and their openness to community feedback on hardware and software design, as well as their quick and helpful open communication is very confidence-inspiring. GPH was originally painted by many as the 'heroic open source hero' breaking away from the 'evil commercial GamePark who hates homebrew'. Things are looking quite different now-- GPH is striking me more as the pissy little company who thought they could make more money on their own, so they broke off from GP, made a more powerful but inferiorly-built product using someone ELSE's plans (MagicEyes'), and doesn't understand how to support it.

This is not even to mention my experiences from before the GP2X was even available for purchase where Craig outwardly lied to and decieved me and the rest of our team regarding commercial game development through them. Their claims about Quake's operation, the GP2X's internal hardware and its capabilities, distribution means, content protection schemes, packaging variety, it was all a crock of lies. That's all I expect from Craig and GPH, now. EvilDragon seems like the only representative of GPH who isn't a lying con-artist.

Those facts nonwithstanding, I do enjoy my GP2X, but I should not have to pour in the time and effort I have to get it to feel 'usable' day-to-day. I'll still keep software compatible for it because not everyone has the same critical nature I do. With how critical YOU are, DaveC, I wonder why you don't hold the GP2X in a lower regard than you do. Instead, you act like GPH's designated GP2X Holy Warrior.

 

[DaveC]

... With how critical YOU are, DaveC, I wonder why you don't hold the GP2X in a lower regard than you do. Instead, you act like GPH's designated GP2X Holy Warrior.

I like it for what it is and does for me. For all of the faults it is still all around the best form factor out for emulation, which is what I mostly use it for. The button layout works very well for me too. The screen aspect and resolution doesn't force that awful stretching to get a decent image size and it doesn't tie you to expensive and dead-end custom batteries. Those two things alone make the XGP very unattractive to me. I don't care if SNES is perfect on it when it is a blurry mess. I could have stretched out but perfect SNES right now on my PSP. 16:9 ruins it. The GP2X stick? yeah it sucked but I don't have that problem anymore with my mod.

The perfect handheld for me would basically be a GP2X with a d-pad and a beefier CPU. If the XGP was that then I would be more interested. As far as "cheap feel" That really doesn't bother me and it is not THAT bad. The M2 is better in that respect. If I had to choose between stretched images or a case that was a little "light" well I don't even have to say it.

 

[Epicenter]

The MK2 is plenty 'light' and hollow, it doesn't have any extra mass to it, the area between the battery compartment and PCB is empty, and even a bit of aluminum EMI shielding would add the requisite wight and strength to feel more 'real'.

D-Pad-- sure, it doesn't bother you because you remedied it, but otherwise you'd likely still be bitter about it. But that doesn't solve the problem for anyone who buys a game I work on and says it's broken because the stick sucks. Then it becomes my problem whether my stick is replaced or not.

The reason I even express interest in the GP2X is because it's a great little idea for a machine, but it has a LOT of faults. It has enough redeeming value to keep me interested. It's just that its best features are brought about by the chipset, GPH did very little to do to actually make it into a game console. And all the parts they had to do, they jacked up severely.

They had to engineer controls. They used a stick that is miserable for games, and buttons with membranes that aren't quite springy enough, or of variable springiness. One shoulder button on my unit pops up quickly and one is mushy. And on all non-MK2 units the stick was MOUNTED THE WRONG WAY. I mean, come on.

They had to design a case. It has a cheap, plasticky feel like a bootleg toy or a POPstation. The rubber covers for the jacks fall off easily and don't fit the holes they cover well without lots of force and wiggling, and the jacks are placed in illogical places that make it hard to play with AC power connected or USB attached, and the headphone jack location makes no sense. Then they mounted the board misaligned so the EXT connector not connect properly and the stick slam into the case on the left.

They had to mount a headphone jack, they picked a faulty one with an issue of shorting out its L/R channels and making them go silent, and the factory didn't solder most of them correctly at all (nearly 0 solder) so they would fall off.

They had to pick a screen, they picked one that was INTERLACED at 60 Hz. That's just nonsense. Then they screwed down the case so tight it was interfered with, and made even worse by an illogical pad they put on top of the MMSP2. The LCD was also incompatible with the MMSP2's LCD controller or it may have worked fine interlaced .... they also picked the wrong value resistor for LCD backlight power causing awful flickering and brightness fluctuation.

They had to add TV-Out and Scaling, so they used parts made for a home device with AC power, so they cut battery life in half, or worse, to a third.

They had to design a power subsystem, they built an illogical and inefficient one that wastes battery power and can't switch from battery to AC-brick power without crashing the machine, and can't charge batteries.

They had to add speakers and USB client capability, this worked acceptably except that they connected the audio channels BACKWARDS and they had to be switched in software, IIRC.

They had to add LEDs for Power and Low Battery support. Many had faulty LEDs that didn't light properly, or recieved too little voltage.

They had to add a 2xAA compartment. They used NO springs and overly-malleable, easily-breakable metal tabs for both sides, unlike every device for over 20 years.

I'm sure even you'd agree, that is ridiculous. I'm not being overly critical to want a handheld with the GP2X's strong points and none of its insane problems.

 

[DaveC]

I'm sure even you'd agree, that is ridiculous. I'm not being overly critical to want a handheld with the GP2X's strong points and none of its insane problems.

Many of those things were fixed in the M2 version. I think they should have added a d-pad but they didn't.

As far as cheap and plasticy, well it is made of plastic and it is cheap. It has to be cheap or everyone would bitch that it costs too much and wouldn't buy it. GPH can't afford to lose money on every unit they sell like Sony or Nintendo. They have to make some money on the hardware as they don't sell games. So yes there has to be some cut corners. You can't have everything. If you want uber build quality buy a PSP.

 

[Epicenter]

The vast majority of the system's problems were not fixed in the MK2. Only the capacitors that fall off, the headphone jacks that fall off, the screen and the ROTATION of the joystick were corrected. The other problems remain.

Adding a little aluminum shielding is in no way expensive and in fact probably required to be approved by the FCC to not emit 'potentially harmful' interference. Screwing in the LCD and some other components instead of wedging them in there with flimsy tabs would add a bit of appreciated weight, and there's still loads of empty space that amounts to bad design, not 'low cost'. You don't add empty space to reduce costs.

Furthermore, decent build quality can be had at low costs. There are lots of GOOD factories out there, and decent plastics don't cost that much either.

Nintendo makes a profit on their hardware. At any rate, the GP2X's hardware is not that expensive, and a DS probably costs about as much to make with its weaker CPU and beefier graphics subsystem, and 2nd display. GPH isn't making them available as a bargain.

And a PSP is not even remotely similar to the GP2X and HAS VERY LOW BUILD QUALITY, but it does not even approach the GP2X in that regard. You know that yourself. How likely do you think it is I'll switch development to a platform with a different CPU architecture that doesn't even run SDL applications correctly yet?