It's the details that take longest (Part 3)

[rygD]

Not touchpads. Not touchscreens. And for the love of all that is holy, no trackballs!!!

I don't think any of them would be good for games.  I know I wasn't very seriously considering any of them, but I still wanted to consider potential positives and negatives of each.  I think I would be pretty unhappy if we got touchpads or touchscreens instead of nubs.  The fun thing about the touchscreens is that people have asked for them, and if a Pyra is mistaken for something in the DS family you can show it to the person and they will see 3(!) screens, plus a keyboard, or at least 2 screens if the keys between nubs were scrapped, which would cause even more confusion about this strange DS.  I wouldn't mind a trackball, but that is probably not a good choice for a gaming handheld.

The most awesome thing would be if the nubs unplug, and then you can put whatever you like there, touchpad, trackball, or nub. But that won't happen, and most games use nubs.

Many games would work fine with a trackball, but many probably would not.  It really comes down to personal preference, and nubs probably resemble the controls most people typically use.

Just out of curiosity, does anyone know if the nubs that were going to be used or the 3DS nubs would be more functional than the other in, say, 20 years?  Do they both have mechanical parts that would likely fail with heavy use? 

 

[dimag0g]

I could be wrong, but I recall something about the red thinkpad nib being patented or some such.

It's this sort of legal concern which prevents a different sort of hinge or a flip-clam/tablet.

In that case, let's settle for the next best thing - the Dell blue thingy!

 

[DrHAX]

When it comes to the 3DS nub, I had one issue with it that I surprisingly didn't have with the pandora nubs (I thought it was just a general issue with nubs until I used the pandora ones). The material it's done in makes it very easy for the thumb to slide off the nub after some minor usage of it (just a few minutes).


Apart from that, I didn't have any issues with its nub - but this issue makes it barely usable for me.

Same problem here - but we will have to create our own disc anyways, so it's not the 3DS top.

You put the Pandora disc the 3DS nub I'd be a happy person ( would even probably buy a disc and put it on my 3DS)

 

[CFWhitman]

Yes, I think the red color of the Trackpoint nub is trademarked (as is the Trackpoint name).

A pointing stick like a Trackpoint is the most like a joystick or nub of any of the common laptop pointing devices. If we had to use a laptop pointing device, it would probably be the best compromise.

I love my Logitich Marble Mouse trackball (don't ask me why they use the term "mouse"). However, there are two issues with trackballs on handhelds. One is that, though you can somewhat emulate the qualities of a mouse with an analog joystick/nub/pointing stick you can't effectively emulate the qualities of a joystick with a mouse or trackball. The other is that small trackballs are not nearly as good as large ones, so trackballs are not nearly as practical on a handheld device (though thumb trackballs do need to be small to avoid repetitive stress injury). A trackball might be OK to have, but it wouldn't replace the nubs.

I think I could live without the click buttons on the nubs, though if having them doesn't hurt anything, extra buttons are almost always nice. If the price of significantly better nubs is to forgo the buttons, then I'm fine with leaving them off. I would consider all options, though. I'm not sure how much travel is ideal for a nub, so I wouldn't assume a nub with greater travel is better without testing both options if I could.

 

[Glyph Reader]

I really like the 3DS's (the 2DS's in my case) nub, and I've been using it with Super Smash Bros. recently.  The feel is quite good, but I have only one issue with it --- the disc part (as in the part that your thumb touches) gets this kind of stickiness to it.  I haven't seen any complaints around that, so I might just be unlucky/super sensitive. 

-Glyph Reader

 

[Kiga]

I've used the 3DS nubs for a long time (got my 3DS on release day) and they are by far the best analog nubs I've used on any device (Pandora, PSP, A couple of BT Game controllers). Can anyone link to an actual analog nub that is widely reported as being better than the 3DS one?

There isn't any nub with that size which is mainstream beside 3DS.

As has been mentioned these nubs are not coming from Nintendo but from the company that supplies Nintendo, so there would be no need to rely on Nintendo.

I mean relying on Nintendo choice. I could have understood if it was build by Nintendo but it's just a model that Nintendo chose 4 years ago. I don't think that Nintendo would chose that at least 4 years old nub now.

It should serve as point of comparison to compare the technology used on another ones and determinate if it is enough sturdy because 3DS one have Nintendo credit and it is  proven that it works, but I don't see why not chose another nub since nubs are easy to compare and determine weak points.

 

[Hồng Thất Công]

But Nintendo DS nubs are for gaming, not sure if it's good as a mouse though.  Trackballs are good for mouse but not for gaming, and Nintendo DS nubs are good for gaming but maybe not good as mouse.  So we can use both of them on the Pyra - one of each

 

[Grench]

There are five ways to do mouse-like movement.

- direct absolute positioning: this is what a touch screen does: instant (x,y) position input. Great for tapping buttons, but no notion of relative movement (so not usable for e.g. an FPS game).

- physically implemented indirect absolute movement: this is what a traditional mouse does: it's a good and intuitive method (but no way you can integrate it into a handheld, you'll have to use an external usb mouse). You're controlling (dx, dy), not (x,y), but in doing so, you also move the device itself, so there's still a mental mapping between the absolute position of the mouse and the absolute position of the cursor, as long as you don't reach the edge of your 'mouse mat'.

- virtual indirect absolute movement: this is what a trackpad does: I don't like it as much as a traditional mouse because you're basically using your finger as a virtual mouse on a very small mouse mat. The advantage of modern real mouses is that they don't need a mouse mat so the area you can move in can be as large as you want. The mental position mapping suffers from such a small 'mouse mat'.

- reversed physically implemented indirect absolute movement: this is what a trackball does: basically it's an upside down mouse, where instead of having a moving mouse on a static surface, the mouse stays where it is and the surface moves (i.e. your thumb). This means that you have to mentally convert 2D movement speeds to ball spinning. To me, this is rather unintuitive.

- relative movement: this is what a joystick/nub does: you're controlling (dx, dy) absolutely, that is, holding the nub in a particular position means (dx, dy) is constant. This is quite intuitive and allows for a good mental mapping between mouse movement (not mouse position) and nub position.

I think it makes a lot of sense to have a touch screen and a nub, with the option to plug in other input devices. Touch screen and nub are nicely complementary because they're at two extremes of the spectrum.

Nice write up, though I think your classifications are too strict.

You classified the track pad as 'virtual indirect absolute movement' only. However, there is no reason that a track pad can not be mapped as relative positioning & movement. Put a virtual paint dot at the exact center of a circular track pad. Declare that as 0,0. Make 0,0 the default output from the device instead of null. Any touch outside of 0,0 declares a point on a grid - exactly as a joystick or nub will. Dragging your thumb slowly from center to right would give the exact same outputs as dragging a nub or joystick center to right. Pick up your thumb and it goes to a default 0,0. Just like a joystick or nub.

Now the hard part. Since most track pads are flat, understanding the relative movement and positioning is difficult for the human brain to keep track of. It needs some form of feedback to know where it's at on the track pad with relation to 0,0. That is where my idea of making the track pad a parabolic dish comes in. 0,0 is the center, lowest point in the dish. As your thumb 'climbs' or touches in any direction out of 0,0, that relative input is recorded AND the user knows which direction they're going AND by how much - or at least as accurately as they would with a nub.

The bad - it would take some getting used to. It would have to be engineered from scratch.

The good - no moving parts. No gaps to let impurities in (lint). Can be as precise or more than a nub.

The fact that this doesn't exist 'off the shelf' is a crusher though.

 

[Yoyobuae]

- reversed physically implemented indirect absolute movement: this is what a trackball does: basically it's an upside down mouse, where instead of having a moving mouse on a static surface, the mouse stays where it is and the surface moves (i.e. your thumb). This means that you have to mentally convert 2D movement speeds to ball spinning. To me, this is rather unintuitive.

My first reaction to this statement was to turn my (optical) mouse upside down and try to use it as a trackball, unsurprisingly doesn't work well. :rolleyes: Works a bit better if a piece of paper is placed over the sensor though. ^_^

On a more serious note, I hope whatever the new nub is, it will at least be more sensitive in the near-zero region. Pandora nubs tend to "jump" from zero to half of max input on either X or Y axis, makes it difficult to do precise/small input, like when trying to make a character walk thru thin pathways.

@Grench: Are you talking about capacitive touch input? Won't that require special calibration and controller modules? Specially since it's a custom design. I imagine something like that would take a big budget.

 

[Silent-Hunter]

On a more serious note, I hope whatever the new nub is, it will at least be more sensitive in the near-zero region. Pandora nubs tend to "jump" from zero to half of max input on either X or Y axis, makes it difficult to do precise/small input, like when trying to make a character walk thru thin pathways.

I know what you mean about that. I hate that too.

 

[Grench]

On a more serious note, I hope whatever the new nub is, it will at least be more sensitive in the near-zero region. Pandora nubs tend to "jump" from zero to half of max input on either X or Y axis, makes it difficult to do precise/small input, like when trying to make a character walk thru thin pathways.

I know what you mean about that. I hate that too.

That -could- actually be made to be software adjustable.

[Rambling]

Using the Pandora Input Tester here...

The nubs have a 0,0 (Top Dead Center). Move left all the way and it is -32,0 right is 32,0. Straight up from TDC is 0,-32, down is 0,32.

An issue point is that down and right at 45* is 22,23 and NOT 32,32.

So the nub maps to a circle and NOT to a whole quadrant plane (as per a joystick would). So it will go straight down or straight left and get a -stronger-, i.e. faster movement signal than it can possibly get when pointed down and left (-22,23).

If this were an engineering project I'd say we just need to remap to radians. OS HID doesn't speak radians.

My solution would be to create a set of linear mapping equations to ride between the nub output and the OS. This would keep 32,0 as 32,0 but interpret 22.5,22.5 as 32,32 to give linear X,Y response even though it's on a circular nub. More or less map it as a toroid section instead of a circle. But then, I've never built a nub driver in my life, so that may not be possible.

To make it more sensitive in close and coarser (faster) towards the edge, make the above mapping non-linear. It could be 1:1 for the first 10 ticks, then ramp out to 5:1 by the time you hit the outer edge to yield a range of 0-150 instead of 0-32. I'm not sure if the nubs are capable of decimal output - good joysticks have insane precision in what they output, way beyond their mechanical limits. I feel that the Pandora's nubs put out far too coarse of measure - well under their mechanical limits.

[/Rambling]

 

[TrashyMG]

We are not getting Pandora style nubs as that was a custom item and it's out of production. Why is there a concern here when dealing with the Pyra? 

 

[Yoyobuae]

I just mentioned it because I want new nubs to be an improvement on this particular respect.

For example as I carefully move Pandora's nub slowly to the right on Input Tester I get this:

0,0

0,0

0,0

0,0

0,0

10,0

15,0

14,0

17,0

19,0

20,0

There's a big jump from zero to 10 and no response for like the initial 40% displacement from the center. Some dead-zone is to be expected but it would be better if it wasn't so big. >_<

 

[_wb_]

You can test Pandora nub dead zones by playing Pandora NubNub and pressing START to pause the game. That gives you a pause screen which shows the raw nub positions in a way that lets you find any dead zone (because it shows not just the current position but also all previous positions).

@Grench: I think it makes sense that nub positions map to a disk shape. Why do you want the max speed to depend on the direction? I think it's a feature, not a bug, that diagonal movement is not faster than horizontal or vertical movement. For a mouse that makes sense, and also for most game situations that's how you would want it to be.

 

[notaz]

@Grench: I think it makes sense that nub positions map to a disk shape. Why do you want the max speed to depend on the direction? I think it's a feature, not a bug, that diagonal movement is not faster than horizontal or vertical movement. For a mouse that makes sense, and also for most game situations that's how you would want it to be.

This is a problem with emulation though, things like PSX analogs map to square. I tried to do some remapping but it still works really bad for PSX.

 

[_wb_]

Multiplying both x and y by (sqrt((x^2+y^2)/max(x^2,y^2))) should do the trick to map a unit circle to a [-1,1] by [-1,1] square.

For the other way around, see http://mathproofs.blogspot.be/2005/07/mapping-square-to-circle.html

 

[Grench]

@Grench: I think it makes sense that nub positions map to a disk shape. Why do you want the max speed to depend on the direction? I think it's a feature, not a bug, that diagonal movement is not faster than horizontal or vertical movement. For a mouse that makes sense, and also for most game situations that's how you would want it to be.

This is a problem with emulation though, things like PSX analogs map to square. I tried to do some remapping but it still works really bad for PSX.

Exactly. Joysticks yield a square so that ABS(MAX(X)) is the same whether Y is 0 or ABS(MAX(Y)).

I've always found it annoying, not a feature, on the Pandora that I can go fast left or fast down, but it drops to 1/2 speed if I go down and left.

 

[CFWhitman]

Yes, in practice, it's definitely a problem that the nubs map to a circle shape rather than a square one as conventional joysticks do. Of course there are software based approaches to correcting this problem. though they have their downsides. A hardware solution would be square nubs (or at least square nub sockets), but I'm not sure if anything like that exists.

 

[_wb_]

I've always found it annoying, not a feature, on the Pandora that I can go fast left or fast down, but it drops to 1/2 speed if I go down and left.

It has exactly the same speed in any direction. If going right means doing (0,+1), and going down means doing (+1,0), then diagonally down right does (+.707, +.707). All those vectors have the same magnitude: 1. If you make it so that diagonally down right does (+1,+1), then the vector has magnitude sqrt(2), so if it represents a speed it is 1.41 times as fast as horizontal or vertical movement.

Why would you want top speeds that are different at different angles? Basically you're giving diagonal movements a speed bonus (1.41) while horizontal/vertical movements are slowest (1.0) and everything in between is something in between. That is not intuitive and does not correspond to physical reality. Walking north or east is not something that is inherently slower than walking northeast. If the nub movement is aimed at representing physical movement of some (fictional) object, like a player character in a game, a ball or spaceship or whatever, or a mouse cursor, then it makes no sense at all to make top speed dependent on direction like that.

I suspect that in most game consoles where the hardware sends square-mapped nub/joystick coordinates, most of the games would implement a conversion to circle coordinates. So maybe you're experiencing a case where the emulator forgets convert the Pandora nub coordinates to emulated hardware square-mapped coordinates, and then the game is converting coordinates again while they are already circle coordinates, meaning that you end up with the reverse bias (weak diagonals). That's an emulator bug then, but it doesn't mean that circle coordinate behavior (however implemented) is not desired for nub-controlled movement.

For a dpad it's different: when they're used as a discrete input method, you do an integer number of {left,right,down,up} steps. Down should do 1 down, Right should do 1 right, Down Right can't correspond to .707 Down and .707 Right because there's no such thing as fractional keypresses. If the dpad is moving something in a grid-like cell structure, then it has to go to cell (+1,+1).

It's different again if the dpad is used as a continuous input method, where the length of time the button is held down determines the amount of movement that is done (like in a platform game), then I think it should in most cases behave like a circular nub again.

Anyway, how anything should behave depends of course on the actual application or game, and it's sometimes ultimately a matter of taste.

 

[TrashyMG]

When the Pandora nubs are working correctly they go the same speed in any direction, diagonal or otherwise. I think the odd situation that happens is when the auto nub calibration goes wrong and things get wonky.. luckily this won't be an issue in future nubs.