liberte@uiucdcs.UUCP (06/30/84)
#N:uiucdcs:13900011:000:14211 uiucdcs!liberte Jun 29 18:04:00 1984 My article comparing trackballs and mice received a few responses which may be of interest to readers. Generally, people disagreed or conceded a few of my points. Only one person expressed preference for the trackball. There was also confusion about what I meant by the mouse position to cursor mapping. Note that I am not particularly enamored with the trackball. I am just trying to make the best argument I can for it. Most of the following is from sun!jfarrell (indented), with my original text and new responses and other peoples comments interspersed. Thanks for your mail - I will, again, summarize new material if I think it is interesting. ---------------------------------------------------------------------------- First the similarities: 1. They are both pointing devices, as are many input devices. 2. They can both operate with proportional body movement. - a more sensitive trackball and/or larger scale cursor motion can give one roll per screen without the need to spin the ball. That will have to be an obnoxiously sensitive or very large diameter trackball. Figure 1000 points you may wish to cross (my screen has 1024; the one in the next office has 1192; Xerox Altos started it all with 606 by 808...). Figure 3 inches usable benefit from the width of your hand + 1/4 the circumference of the ball to define the maximum displacement. If you can reliably move 1/100th of an inch, you need 10" total displacement => 6" 1/4-circumference => > 7.5" diameter. Alternatively, you get to move your hand (steadily) less than 1/100th". And I claim 1/4 the circumference available is pretty optimistic. You can get around this with a variable-ratio mapping; because the trackball doesn't have any absolute positioning to start with, the incremental loss is less than for a mouse. The variable-ratio mapping you mention in the last sentence is what I had in mind. By your calculations, a 7.5" diameter trackball is required to get 1000 point resolution in 1/4 turn. I was assuming a rather gross resolution, about 100 character widths (~10 pixels per character), for moving over the whole screen. Additionally, with fingers spread, at least 4 to 6 inches of displacement may be achieved (I can reach 9", thumb to little finger, but that is unusual). Thus a much smaller ball with less demanding sensitivity may be used. A fine resolution mode (automatic or manual control) may be entered in a local area to work at the bit level. Another point (I only recently thought of) is that the wrist may be turned relative to the trackball so that the fingers may be used to move the trackball diagonally thus avoiding using the palm. With the mouse, such wrist turning merely changes the angle of the mouse relative to the surface, unless it is rotation sensitive. ---------------------------------------------------------------------------- The trackball is better because: 1. The trackball stays in one place Well.... I leave my mouse in the same place -- at least the pad normally stays in the same place, which is good enough for me to acquire the mouse without looking. I think that's the point you're driving at. Partial credit to the trackball. - no clear space on the desk is needed. Again, partial credit: the trackball DOES take up space; it's just figured into the footprint of the keyboard housing. Or it's separate, and your point is simply wrong. Admittedly, most trackball housings are smaller than my 8" X 11" mouse pad. - it can be operated without looking at it, more so than for the mouse. Bull. I never look at my mouse. I will believe I use peripheral vision to acquire it, but I don't have to take my eyes from the screen. Looking at something with peripheral vision still takes brain activity - admittedly not much, but possibly a mental distraction? Because the trackball stays in one place, kinesthetic memory may be used to locate it, whereas peripheral vision (or gropping) must be used to find the mouse if you dont remember where you last left it and you dont look directly at it. - it can be located close to the keyboard for faster access. Again, partial credit: the biggest penalty is taking your hands from home position; that costs about 4 of the 6 keystroke times Card & Moran observed as the cost of acquiring the mouse; both devices pay that penalty. Note also that if your interface allows **keeping the hands on the locator**, the significance of this cost decreases. A little easier to do with one hand pointing & clicking while the other chords functions. - mouse requires cleaner, more secure environment, whereas trackballs are suitable for video games even. True for mechanical mice; not for optical mice -- my mouse is continually cleaning the surface it reads (it rides on felt)! I believe the optical mice are more environmentally secure than trackballs, unless somebody tries to drag the machine by the mouse cord. I meant not only clean of dust, but also clear of objects. An optical trackball is also a possibility to avoid dirty trackball problems. 2. The trackball is omni-positional - that is, the cursor can be located anywhere on the screen whereas the mouse cursor is usually mapped from the mouse position. This allows cursor jumping and returning without large movements of the arm. - wrap-around is reasonable with the infinite range of the trackball. - stopping at window edges and ignoring further motion is possible. This whole point is fubar. Structurally, a mouse is an upside-down trackball. Anything you can do with the trackball, you can do with a mouse. In particular, my system has all the features you mention, implemented on a mouse. If you complain that I must raise the mouse to take another swipe, I will observe that you must raise your hand for ditto. I agree that the mouse can do the same - but in raising the mouse and replacing it elsewhere, the (approx.) 1-1 mapping of mouse position to screen is lost. The trackball does not have any surface to move on and so the mapping is implied only by the motion. Perhaps the mouse position (as opposed to mouse motion) mapping is not so important, but I thought it was one of the selling points given for mice. You make a *big* point of it later. Additionally, the trackball CAN be spun to keep it moving while taking another swipe, although I would prefer to avoid that. Several people thought I thought the mouse uses absolute positioning. But notice I said "the mouse cursor is usually mapped from the mouse position". I didnt say "absolute" postion. I fully understand that it is the mouse motion that is translated into relative position and that this mouse motion may be translated alternatively into cursor velocity or acceleration or whatever. I guess I wasnt all that clear, but it was obvious to me. 3. Scale of cursor motion can be adjusted - very small trackball movements may correspond to large cursor motions and large trackball movements may correspond to small cursor motions. - one can also adjust mouse cursor motion scale, but the mouse to cursor mapping is shifted. Occasionally one must adjust the mouse. As in physically move it back toward the center of the pad? I suppose. I haven't noticed this effect; the discrepancy may be due to the larger effective range of the mouse (see my first response above). 4. A third dimension of trackball rotation in-place is possible. - screen object translation and rotation can be combined in one device. - pressure sensitivity may be added for another, short dimension. You mean rotation in Z? I'll believe it when I see it. I know of 3-D *joysticks*, but they have limited deflection. Think about trying to spin the ball in Z without introducing unwanted translations in X or Y! As it happens, my mouse is rotation sensitive; it's a loss. Similarly, I think you'd have a hard time getting usable pressure from the trackball -- at least no more than the slightly less than 3 bits I get off my mouse buttons. From mark@Umcp-Cs.ARPA You list one of the advantages of track balls over mice as the "third dimension" of rotation in place. Mice have this too, at least the Mouse Systems optical mouse on my sun workstation does. ============================================= From TYSON@SRI-AI.ARPA I don't really see that there is any more degrees of freedom for a ball in a box whether the ball is on top or bottom. I expect that anything that can be done with a trackball can be done with a mouse and vice versa. (The question is, how well can the user use it.) ============================================== Me again: Two or more fingers may be used to twist the trackball on the Z axis. I can well imagine that spinning it could skew it in X or Y. A mouse ball that only contacts the surface at one point cannot be rotated in the Z axis. A "sticky" surface would allow Z axis rotation. Still, the rotation is not "free" in that you would not want to rotate the mouse more than the maximum wrist turning angle. The few bits of info derived from pressure on the trackball could be used to scale the motion of the cursor. I.e., more pressure would translate into finer cursor motion. Or perhaps the opposite would be better. Of course, the mouse could do the same. -------------------------------------------------------------------------- The mouse is better because: 1. The mouse has more direct mapping between arm motion and cursor motion. - consistency has an effect of simplicity and ease of use. Jumping of trackball cursor can be confusing. This is a *big* point -- we have kinesthetic memory: we remember the bone-muscles configuration associated with a space-time event. Thus, having been "there", our arms & hands know how to move to "there" again. This *big* point is the position-screen mapping argument that has one problem. Given the remapping of position to screen whenever the mouse is raised and relocated, "there" is no longer necessarily in the same place. Thus "there" is a relative "there", whereas with the trackball, the kinesthetic memory would be of motion in a given direction, always absolutly centered over the trackball. - the pointing effect of the arm motion may map better in the brain than the rolling of a trackball across fingers and palm. This is another *big* point -- our muscles and neurons are optimized for putting fingertips, not palms, in precise positions. Observe which muscles move your hand across the trackball (biceps, triceps, and wrist rotators) versus moving the mouse (a little wrist rotattion, mostly finger flexing). This has effort / fatigue as well as precision benefits. I would argue that the mouse requires more full arm motion (biceps, etc.) whereas moving the trackball is done more with the wrist and fingers. The finger tips may be used with trackball on a fine (or gross) resolution, but even when using the palm, the fingers would still point to the abstract point on the trackball "surface". I found I stayed away from using my palm, nevertheless. 2. Buttons are easily used on mice. - single arm interaction is possible. - trackball would usually require two arm interaction. Agree. 3. It is nice to have a clear space on your desk. Isn't really there -- if I want to be able to use the mouse, that space is effectively full. I move my keyboard to clear space on my desk. It is not FREE space, but it IS clear space. ---------------------------------------------------------------------------- So what am I missing? The fact that a mouse stops when you do, and stays in that position, makes it effective for moving directly to a destination, rather than piloting it along a course -- the difference is how often you correct in-course, and it's the major determinant of how long it takes you to get to a destination. I believe most of the games you mention have more of a "piloting" flavor, in fact than the "selecting" style of good mouse menus & editors. The trackball certainly stops if you just stop your hand down on it. And it stays there. By "piloting" I presume you mean spinning it. I would prefer to get away from such spinning, but greater sensitivity would be required. The other major determinant of how long it takes you to get to a destination is how far you have to go. Both mice and trackballs could be made more sensitive, but the trackball must be made more sensitive to be sensible. An advantage of the trackball in the game environment is that it is enclosed -- to keep it from being obnoxiously inflexible, the mouse cable is kept very light, hence fragile. A pissed-off drunk could break it easily. This advantage weighs more heavily in a game parlor or bar than an office. From TYSON@SRI-AI.ARPA My experiment with a trackball led me to believe the biggest problem is the momentum of the ball. Because it is relatively big, the momentum must be fairly large. That tended to cause me to miss where I wanted to go. Another problem was the ball tended to move if I bumped the table (ie, it wasn't rock steady when at rest). Another problem is that it takes more coordinated hand movement to turn the trackball one full turn. It takes some back and forth movements. Actually, a lot of the problems I had are due to the mechanical properties of the physical device. Except for that last complaint, there is nothing that precludes using something like a laser device to read hand motion. Interesting idea.... Turn an optical mouse upside down, effectively. ========================== Greater momentum is better for spinning, but, as I have said, I would rather steer clear of spinning. Thus greater sensitivity (and/or grosser cursor motion) would be required to substitute for large trackball motion. I like the idea of directly reading hand motion. How about having finger "pick" like devices on each finger tip which you touch to a pad to do the same as pressing a key on the mouse? Daniel LaLiberte (ihnp4!uiucdcs!liberte) U of Illinois, Urbana-Champaign, Computer Science {moderation in all things - including moderation}