cprice@mips.COM (Charlie Price) (08/01/90)
In article <10048@pt.cs.cmu.edu> lindsay@MATHOM.GANDALF.CS.CMU.EDU (Donald Lindsay) writes: > >Beats me. However, I would certainly hope that the access mechanism >involves fewer moving parts than it does now. > >Specifically, it would be nice if we scanned a laser beam over the >media, rather than rotated the media. Or, at least, did the "head" >movement that way. There was hope for this sort of thing, a decade >ago, and somehow it never happened. I believe that the best spatial >modulators had very limited angular effect: perhaps we should be >trying to shrink the CDROM disks, just as we've been shrinking the >magnetic disks. >-- >Don D.C.Lindsay Do you have any reason to think that "scanning" a memory media at varying angles could be made to work (economically)? Don't CDs at least, pretty much *depend* on the read beam being delivered orthogonally to the media since the effect used to sense bits is interference in reflected light, not the amount of light reflected? A second problem is knowing whether you are really looking at the information that you think you are. A disk has a lot of "servo" data that lets you position to read and lets you know that you are in the right place. If you don't have a separate servo surface, you need some servo data embedded in the track every now and then to correct on-track positioning, make sure that you are on the track you think you are on, and maybe resynchronize to a data bit rate. An embedded servo system depends to some extent on the mass of the head arm to keep it from drifting very far off track over short periods. It also depends on the mass of the rotating spindle to keep it from changing velocity too quickly and thereby changing the data bit rate too much over short periods. A system that depends on mass is immune to some level of vibration because you don't move the masses quickly enough that you can't compensate for them. Suppose that orthogonal illumination isn't a problem, and you can compensate for light delivered and reflected at an angle. You have to LOOK at the reflected light, right? I can't think of any way to do this without moving something that is at least as large and massive as the optical head itself. I can imagine an arrangement that would allow you to scan "straight lines" at high speed, but would require something to move to get to another line. Moreover, if you end up with the illuminating beam and the sensor not rigidly attached together, you have introduced another spatial relationship that must be managed to make the drive work. Other problems are storage density and precision. If you want high density, you need small recording domains (spots) packed very close together. Illuminating the media at an angle changes the spot size at varying angle and changes the size of the spots. It also increases the absolute position error on the media surface, so the spots are both larger and spaced further away from each other. You get fewer spots per surface. This is OK, just a tradeoff to be made (and it is made in mag drives). You need some way to either compensate for or render unimportant defects in the media (nothing is perfectly flat, perfectly smooth, identically transparent, ...). This is my linear thinking speaking. Probably this will be solved by doing something that is a posteriori entirely obvious. Anybody for the optical christmas tree ornament? You could make spheres (or partial spheres) that were rigid enough and speherical enough that you could position an illumination source and sense cluster in the middle and you would illuminate/reflect orthoginally without moving the media or moving the laser very much. You just snap the new media over the laser head and away you go. I can also imagine spinning a disk like you do today, but doing a very fast scan along a slighty curved line on the surface by scanning a laser beam across a radius of the disk. You would use the disk surface something like a VCR uses the recording surface of a tape. You would still have latency to an individual track, but you could potentially read the entire disk surface out in a few revolutions. -- Charlie Price cprice@mips.mips.com (408) 720-1700 MIPS Computer Systems / 928 Arques Ave. / Sunnyvale, CA 94086
pcg@cs.aber.ac.uk (Piercarlo Grandi) (08/03/90)
"cprice" == Charlie Price ??? writes: cprice> In article <10048@pt.cs.cmu.edu> cprice> lindsay@MATHOM.GANDALF.CS.CMU.EDU (Donald Lindsay) writes: lindsay> Beats me. However, I would certainly hope that the access mechanism lindsay> involves fewer moving parts than it does now. lindsay> Specifically, it would be nice if we scanned a laser beam over the lindsay> media, rather than rotated the media. cprice> Don't CDs at least, pretty much *depend* on the read beam being cprice> delivered orthogonally to the media since the effect used cprice> to sense bits is interference in reflected light, not the cprice> amount of light reflected? Here you have the possibility of using a cup-shaped medium (or cylinder shaped, if you use a light source array instead of a point). There is no reason to restrict oneself to flat recording media, as you later observe musign about recording spheres. cprice> A second problem is knowing whether you are really looking at cprice> the information that you think you are. A disk has a lot of cprice> "servo" data that lets you position to read and lets you know cprice> that you are in the right place. [ ... and heavvy heads cprice> providing hysteresis in positioning ... ] You would use television like scanning, like CRT memories of the fifties. Precision positioning can be done, and then you can inscribe the medium with servo information within each scan line, to keep the scanning beam in synch. Or you could instead random position the beam, by using a less precise homing beam onto larger features of the surface, and then the read/write beam. cprice> A system that depends on mass is immune to some level of vibration cprice> because you don't move the masses quickly enough that you cprice> can't compensate for them. But we are using granite cabinets here, of course. Thermal problems can get us pretty badly, not just vibration. We also would like to use phase conjugated mirrors as well, for the same reasons. I have little doubt (let's be optimistic) that the technology exists, if only because it has been developed for telescopes, and optical and electronic microscopes, not to mention chip litographers of various types. cprice> You have to LOOK at the reflected light, right? Why ever? You can just use as recording surface something wih holes in it, and put under it a photoconductive surface. Remember that we could be scanning the recording surface strictly sequentially, so looking at only one bit at a time, of known coordinates. cprice> You need some way to either compensate for or render unimportant cprice> defects in the media (nothing is perfectly flat, perfectly smooth, cprice> identically transparent, ...). This was a problem also with CRT memories (carbon dust from the coils falling on the phospor, and thus literally blacking out a bit), and that is why they (Manchester University) started working with ECC like ideas. In case this is not already obvious, I have been thinking about a granite pyramid a few feet tall with a cup shaped recording medium in its base and a laser eye at its top. :-) :-) This would have no moving parts at all (if Fabry-Perot devices are used to do the modulation and deflection), nanosecond access times and terabit capacities (if we can find a recording medium that is good enough), write once properties, and a fairly steep price. One would use a very fast photoconductive/sensitive recording medium in which you burn holes to write; when you read, if the laser goes thru the hole and the surface is not illuminated, if there is no hole, the laser hits and illuminates the surface. The surface would be black and very thin to avoid requiring a large increase in the laser power to write thru, and writing would be done with a slightly wider beam than for reading. There could be endless variations; the surface could just be "black paint", and the photoconductor/sensing surface coated beneath it, or a transparent surface that blisters becoming opaque when burned, etc... The surface could become diffractive, or shorter/taller by a wawelength, getting beam cancellation, or whatever. There are loads of optical effects. Does anybody know what could be used? Do we have any hope? -- Piercarlo "Peter" Grandi | ARPA: pcg%cs.aber.ac.uk@nsfnet-relay.ac.uk Dept of CS, UCW Aberystwyth | UUCP: ...!mcsun!ukc!aber-cs!pcg Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk