stever@videovax.UUCP (Steven E. Rice) (09/10/86)
In article <277@pttesac.UUCP>, Marnix van Ammers (vanam@pttesac.UUCP) writes: > In article <6802@sun.uucp> cmcmanis@sun.uucp (Chuck McManis) writes: >> Three problems with putting the sidecar on top of the Amiga : >> o Noise (from the buss cable) >> o Timing ( every nanosecond counts on this bus) >> o FCC (unless you put the thing in a faraday box you would blast >> recption for miles) > > Sorry, I'm not an expert, but I don't buy any of those reasons. > I feel the same dissapointment as Chuck McManis does. I don't > have space to the side of my Amiga. The buss cable could be shielded. > It would only have to be about 4 inches long. I don't believe that > would screw up the timing (I understand that a signal propagates > more than 12 inches in a nano-second, so 4 inches would effect > timing by 1/3 nano-second or so -- can't be that critical -- > the CPU isn't going all *that* fast afterall). In free space, light (and radio waves) travel a bit over a foot per nanosecond. In a cable, the propagation velocity tends to be in the neighborhood of 2/3 c -- about 8 or 9 inches per nanosecond. But that's not the real problem! Whenever a signal propagates down a cable, it behaves according to the laws that govern transmission lines. Typical shielded cables (ribbon cables and such) have *very* low characteristic impedances (in the 30 to 50 Ohm range). Because of the characteristics of the transmission line, the driver sees a very low impedance, and strains to output the very large current required whenever it switches. But once the current gets to the other end of the transmission line, it has no place to go except back where it came from -- unless the cable is terminated in its characteristic impedance, which would herniate the poor 68000! So you get reflections. The result is very large voltage spikes on the bus lines, which often go above Vcc and below ground (this is a no-no for MOS circuits). If the round-trip time for the reflections (2 times the cable delay plus 2 times the propagation delay contributed by each circuit board) is of the same order of magnitude as the rise time of the signals (less than 5 nsec for the 68000), the reflections can cause the bus signals to cross the transition region (+0.8v to 2.0v) several times for each signal transition. This tends to cause "unreliable operation". . . If you are going to design a system using a cable in the system bus, the cable should be buffered at both ends with bipolar bus buffers. If the system is especially fast, a terminated bus will be required (read "lots of power and big bucks"). And whenever you add bus buffers, you add delay (maybe another wait state?). Since the Amiga was designed for a mass market, it is not surprising that that it does not have a fully buffered bus. (DEC and others make some really nice machines with buffered busses. . .) Steve Rice ---------------------------------------------------------------------------- {decvax | hplabs | ihnp4 | uw-beaver}!tektronix!videovax!stever