KEN@JPL-VLSI.ARPA (09/10/84)
From: Ken Adelman <KEN@JPL-VLSI.ARPA> Rumor has it that Woz (or someone that he programmed) thought that apple could save a few dollars per apple by using a timing signal that already existed on the motherboard instead of a 1.84320 Mhz signal that the 6551 requires. Since I have yet to hook my modem to a //c, I haven't experienced this 1165 baud problem myself. The same rumor said that if you returned your //c to your dealer and complained about it, that he would solder a 1.84320 Mhz crystal onto the board in a pad that apple left for such a purpose. Attempts to get our local dealer to this resulted in responses like "but I don't even have the service manual...". If you don't mind voiding the warranty, adding the crystal would be quite easy. I recall that the 6551 will support either a crystal being directly connected to it or an external timing signal. Sounds like cutting one wire and soldering two. I would be willing to look up the pin numbers on the 6551 which the crystal should be connected across, in case it is not obvious upon opening the apple. I would be more interested in hearing from someone who has managed to get apple to fix it. I don't know if apple's modems will work with it, but I suspect that if they send the characters over the phone at the same rate that their modem won't talk to a Vadic (1205 baud) on the other end. -Ken Adelman Caltech ------
darrelj@sdcrdcf.UUCP (09/10/84)
>From the various descriptions of the problem, the Apple is providing
one half the color burst frequency instead of 1.8432Mhz for baud rate generation.
The former is easily derived from the 4x color burst crystal used in all
Apples (makes generation of color images an incidental (but wanted) side
effect of the B&W video circuitry. The latter is one of a half-dozen
popular frequencies for Baud rate generation (1.8432MHz = 2^13 * 3^2 * 5^2)
which get most popular rates exactly (e.g. 1.8432MHz/3/512 = 1200) with
simple counters. (For asyncronous transmission, you normally want 16*baud).
There are also three completely different circuits used for baud rate
generators.
a) A couple of chains of low divisor counters with output taps between most
stages (e.g. once you get 9600, sucessive halvings get most other common
rates
b) A polynomial (mod 2) shift register which can divide by almost number but
requires initialization values which are hard to compute. Generally these
values are hardwired into a small table. It's advantages are that you
can select almost any (fixed) set of input/output relationships with less
circuitry than (c).
c) A general divide by n counter of 12 to 16 bits. Takes more circuitry,
but easy to program. E.g. to get 1200x16 clock for most UARTs at 1200 baud,
program divide by reference/(1200x16).
Apple seems to be using the divisor needed for a 1.8432MHz crystal (96)
with 1,789,773 Hz reference (which needs a divisor just over 93).
[With type a circuitry, you can't divide by 93, so you're stuck with
1165.22 baud; with type b, your stuck if you use a part meant for 1.8432MHz,
instead of paying for a custom version; with type c or custom b, you have
your choice of 93 as a divisor (giving 1203 baud) or 94 (giving 1190 baud).
The fastest fix is to provide the right crystal for the serial chip (note
that you will need a resistor or capacitor in addition to the crystal;
you will have to check the data sheet for the proper crystal hookup).
It also fixes all baud rates.
With a type c baud rate generator, an alternate fix is software, but waiting
for Apple to produce and distribute a revised ROM will take months, and will
still leave high baud rates on shakey grounds; at 9600 baud, you are stuck
with dividing by 12, giving a rate 3% low again.
[An amazing amount of trouble to save $5 for the crystal!!!]
--
Darrel J. Van Buer, PhD
System Development Corp.
2500 Colorado Ave
Santa Monica, CA 90406
(213)820-4111 x5449
...{allegra,burdvax,cbosgd,hplabs,ihnp4,orstcs,sdcsvax,ucla-cs,akgua}
!sdcrdcf!darrelj
VANBUER@USC-ECL.ARPABuehring%ti-csl.csnet@CSNET-RELAY.ARPA (09/16/84)
From: Walt <Buehring%ti-csl.csnet@CSNET-RELAY.ARPA> Spoke with the local Apple office (Dallas) about the //c modem problem. Didn't expect much but was given sympathy, an acknowledgment that the problem exists, and the promise that a notice would go to all dealers sometime in September which would instruct them on how to correct the problem (you may have to "prove" that your //c doesn't work with your modem and comm package). No quick fixes. Guess I'll twiddle my thumbs a few weeks then start hounding my dealer. Re: putting in your own crystal: I *love* to hack hardware -- other people's. /Walt/ -------