[rescue] SS Classic Serial ports

[Scott Newell]

At 08:00 PM 4/23/2003 -0400, Charles Shannon Hendrix wrote:
>Now, that's not because the chips are so terrible, but because Sun did
>odd things in the implementation.  They use seperate clocks for in and
>out, which drift and every N frames cause an error at high speeds.

Hmmm...  Do you mean that Sun ignored the internal baud rate generators on
the 8530 uarts, and generated separate clocks for the shift registers
through the RTxCA, RTxCB, TRxCA, and TRxCB pins?  I don't think so.  The
uart has a baud rate generator for each channel, and the clocks for tx and
rx on a single channel are shared.  Looking through the OpenBSD source, it
sure appears to me that the default situation is to use the internal baud
rate generators.


>You are supposed to be able to give them an external clock on one of the
>serial port pins, and get higher speeds, but I don't know of anyone who
>has done this, just a lot of talk between various serial code hackers
>like der Mouse on the netbsd lists.

Doing this would bypass the internal baud rate generators.  If you fed
separate clocks to the receive and transmit sections, you might end up with
the split clock problem you mentioned above.  (Although I'm not sure that
both the tx and tx clock are brought out to a connector--need to do some
beepin' to find out for sure.)


newell