PC Ports at 31.25 Kbps...

Magnus Danielson e93_mda at drum.it.kth.se
Thu Feb 20 04:01:58 CET 1997


>>>>> "g" == gstopp  <gstopp at fibermux.com> writes:

 g>      I'll have to investigate the Pentium data rates! I've only hacked into 
 g>      486's and down so far...

This is a new one... are you sure that this is not the MPU-401 UART
mode being integrated with the gameport once again?

Besides, however you play it you will have a suitable known integer
multiple clock on the motherboard.
     
 g>      I've found that on serial cards (old ones, with 8250 UARTs etc.) you 
 g>      can replace the 1.8432MHz crystal with a 6MHz unit, and now your 
 g>      9.6Kbps setting will yield 31.25Kbps. If you can't find a 6MHz, use a 
 g>      12MHz and set the baud rate to 4800.

Are you sure it is a wise thing to jump so much in frequency? I would
no jump to far cause I would suspect the capacitors on each side of
the crystal migth need adjustments in order to stay in tune properly...
     
 g>      Even though the older PC hardware architecture gives you the ability 
 g>      to derive the 16x UART clock using a 16-bit divisor, the divide-by-1 
 g>      starts with a base frequency of 115.2Kbps, with /2 giving 57.6K and /3 
 g>      giving 38.4K. The /4 gives you 28.8K, so 31.25K falls between with a 
 g>      divisor of something like 3.6 (which is of course not attainable).

Not that many PCs have it, but a real RS232D serial port will also
have transmitt and recieve clocks. If PCs had this, we could use them
for a much wider range of stuff, unfortunatly is such great stuff
removed by the cost optimation that has happend over time... but I
think that some older cards migth still have this feature.
     
 g>      Also - I have been eyeballing those motherboards, such as Gateway2000 
 g>      and Compaq, that have everything integrated (serial ports, video, disk 
 g>      controller), which would make great one-circuit-board embedded CPUs 
 g>      for projects, but they use custom VLSI chips that emulate the UARTs 
 g>      and if you change the 24M or 48M crystal to something that's a 
 g>      multiple of MIDI rates, then the CPU won't even boot. Rats.

Throwing in the frequency conversion stuff in a corner of todays VLSI
chipsets is done at a very low extra cost... only a few components
cause a trouble in the process... dividers and phasecomparators is cheap.

Cheers,
Magnus



More information about the Synth-diy mailing list