Thanks for the insights guys. I will restate my problem.

I have two PIC16F88 and two PIC16F877A on a single 100 mm square board. I have been using serin/out2 with the mode number set to 16390 which gives 38,400 bps. I use an RTS/CTS system to pace the communications and send every packet with a checksum. This has been a reliable means of communicating between the PICs but I now need it faster. I don't care what the speeds are as long as they are faster and preferably up around 100,000 bps.

The boards are built and shiftin/shiftout traces are not on it. I also don't see how shiftin/shiftout can easily work in a peer to peer system. Shiftin/out needs a master to provide the clock. I cannot, or don't want to, use the USART HSERIN/OUT as this is reserved for the bootloaders and physically wired on the PCB to the bootloader ports. In a previous project I had intermittent trouble at startup with USART to USART comms where both PICs had MCSP bootloaders running. I think that on occasions the bootloaders would see each other and think mother was calling. They appeared to start a brief but merry dance between themselves and the system would then crash. The boot code was never overwritten thankfully.

From a peek at the SERIN2 tables on the Melabs web site it seems that there are only certain speeds available to SERIN2. I don't know what happens if you just pick mode numbers at random but I doubt it will be pleasant.

Because the baud rates are the result of some unknown (to me) math operations, and we have integer math divisor truncations, I am pretty sure there will be a few optimum and many sub-optimum baud rates with DEBUGIN and DEBUG. I don't care if the speeds are 'non-standard', I just want fast reliable comms.

Does anyone have a table of preferred DEBUG/IN speeds similar to the SERIN/OUT2 speeds on the melabs web site or can they point me to an algorithm that computes this?

Cheers
Brian