>     The errata involving BRGH=1 for the USART is present on all midrange
>     microcontrollers with the USART module including the newer 'A'
>     devices.
>
>     This errata only affects the reciever of the USART, the transmitter
>     remains unaffected.  We have found that some applications do not
>     experience the errata.  It seems to be dependent on the oscillator
>     frequency of the receiver, the baud rate, and the baud rate error of
>     the transmitter.  The type of errors you would see are either framing
>     errors or the data will simply be wrong.
>
>     We have not been able to precisly define which combinations of osc.
>     frequency and baud rate that do not produce errors, therefore we have
>     suggested in the errata to not use BRGH = 1.
>
>     Regards,
>
>     Rodger Richey
>     Sr. Applications Engineer
>     Microchip Technology Incorporated
>
>

Our experience seems to confirm this, we are using the 16C74 in
applications with
BRGH=1, and after **extensive** error rate testing concluded that there was no
difference in error rates with BRGH=1 and BRGH=0.

The tests we carried out were using short (8-10) byte packets running HDX RS485
at 156250 baud with fosc=4.9152.  We are talking to a Host Processor running our
own network protocol and thus have complete control over both ends of the link.

It may be that the problem is increased sensitivity to noise and mismatched baud
rates with the lower divisor. This would seem to indicate that the problem is
not really a bug in the chip as such. More of a specification question.

                                        regards

Ray Gardiner, Shepparton, Victoria, Australia       ray@netspace.net.au
Technical Director dsp systems               http://www.dsp-systems.com