On Jun 24, 2021, at 8:51 PM, Grant Taylor via cctalk
<cctalk at classiccmp.org> wrote:
On 6/22/21 9:12 AM, Ethan Dicks via cctalk wrote:
The OP said he meant with "real"
connectors, but in my case, I've encountered strange buffering issues with USB serial
dongles (since they are really block-mode devices, not character-at-a-time) and I've
definitely had problems supporting lines with odd parameters (especially speeds slower
than 300 baud or with 5-bits-per-char, like one would use for a Model 19 or Model 28
teletype). The hardware UARTs on AVR processors implement those juse fine (though for
"50 baud", you often have to put a slower crystal on the processor because the
16-bit divisor overflows at 16-20MHz). The "soft serial" libraries often just
hard-code 8-bit implementations. Fine for modern stuff but I have uses for connecting to
electromechanical serial devices.
These seem like real problems, which can't be overcome by a passive physical
adapter.
These also seem like implementation problems to me. At least more than they seem like a
USB spec problem. I naively assume that if someone wanted to produce a USB-to-Serial
adapter that supported the things you're describing that they could do so. But sadly,
I believe that RoI will be on the wrong side of the demand curve.
An Arduino or something of that size can easily do a USB to serial adapter in software.
That would let you do any data rate and character length the device can so (if you use its
UART) or whatever you can generate in bit-banging software. For example, a Raspberry Pico
can clearly do any rate you want, including strange slow ones or oddballs like 134.5 baud
6 bits (or is that 7 -- for the serial line 2741). 45.45 baud 5 level and some strange
fractional stop bit, not quite 1.5 -- no problem. 10 bit characters (for the classic
PLATO terminal to host direction) -- no problem.
paul