[Coco] Serial Port file transfers

[Gene Heskett]

On Friday 10 November 2006 14:23, L. Curtis Boyle wrote:
>On Mon, 06 Nov 2006 00:30:52 -0600, Gene Heskett
>
><gene.heskett at verizon.net> wrote:
>> On Sunday 05 November 2006 23:03, Jim Hickle wrote:
>>> What sort of speeds should I be expecting from a 512k CoCo running
>>> NitrOS9, with an RS-232 pack connected to a Linux system's serial
>>> port? I'm trying to transfer files and it's taking forever using
>>> floppies and the PCDos program.
>>>
>>> Using rz/sz on the coco I can set the ports no higher than 2400 bps. 
>>> I can bump it up to 4800 if transferring using Supercomm's Xmodem
>>> transfer.  Is this really the limit?   Maybe I should give up on
>>> serial transfers and use CD-ROM's?
>>
>> Years ago now, I was using an amiga-2000-040 and Olaf Barthels term4.7
>> to measure that.  I was getting around 740 cps with the interfaces set
>> for 9600 on a nitros9-1.15 system, and about 430 cps with a stock 512k
>> coco3. This was using the auto-trigger of rz/sz-3.36 with supercom-2.3
>> for the term program on the coco3's.  Using the 7-wire protocol for
>> flow control, not xon/xoff.
>
>     This was using Nitros9 as originally intended on a 6309 chip... I
>don't know if Jim is using a 6809 or 6309 chip in his scenario. I know I
>used to get about 700-800 CPS with a 9600 connection with RZ/SZ (which
>really needs a rewrite to use proper buffering to speed it up), 

I've looked at that myself a decade+ back when I did the 3.36 version I 
think everyone is still running, but the character orientation of the 
code flow is so endemic, it litterally would be a major rewrite.  Paul 
thought he could just add a buffer, but if it ever worked I never saw it 
help. 3.36 doesn't buffer till the character is fully processed and the 
crc has been determined.  The table lookup crc was my contribution. With 
that switched off, the speed fell like a rock, to about 390 cps in my 
tests at the time.  That was something else Paul refused to use, simply 
because he never understood assembly and couldn't for the life of him 
understand that a 'tfr b,a' & 'clear b' was in fact an 8 bit left shift.  
Sigh...

>and with 
> a 38400 connection (Eliminator from Frank Hogg Labs), I could get 2000+
> cps with Ymodem/Ymodem batch. If you have the Randy RS-232 pak, then
> about 1700-1800 cps would probably be the max with X/Ymodem (x is
> slower, since it has more overhead) using a decent term program like
> Supercomm or OSTerm. Once again, running the 6809 version of Nitros9
> might be a bit slower; I am basing on the 6309 version.

Yup, my figures above were all on a 63C09.

>> Using xmoden or ymodem, speeds in the 3000 or more cps range should be
>> doable on a stock coco3 since they have a much simpler error checking
>> method and send whole 128 byte blocks at a time whereas rz/sz is
>> character
>> at a time which makes its 24 bit crc checking a huge cpu hog.  But,
>> rz/sz absolutely guarantees the data sent is good, the other two
>> methods can let
>> errors slip thru, and file sizes need trimmed as they record whole 128
>> byte blocks even if the last byte of the file is the first byte of the
>> last block, which makes the rx'd file 127 bytes too long.
>
>    Y-batch also does exact file sizes correctly, like Z, as well as
>multiple files in a single transfer.

y-batch was not a commonly available thing when I was working on rzsz.  I 
think it came out about the time I was well into the amiga scene.

>> If you can't set the ports any higher than 2400 baud, you have a flow
>> control miss-match I think.  And because the acia chip has a bug,
>> there was a minor re-wire of the rs-232 pack and a patch for sacia
>> that fixed that bug if you wanted to use hardware flow control,
>> commonly called 7-wire.  That patch should be in the later sacia
>> modules, its enabled by one of the xmode options and isn't documented
>> anyplace but in that patch's
>> readme that I know of.
>>
>> Unforch, I don't recall the name of the patch, and I'm not sure that
>> it survived the hack job when rtsi got hacked into several years ago. 
>> I don't believe I have it myself now, as it was a casualty of a flaky
>> hard drive & controller a decade+ back up the log.

Can you or anyone else recall the status of that hack&patch, and the 
current state of the sacia driver code to make use of it?

>>> -jim hickle
>>>
>>>
>>>
>>>
>>> ---------------------------------
>>> Get your email and see which of your friends are online - Right on
>>> the new Yahoo.com
>
>--
>L. Curtis Boyle

-- 
Cheers, Gene