The Atari 850, the most pointless device Atari ever made...

[Level42]

I bought an 850 only when the price had dropped to just 150 Dutch guilders at a local Atari specialized shop BITD. This was about as expensive as a dedicated parallel interface from third party so I figured I'd go with it because the serial ports might be nice,

 

I never used the serial ports for anything. I never had a modem on my A8's.

 

Only the C.Itoh 8510 printer on the parallel port. It worked fine and I liked the 800 style design, it showed it's build quality.

 

Today I had a look at the schematics. I already knew BITD that the 850 had its own processor and was thus a small computer by itself.

 

The thing mainly exists of 4 chips:

1x 6507 processor

1x 2532 EPROM

2x 6532 RIOT

 

Looking at the design I can only have one feeling, one question......why on earth didn't Atari decide to put 1 6532 on the A8's main boards ???

As far as I can see this would allow one parallel and one RS-232 interface. (Who needs 4 RS-232's ?).

 

This would require 1 6532 maybe some buffer chips and 2 connectors....

 

Atari had and used truckloads of 6532 since it was used in the 2600 so the price could have barely been a problem.

 

This would have given the A8's industry standard connections from day 1 and would have really given it (yet another) edge over the competition.

 

It really bugs my mind why they decided to go with an external case, which needed its own power supply (externally and internally), power switch, 2 SIO connectors, SIO interfacing, 6507 and EPROM. All of this is just extra ballast and added up to the hefty initial price tag of the 850.

 

An internal RIOT could have simply been joined on the internal busses, drivers could have been included in the OS-ROM.

 

So what do you guys think ? I think Atari was too arrogant and too protective (wanting the user only to buy their external,devices) in the days of the 400/800.

But even when the XL series was introduced they would have had a chance to get a big advantage over competitors like the C64 by including this single chip of which they already had truckloads....

Edited July 31, 2015 by Level42

[pixelmischief]

Who needs 4 RS-232's?

 

I'm pretty sure this was a big deal for BBS hosts.

[Level42]

I'm pretty sure this was a big deal for BBS hosts.

OK but that was a very tiny market segment for the A8 platform.

Edited July 31, 2015 by Level42

[Joey Z]

 

I'm pretty sure this was a big deal for BBS hosts.

but did any BBS software actually support more than one user at a time? The 850 itself only really supports one port open at a time anyway.

[FifthPlayer]

I suspect it gets back to RFI emissions. The 400/800 designers were scared of running afoul of the FCC for certification, so they were very conservative in the system design.

 

SIO came about because they wanted the 800 to have an expansion bus with plug-in cards like the Apple II, but would never have been able to meet the (then-strict) emissions standards for a consumer electronics product.

Edited August 1, 2015 by FifthPlayer

[+David_P]

The 850 has problems with multiple concurrent connections, since it's bit-banging on a a CPU that doesn't have interrupts.

[Rybags]

850 was overkill for what a lot of people needed and the price reflected that.

 

Pointless - not really. The industry at all levels was full of such black boxes used to interface incompatible host/peripherals back then and even now to an extent.

It was just a sign of those times - everyone had their own standards. Today we have some universal standards and the ones we ended up with weren't necessarily the best or cheapest ones.

[tschak909]

If you look at the serial port configurations for each of the four ports, they are different variations of RS-232, each with a different combination of pins.

 

Port 1 was best for MODEMs, as it had DTR, DSR, RTS, as well as carrier detect (DCD).

Port 2 and 3 worked well for devices that only required working RXD/TXD pin pairs, and a single DTR pin, such as plugging up terminals, or devices like serial printers.

Port 4 was very special, and was ultimately designed for 20ma current loop devices. You could for example, plug a Teletype here, or plug into an industrial control system, or one of the 20ma comms ports of a minicomputer.

 

Atari ultimately made these design decisions because they knew the entry cost of the expansion was going to be higher, due to the intelligent device requirement for SIO, FCC, and the like. Useless? Maybe to you, but I've used all four serial ports, for different things, precisely because Atari decided to put them there.

 

-Thom

[fujidude]

If you look at the serial port configurations for each of the four ports, they are different variations of RS-232, each with a different combination of pins.

 

Port 1 was best for MODEMs, as it had DTR, DSR, RTS, as well as carrier detect (DCD).

Port 2 and 3 worked well for devices that only required working RXD/TXD pin pairs, and a single DTR pin, such as plugging up terminals, or devices like serial printers.

Port 4 was very special, and was ultimately designed for 20ma current loop devices. You could for example, plug a Teletype here, or plug into an industrial control system, or one of the 20ma comms ports of a minicomputer.

 

Atari ultimately made these design decisions because they knew the entry cost of the expansion was going to be higher, due to the intelligent device requirement for SIO, FCC, and the like. Useless? Maybe to you, but I've used all four serial ports, for different things, precisely because Atari decided to put them there.

 

-Thom

 

My understanding was that the 850 did not support hardware flow control? That according at least to Bob Puff, but I'm pretty sure other sources too which I cannot put a finger on. It's one of the big reasons I bought a Black Box. It also supported a 19.2Kbps connection to a DCE, rather than 9600bps of the 850.

[tschak909]

Sorry, yes, no RTS.

 

-Thom

[thorfdbg]

So what do you guys think ? I think Atari was too arrogant and too protective (wanting the user only to buy their external,devices) in the days of the 400/800.

But even when the XL series was introduced they would have had a chance to get a big advantage over competitors like the C64 by including this single chip of which they already had truckloads....

 

The 850 is actually one of the more useful devices, it's certainly not pointless. It is just badly engineered. The parallel port is very useful for printing indeed, and it is a pity that this did not go into the main system. As far as the serial ports were desiged: Well, unfortunately, you could only use one of them at a time because the SIO protocol does not allow more. Or rather, because you had to disable the SIO protocol for any serious serial transfer in its "concurrent" mode. Indeed, the 850 does not support hardware flow control: The problem is, as soon as the device is in "concurrent mode", you can no longer switch the status of the ports because SIO is blocked by the serial transfer itself.

 

The funny part of the 850 is: As soon as you enter the concurrent mode, and that is the only mode that allows receiving serial data, the serial protocol is not decoded by the 850, but by POKEY in the main unit. All the 850 does then is to sample the serial input line at the frequency of the CPU, and mirror the state of the serial data on its SIO interface, so it acts as a "dumb passthrough". The limit of 9600 baud is simply because the CPU in the 850 cannot sample higher.

 

What Atari should have done is to add a small buffer of probably 128 bytes in the unit, and trigger an interrupt (yes, there is a line for that on SIO) to signal that the buffer is close to overrun or as soon as data is ready, so the main unit could fetch incoming serial data without requiring to mirror the SIO line.

 

Unfortunately, Atari went cheap (again) and used only the RAM in the RIOT for all of the memory in the device (for stack and zero page), so nothing was left for the buffer. The end result is a very cludgy design: You could run it as pure output device in the "small block mode" in which the 850 handled the serial communication and SIO was free for other data, or you could run it in "concurrent" mode, in which you could also *read* serial data, but SIO was blocked and you could not save your data to disk without leaving concurrent mode. Even worse, hardware handshake also requires leaving concurrent mode. *UGLY*

[+David_P]

There are actually two interrupt lines in the SIO - INTERRUPT and PROCEED. Imagine an intelligent peripheral that could alert for two different conditions, and have the computer react accordingly.

[Level42]

In that light, it might very well have been that they actually chose to go with the 6507 and 6532 (2/3 of a 2600) since they had these around cheaply (the more volume, the lower the price) when designing the 850.

 

However.....I still wonder, would it have been as easy as I think to include a RIOT into the A8 design ? Would the supporting routines still fit in the OS-ROM and more generally....would it have given the A8 platform an advantage over the competition ?

[Level42]

Hah, if they had popped in a TIA they could have made it to double as a 2600 "adapter" ...

[Bryan]

Hah, if they had popped in a TIA they could have made it to double as a 2600 "adapter" ...

Same with the 1050!

 

The 850 existed because the cost of the 400/800 was already pretty high and making a computer with SIO and a slew of industry ports would have meant $100's more in retail cost for things most people wouldn't use. If you eliminated SIO, then you'd need another way to handle floppy drives.

 

I do think the 850 is underpowered, though. Had it been Z80 based, it could have been a CPM module as well.

[Joey Z]

and many of these are reasons why I want to make a R: device emulation within RespeQt which uses the full power of a modern machine. The possibility for a gigantic TX and RX buffer on the emulated 'device' is just one thing. Additionally, the existence of these buffers allows the device to use the SIO protocol like a normal device does. Read and write individual blocks, rather than 'concurrent mode' style. This would allow handshaking changes during comms, even disk access. No data loss would occur, unless you really take a long time. I am also looking at implementing the interupt and proceed lines in an SIO2PC device, and seeing how that does.

 

However, before all of that, I'd like to implement 850 emulation.

[Kyle22]

...

I do think the 850 is underpowered, though. Had it been Z80 based, it could have been a CPM module as well.

That's the ATR-8000.

[Bryan]

That's the ATR-8000.

Or an Indus GT.

[Joey Z]

Or an Indus GT.

that didn't have a serial port though, did it? Kyle22 said the ATR-8000 because it was a serial and a printer interface as well as a disk drive and CP/M machine.

[Bryan]

that didn't have a serial port though, did it? Kyle22 said the ATR-8000 because it was a serial and a printer interface as well as a disk drive and CP/M machine.

Nope, just an option for CP/M.

[Mathy]

Hello guys

 

Or an Indus GT.

 

Or Atari's own 1060.

 

Sincerely

 

Mathy

[+MrFish]

Or Atari's own 1060.

 

Or the 1066 for that matter...

[Bryan]

I wonder how big of a draw CP/M would have been, even with a cheap device.

[gozar]

I wonder how big of a draw CP/M would have been, even with a cheap device.

 

Reading some of the magazines from the early 80s, it would have been a pretty big deal. It looks like CP/M was still a player up until 82 or 83.

[+DrVenkman]

 

Reading some of the magazines from the early 80s, it would have been a pretty big deal. It looks like CP/M was still a player up until 82 or 83.

 

I have no dog in this hunt except that just today I happened to be reading through A.N.A.L.O.G. #13 (Sept. '83) on Archive.org and there was a writeup on the announced XL CP/M module. There's a fairly extensive writeup of the product along with a list of expected CP/M software believed to be in the works such as WordStar, MultiPlan, dBASE II, PeachPak accounting software and more.