XL Design Change Intent

[TemplarXB]

My understanding of the intent of the original 8-bit was to have internal expansion like the Apple II line, but the FCC regs resulted in it being encased in aluminum and using the SIO port being developed.  

 

After the FCC regs were changed, why didn't Atari design the XL line with that internal expansion?  Did the priority change to cost cutting, or was that the intent of the 1090XL?

[xrbrevin]

Tramiel?

[Faicuai]

1 hour ago, TemplarXB said:

My understanding of the intent of the original 8-bit was to have internal expansion like the Apple II line, but the FCC regs resulted in it being encased in aluminum and using the SIO port being developed.  

 

After the FCC regs were changed, why didn't Atari design the XL line with that internal expansion?  Did the priority change to cost cutting, or was that the intent of the 1090XL?

Does not seem fully supported by the state of the industry, back then.

 

The Apple-II design, system-architecture and enclosure all pre-date the 400 / 800... yet the Apple-Ii is (by far) the most flexible and expandable 8-bit system (post 1976-1977) that I am aware of.

 

The 800 itself does provide expandability on its internal bay, at the expense of higher component density and difficulty with routing-out cabling (which can be solved, relatively easily and elegantly)... and it enables such without the presence of PBI / system-bus on the bay. Even the 800's architecture is clearly different (and superior, in my opinion) that anything Atari produced afterwards, 8-bit wise.

 

As for the rest, the XL and XE lines were (essentially) a capitulation for Atari. That is, those products were the result of the company's realization that it will never be able to catch up with Apple in any significant way, and its only hopes were to compete against the Tramielized C64, which signaled the end of most of this industry's segment, thus paving the way to consolidate Apple and IBM (real) success... and only survivors of that era, in today's time. 

Edited May 12, 2020 by Faicuai

[+DrVenkman]

The XL machines were designed well before Tramiel’s involvement. They were intended as simplified construction, cost-reduced versions of the line, with refreshed styling and physical design. Take a look at how many separate boards and components were needed in an 800 versus an 800XL for example. 

 

[+slx]

The 800XL was supposed to be expandable via the PBI connector (and 1090 box).

[gilsaluki]

There were a few companies that supported the PBI connector (ICDs MIO box, CSSs Black Box), but not enough companies saw a profit in doing so.  Very sad.  There is no 8-bit that had more capability and built so reliably than the Atari Eights.  But sadly, Atari always treated their computers as an afterthought.   Game machines were their bread and butter.  Funny enough, they outlasted Commodore (by a bit).  Marketing always was on the cheap and an afterthought.   Tramiel just extended it a a few years. 

[Faicuai]

3 hours ago, slx said:

The 800XL was supposed to be expandable via the PBI connector (and 1090 box).

 

Indeed.

 

But the proof is in the pudding: have you noticed that, not even TODAY's modern-day HW expansions actually use PBI (!?)

 

With the exception of a couple coming from tf_hh (diagnostics & RAM expansion), and the Incognito board, sitting elegantly on 800's Personality Slot (#0), and delivering a functional / power-passing PBI bus, at the expense of killing (in its current implementation) all those must-have expansions like BitD, Franklin Austin, none of which (never needing a PBI bus to operate), ever saw the light of day on the XL/XE platform... and much less on the incoherent-looking, hard-to-place 1090 (!)

 

I think it says it all...

Edited May 12, 2020 by Faicuai

[+DrVenkman]

2 hours ago, gilsaluki said:

Marketing always was on the cheap and an afterthought. 

I don't think that's the case. Atari, Inc. threw money at marketing! The beautifully illustrated and photographed catalogs, the incredible store kiosks and demo carts! Hell, they even developed a freaking interactive LASERDISC (!!!) system for sales too. I'm also sure Alan Alda didn't do all those commercials for cheap, being a household TV star name and coming off a long and very successful TV run on M*A*S*H. 

No, to me the issue was always a matter of focus and direction, not money. Atari couldn't decide if the A8 was a games platform, a professional computer system, a home computer system, or some nebulous combination. Of course, in 1978 the idea of a personal or "home" computer that wasn't some kludgy S100 thing was still pretty new and unfamiliar to the average consumer. No one quite knew how it all was going to shake out; what uses people would have or want to have for these things, and the technology underpinning it all was literally changing yearly. 

[evilmoo]

6 minutes ago, Faicuai said:

But the proof is in the pudding: have you noticed that, not even TODAY's modern-day HW expansions actually use PBI (!?)

That's because SIO and the cartridge slot have not changed all the way from 400/800 -> XEGS.  If you design a cartridge or SIO device now, you design it once and it works everywhere.

If you want PBI on the 400/800, you have to make it an internal card.  If you want PBI on the XL, you need a 50 pin header.  If you want PBI on the XE, you need a bizarro adapter board to merge the cartridge slot with the ECI connector.

 

Additionally, you either need to find 5V parts to make all of your PBI device, or you need to level shift down from 5V to 3.3V or whatever your modern part set requires.

If there had been a standardized card slot (like there was supposed to be in the 1090), a lot of the circuit engineering work would not be so painful now.

[ACML]

I think the 1200XL was Atari's attempt to reduce production cost compared to continued production of the 800.  It also armed the Atari 8-bit with a 64K flagship.  Did they know the C-64 was just around the corner?   At the original price tag of $899, Atari had not yet realized that price would decide who would conquer the casual home computer market.  I also think Atariwriter came out a year or two too late.  It took something like Atariwriter to make just about everyone want/need a home computer.  Once Atariwirter came out, my sister, cousin, younger brother all wanted and bought an Atari home computer.  By then it was too late, the C-64 had an all out war for market share and won it all. 

 

As to the PBI, I never really cared except for the time a owned a 1meg MIO.  It was cool, but even then it still felt like a gimmick.  I mean, it was fun for a while, then you realized that you really didn't need it.  Same goes for 256K of RAM.  There was really nothing Atari could do to survive once Tramiel started selling C-64s for $199.  I was saving for a 1450XLD, but Atari would have lost millions on its production.  Atari had a small window (1979 - 1982) to assert itself as the one to bring home, but they just couldn't capitalize on the opportunity.  I say this as Atari was out selling Apple by 2:1 in 1981.  I attribute that impressive market share on the decision to sell the 16K 400 for $399 (May 1981).  If the C-64 never existed, maybe we'd all be buying 1450XLD's and 1600XL's on eBay today.

 

Edited May 12, 2020 by ACML

[Faicuai]

1 hour ago, evilmoo said:

That's because SIO and the cartridge slot have not changed all the way from 400/800 -> XEGS.  If you design a cartridge or SIO device now, you design it once and it works everywhere.

If you want PBI on the 400/800, you have to make it an internal card.  If you want PBI on the XL, you need a 50 pin header.  If you want PBI on the XE, you need a bizarro adapter board to merge the cartridge slot with the ECI connector.

 

Additionally, you either need to find 5V parts to make all of your PBI device, or you need to level shift down from 5V to 3.3V or whatever your modern part set requires.

If there had been a standardized card slot (like there was supposed to be in the 1090), a lot of the circuit engineering work would not be so painful now.

The format itself of the physical interface is NOT a real issue. The SPECS of the bus, however, are, and that does not seem to have changed either, other than enabling power pass-through on the XE (which the 800 XL DOES NOT pass, forcing you to have YET another power-supply cable-jungle for the 1090 !!!)

 

Furthermore, NOT having an on-board internal bay is the most serious offender, here, which makes implementing the PBI expansion-chain a real kludge. Let's just face it.

 

And here is the proof, again: finally built my Card Edge-to-IDC (50 pin) adaptor and gender conversion, for testing PBI devices, cross-platform (works like a charm for 800 / Incognito):

 

 

 

The best part of this is that, on the 800, you can run a dual internal / external setup, without a problem, and retaining expansion-bay top cover without any changes its plastic shell.

 

 

 

Edited May 13, 2020 by Faicuai

[+slx]

I don’t think there was a way to build a computer that would be able to compete with the C64 pricewise and have a chassis that would allow for expansion. I doubt ECI had any reason but to save a few cents off each PCB, otherwise it would not have made sense to change that port again. 
 

Don‘t forget C= did not price the C64 to make a profit, just to break even after the first round of price cuts. The money was made with peripherals. 

[Sugarland]

Atari should have had a 'high end' machine early on built more like an Apple II or PC with slots and similarly NO RF.  It didn't need a super expensive case, just something that functioned.

 

Looking at my PC today, it has only one expansion in it, a video card. All the other stuff is integrated into the motherboard including USB, ethernet, sound, hard drive controllers, etc. Well it does have RAM installed too..  Point is the XL's weren't a bad idea, to integrate all former expansions/cards onto the motherboard. Modern PC's still do the same. It's just that it was locked in, otherwise SIO.

 

There is a misconception that cheaper=more competitive. Not always true. PC's and Apples cost more but still sold well. It was also about usefulness and return on investment.

 

[Sugarland]

12 hours ago, TemplarXB said:

My understanding of the intent of the original 8-bit was to have internal expansion like the Apple II line, but the FCC regs resulted in it being encased in aluminum and using the SIO port being developed.  

 

After the FCC regs were changed, why didn't Atari design the XL line with that internal expansion?  Did the priority change to cost cutting, or was that the intent of the 1090XL?

 

Only because Atari wanted RF video coming out of the machine it fell under the FCC regulations. That meant there could be no slots/cards that were externally accessible like a PC or Apple. Apple II's did not output RF. Big mistake for Atari (marketing?) to mandate RF on all models.  FCC relaxed the regs a few months after the 800's shipped.

 

[BillC]

48 minutes ago, Sugarland said:

Looking at my PC today, it has only one expansion in it, a video card.

Many modern PCs also have the video integrated into the motherboard or CPU, adding a video card is like installing a SOPHIA or VBXE.

[MEtalGuy66]

God, this question has been asked a million times in here. I bet I've written 50 pages on it personally.. 

 

The 1400XL and 1450XLD machines were the last to have a buffered data bus between the CPU and the hardware. The original PBI expansion bus and the 1090XL both depended on this design feature/characteristic in order to have some semblance of stable operation. When it came time to actually bring machines to market, in the early 80s, many "cost cutting measures" were implemented in the design before the actual production machines (1200xl,600xl,800xl)  came to market. One of these was the REMOVAL of the buffers on the data bus. Instead, all hardware bus traffic would be directly synced by the PHI2 signal from Sally. This is fine as long as your bus capacitance is finite and known.. Which it can not be, when there's a possibility of plugging many devices directly onto the bus, as in the case of the PBI (and cartridge too, actually). So what we ended up with is a machine that, although the physical connector was still there, could never run an architecture of parallell bus expansion reliably. Atari essentially made a decision to abandon the possibility of bus expansion in favor of getting the XL series to market at a price point that they felt the market would bear..  Sad but true.. This is why the only device ever actually sold by Atari to utilize the PBI connector was the 64k upgrade for the 600XL. 

 

Aftermarket developers, however, chose to utilize the PBI bus for various interfaces such as the Supra, ICD MIO, and CSS Black Box..  But if you know anyone who has owned these, they usually require "fine tuning" of the bus timing on the machine in order to operate with any degree of stability. In fact, the more things you hang off of the data bus in general, the more phi2-to-bus timing gets skewed by increased capacitance (and thus propagation delay), and it will cumulatively cause the machine to become irrattic/unstable/unoperational unless appropriate hardware measures are taken to compensate and bring the bus timing back into valid sync with PHI2. 

 

This includes internal expansions that plug into chip sockets, cartridge based expansions, and PBI devices.  The more crap you hang off the bus, the worse it gets. And once you get a heavily expanded system like that "tuned" for stable operation, any major changes (such as removal of a cart or PBI device) often throw it "out of whack" the opposite way because now you've reduced bus capacitance/propagation delay. So you've got to "retune" again. 

 

This "tuning" can be done in a variety of ways. These range from the simple addition of capacitors on bus lines to slightly increase delay, to shortening/lengthening of entire bus interconnect cables, to complex compensation circuits. MANY hardware people in this hobby have devoted untold hours to this problem. Unfortunately, short of redesigning the XL/XE machine spec, there is no universal fix. And without that, a 1090XL like device, while not impossible, is really "asking for it" in terms of installation specific troubleshooting and "hair pulling".. 

 

 

Whew!

I hope this is the last time I have to explain that. But it probably wont be.. 

 

Short answer to your actual question:

regardless of which type of interconnects you use, internal, external chain, external back-plane ala 1090xl...   For extensive parallel bus expansion, you need reliable bus timing.. The original 1400xl/1450xld designs had this. The apple II has this..  Machines like the production atari XL/XE, the commodore 64, etc.. Do not..  

Edited May 13, 2020 by MEtalGuy66

[_The Doctor__]

The 1090XL did re incorporate buffer isolation in one of the cable revisions, I'd bet Curt would have information on the cable... I'm thinking a girls name like julia or something attached to the nomenclature... Just best to put the buffer back on the mobo though...

Edited May 13, 2020 by _The Doctor__

[MEtalGuy66]

5 hours ago, _The Doctor__ said:

The 1090XL did re incorporate buffer isolation in one of the cable revisions, I'd bet Curt would have information on the cable... I'm thinking a girls name like julia or something attached to the nomenclature... Just best to put the buffer back on the mobo though...

Not even remotely the same thing. 

[Faicuai]

On 5/13/2020 at 2:52 AM, MEtalGuy66 said:

God, this question has been asked a million times in here. I bet I've written 50 pages on it personally.. 

 

The 1400XL and 1450XLD machines were the last to have a buffered data bus between the CPU and the hardware. The original PBI expansion bus and the 1090XL both depended on this design feature/characteristic in order to have some semblance of stable operation. When it came time to actually bring machines to market, in the early 80s, many "cost cutting measures" were implemented in the design before the actual production machines (1200xl,600xl,800xl)  came to market. One of these was the REMOVAL of the buffers on the data bus. Instead, all hardware bus traffic would be directly synced by the PHI2 signal from Sally.  (..)

 

Just making sure I get this right:

 

When you mention "buffered data", do you mean "buffered PHI-2"?

Edited May 14, 2020 by Faicuai

[+mytek]

7 minutes ago, Faicuai said:

 

Just making sure I get this right:

 

When you mention "buffered data"do you mean "buffered PHI-2"?

I believe he is referring to Octal Bus Transceiver chips (example: 74LS245). It would have taken 3 of them to fully buffer the data and address lines. Buffering Phio2 would have taken a separate gate, so I can't speak for if that was also done.

 

[Faicuai]

7 minutes ago, mytek said:

I believe he is referring to Octal Bus Transceiver chips (example: 74LS245). It would have taken 3 of them to fully buffer the data and address lines. Buffering Phio2 would have taken a separate gate to buffer, so I can't speak for if that was also done.

 

Interesting... plenty of extra chips for achieving that (are those present / visible on the 1400XL or 1450XL MoBo's?)....

 

BTW: what's your clinical opinion of this:

 

 

Buffered? Unbuffered? Crooked?

 

Edited May 14, 2020 by Faicuai

[+mytek]

2 minutes ago, Faicuai said:

Interesting... plenty of extra chips for achieving that.... BTW: what's your clinical opinion of this:

 

 

Buffered? Unbuffered? Crooked?

 

Usually that's ringing which can be the result of transmission line (traces) reflection caused by a poor impedance match, or normal ground bounce that our systems are notorious for. Buffering can help, but the best practice is to beef-up the ground and power planes if its a 2-layer board, or better yet go to a 4-layer board that has mid-power planes like what was done on the 1450XLD (and possibly the 1400XL), and is also the route I chose for the 1088 series motherboards. Buffering is mainly needed and/or desired when you go off board with your system bus signals, such as what the PBI was meant to do. Sometimes adding series inline resistors on each end of the transmission line can help to absorb the reflections, which can be used with or without buffering.

 

Also the noise you are seeing, will become worse the farther away you place the scope probe GND lead from what you are measuring.

 

[Faicuai]

1 hour ago, mytek said:

Usually that's ringing which can be the result of transmission line (traces) reflection caused by a poor impedance match, or normal ground bounce that our systems are notorious for. Buffering can help, but the best practice is to beef-up the ground and power planes if its a 2-layer board, or better yet go to a 4-layer board that has mid-power planes like what was done on the 1450XLD (and possibly the 1400XL), and is also the route I chose for the 1088 series motherboards. Buffering is mainly needed and/or desired when you go off board with your system bus signals, such as what the PBI was meant to do. Sometimes adding series inline resistors on each end of the transmission line can help to absorb the reflections, which can be used with or without buffering.

 

Also the noise you are seeing, will become worse the farther away you place the scope probe GND lead from what you are measuring.

 

THANKS!!

 

I ask because that is Pin #31 from Incognito's PBI port (PHI2 signal). Sampled-ground is GND next to it (two pins available).

 

Considering the sharp-rise and sharp-fall on the edges, it seemed (to me) like it was BUFFERED.... but I did wonder about that Rocky-Mountains overshoot and Lunar-craters tails... for one second I thought that my beloved Fluke-99B was faltering (I think it is the piece-of-shit chinese Card-edge connector that I crimped to the IDC Ribbon... compared it to Jurgen's, mine seems pure crap). I will try to sample PHI2 directly from anywhere available in SLOTS 0 to 3, and compare with what I got outside.

 

As a reference, I sampled PHI2 directly from my 800XL PBI bus, and exhibits the same sharp rise and falls, but clearly smoother crests and tails.

 

Edited May 14, 2020 by Faicuai

[Faicuai]

On 5/13/2020 at 1:58 AM, Sugarland said:

Atari should have had a 'high end' machine early on built more like an Apple II or PC with slots and similarly NO RF.  It didn't need a super expensive case, just something that functioned.

 

Looking at my PC today, it has only one expansion in it, a video card. All the other stuff is integrated into the motherboard including USB, ethernet, sound, hard drive controllers, etc. Well it does have RAM installed too..  Point is the XL's weren't a bad idea, to integrate all former expansions/cards onto the motherboard. Modern PC's still do the same. It's just that it was locked in, otherwise SIO.

In all fairness, modern PC architecture has evolved into a gradient of options in which, the most "integrated" seem to be the most cust-cut too!

 

But that gradient (and main strength of PC architecture) is also a reflection of its ability to meet the markets and needs being addressed (that's why it won)... without mentioning the staggering pace at which Intel itself has gobbled up system components INTO its processors themselves!

 

In my case, those expansion slots have saved my sorry ass many times, and I still use them actively today. Thanks GOD the Z840 has a few to spare (my QAM-capture card, Firewire, Video and NVME drive are all plugged in there):

 

Edited May 14, 2020 by Faicuai

[Faicuai]

6 hours ago, Faicuai said:

THANKS!!

 

I ask because that is Pin #31 from Incognito's PBI port (PHI2 signal). Sampled-ground is GND next to it (two pins available).

 

Considering the sharp-rise and sharp-fall on the edges, it seemed (to me) like it was BUFFERED.... but I did wonder about that Rocky-Mountains overshoot and Lunar-craters tails... for one second I thought that my beloved Fluke-99B was faltering (I think it is the piece-of-shit chinese Card-edge connector that I crimped to the IDC Ribbon... compared it to Jurgen's, mine seems pure crap). I will try to sample PHI2 directly from anywhere available in SLOTS 0 to 3, and compare with what I got outside.

 

As a reference, I sampled PHI2 directly from my 800XL PBI bus, and exhibits the same sharp rise and falls, but clearly smoother crests and tails.

 

Ok, had to really check this out and this is what I am getting from Slot-3's PHI2 (and GND) pins:

 

 

With more temporal resolution:

 

It's about a wash as far as I can see "outside" (on pin 31 of Incognito's PBI).

 

Still wondering if that waveform is (or not) what you would expect from a buffered PHi2.