WHERE OLD COMPUTERS ARE WELCOME
In our last “Gotcha” a few manufacturers were taken to task, and fewer still were praised for the quality of then- products and documentation. This time we take a brief look at the familiar Altair and IMSAI chassis, and a long look at the very interesting Poly 88.
I refuse to revive the old Altair vs. IMSAI debate. As everybody now knows, the old Altair power supply was feeble. My Altair worked fine after I had replaced its supply with a custom 40 amp at 8 volt wonder. [In an external box that should never land on your foot.] I actually liked the Altair case better than IMSAI’s, and am glad to see MITS has carried the design over to its new machine. Just two screws and the top slides off. If the screws are omitted, the case is just as strong to top loading. On the IMSAI there are four load bearing screws which I could never get at because something was always sitting alongside the computer. If the screws are omitted, the top sits rather low. Not good.
Since I get letters asking: Yes, the heavy duty supply on the IMSAI is excellent. I haven’t tested the new MITS supply, but it looks good. I don’t think they’ll make the same mistake twice. After all, if they’re smart enough to make a computer
I find a serious flaw in the IMSAI front panel. Those big paddle switches that make the IMSAI look sort of like a PDP- 1 1 have a small space between their tips. IMS should take note, as it prevents errors, and still makes it easy to hit two at once on purpose.
The Poly 88’ s almost S-100 bus (hobbyist bus, Altair bus, I could care less) has the best switches of all: there aren’t any. Polymorphic Systems, in Goleta (rhymes with “Lolita”) California makes this very unassuming little box that does a lot of things right, which the bigger names (with bigger boxes and price tags) are doing wrong. Not that Poly is perfect. My corrections to their manuals were extensive and numerous.
But they listen harder. For example, when I called up IMS with a long list of carefully annotated errors in their manuals, they put me off, promised to call back, never did, put me off when I called again, etc. In the end, my careful documenta- tion of their manuals did them no good at all, and frustrated me. Polymorphic Systems listened, sent me extra manuals so that I could send them mine with corrections and the like. They’re not dumb; they’ve got a proof reader working for free. It is my opinion that manufacturers should hire a proof reader before sending manuals out, but the way that is now used is cheaper and only has the drawbacks of having a few hundred frustrated customers out there. And, they get hun- dreds of phone calls of the form, “Where does R21 go?” (Maybe the phone company is behind the bad manuals.)
The Poly-88 system — which has replaced my Altair 8800 and my IMSAI 8080 — has but two controls on the box. An on- off switch with a power-on indicator fight, and a reset button with a halt fight. That’s all you get; that’s all you need. It surely doesn’t look impressive. Sort of like a toaster in size and shape. The Poly is by far the easiest of the S-100 bus computers to build. The backplane and power supply are all on one well-designed motherboard. The only wires leading to it are from the transformer and the front panel button via two Molex connectors. It is all very neat with almost no point-to-point wiring. Someone was thinking when they designed this one. To take the backplane/motherboard out, just pull the two connectors and undo six easy screws (which go into captive nuts on the board, nothing to get lost inside). Have you ever tried to take out an IMSAI backplane? More screws than an X-rated movie, and then there are wires screwed onto the board. Dumb.
For some reason Poly’s tiny little case requires eight screws! A bother. The kind of thinking that went into the electronic design was absent when the case was created. It is on such small shoals that great ships are wrecked. There are more bolts than slots. There are 5 slots. Is that enough? Let’s see, one for the CPU, one for the video board (it’s a dandy), one for a ROM board to hold something comfy, like BASIC, and, say, a 16K RAM board. With a Pixie Verter and a keyboard you’re ready to go into any American household with a TV, plug in and program away. And you’ve a lot left over — another 16K? Voila! (or ’Cello! for that matter) we have a 32K computer with serial port, cassette interface, video interface, and software aplenty, power supply and on/off switch tucked away in our viola case. I must mention that a two-port serial interface is built into the CPU, and the RS- 232 and cassette interface cards are snug against the connec- tors in the back of the case. That is a lot of computer in an itty-bitty base.
That isn’t enough? You say you have a pile of old 4K boards? I do. Poly has the Idea of the Year (at the rate we get new ideas in this business, maybe the idea of the month): At one side of the chassis the backplane terminates in a male S-100 bus connector. On the other side of the chassis is a female connector. Aha. You can buy another chassis, put it alongside the one you already have, sort of nudge them together and guess what. Nope, you don’t get a litter of 4040’s. You push them together and you get . . . a ten-slot chassis. Actually, eleven slots ’cause you can put a card into the end and let it stick out instead of yet another chassis. This is useful and saves need for an extender board. The power supplies are separate and not bussed together. Like the IMSAI this machine has a substantial power supply. They rate it 6 amps at 5 volts. 1 loaded it down with 9 amps worth of boards and a length of nichrome wire and it was still putting out 8.2 volts. The point is, as you expand the chassis, you expand the power supply as well. Each expansion chassis costs $155. Takes two or three hours to build. Polymorphic Systems forgot (so typical of manuals) to tell you that R6 should be omitted on slave chassis. So I tell you.
Before I find some drawbacks (I am not in the employ of Poly) there was an advantage to the multiple chassis that I hadn’t suspected when I ordered the miniature monsters.
When working on a board, sometimes it’s handy to have a program sitting around, but it disappears when you turn off the power to make a change on the board. But with the separate chassis idea, you put the CPU and memory in one box, the board under test in another. Just turn off the one chassis to remove the board, make the changes, replace and turn on the chassis. Program still there, testing continues.
Not all is peaches and cream . . . The diodes supplied with three of the four I’ve built— you know, the little ones for the plus and minus 1 6 volt supplies— were small signal diodes instead of power rectifiers. The smoke test lived up to its name. A quick trip to Radio Shack (it was Sunday) got me a handful of diodes of the right rating. 20 for $ 1 .98 or something like that. The first one I tested was bad. But there were 21 in the bag! The other 20 were good, so no complaint there. But, dear reader, always test.
The assembly instructions were terrible. There were as many many errors as the other brands had [see DDJ, Vol. 1 , No. 8] .
Lots. They say they’re coming out with a new manual. If it’s any good, I’ll probably write it up. Volume II of the instructions, however, is on the side of the Angels. This is the clearest manual on the 8080 instruction set I have ever seen.
I leave it on the living room floor for people to pick up and read. It’s that good. Someday I may even get a coffee table for it. So if you want to learn the 8080, get that manual. Maybe some magazine will serialize it (in good serial style:
“Last month our hero got saved from the evil Dr. Halt when an interrupt arrived in the nick of time . . .”).
What is life like without a front panel? Sheer joy, my friends, sheer joy. You can deposit, examine, single step, everything. You can do it in style, from a keyboard. When you single step you see not only the address and the con- tents (and in hex, not in binary [less than joyful to those who prefer octal]) but you also see: the accumulator, the flags, the B-C register, the D-E register, H and L, the program counter and the stack pointer.
And you also see the first eight bytes of the stack, the location the PC points to, the next seven bytes thereafter, and the eight bytes pointed to by each of the B, D and H registers. It certainly beats lights; it also beats the hex displays found on a few other machines.
I hate to say this, but the Polymorphic advertisements understate the advantages of their machine. Too bad for them. It should be clear that the conventional front panel is a holdover from an earlier era. It’s too bad that those lights and mysterious switches appeal to so many of our computer cult. Like those famous tailfins on cars, it im- presses the neighbors, but doesn’t make the machine run better. Of course this goes for all the ROM replacements for front panels. Having both a front panel and a ROM monitor is fine; you just have to pay for it.
For just under $600 you get, with the Poly system, the bos, power supply, video board, the monitor in ROM, 512 words of RAM, room for 3K more of ROM, and all the sockets you need for the ICs. Of the S-100 bus machines, it is the only one where the minimal system has to do real programming. (O’ course, you have to add a keyboard and a TV monitor — but nobody includes them for the price.) Enough free advertising for Poly. I am not so much interested in selling computers for them as I am in seeing my computer cousins not wasting their time flipping switches and misread- ing lights. Any system (as I said) with HEX display is better for a human being than the same system with a BINARY display, and of the S-100 systems available this week, the Poly will get more done per your hour than any of the others — that’s building hour, programming hour, and even earning hour. Other manufacturers, if you’ve got a better system, tell me about it. Don’t bother, unless you use the S-100 bus (so we can go to others than just you for add-ons). But, do tell me if you’ve got something really different like 8K for $50.
A disadvantage of not having the conventional front panel (after all this, I do know one disadvantage) is that the CROMEMCO Bytemover program won’t work. It needs switches. You can get a parallel port and eight switches and wire it up for port address FF, but that’s a bother. So I called up CROMEMCO (if their documentation had been better this would not have been necessary) to find out how to write a 2707 EPROM without their program. The method, they told me, consists of writing each PROM in its entirety, from beginning to end, a number of times. Say a hundred to three hundred times. “How many times does the Bytemover write each PROM?” I asked. “Thirty-two,” I was told. So I wrote a little program that wrote the stuff into all the PROMS 255 times (you can guess why). It worked. I sold my PROM containing the Bytemover program. [Another disadvantage of no switch register is that it means the user has no sense
There are a few devilish “gotchas” in the Poly system. The first problem showed up in trying to run MITS BASIC (duly purchased from MITS). Since the monitor likes to reside in low memory, and so does BASIC, there was a conflict. The solution: Poly provides a jumper to cut, and one to add to allow you, under program control, to switch back and forth between the processor-board memory (3K ROM, 14K RAM) and other memory having the same address. In my case the program that copied BASIC from PROM resided at location E400, so I used the following program to get BASIC up:
3E 20 D3 04 C3 00 E4
The seven bytes say: put 20 (hex) into the accumulator, then send it to port 4. Sending the 20 to port 4 turns the onboard stuff off. The next three bytes of the program jump to E400 to get the program started. Now, of course, that little program is on the PROM.
The next problem was with the old MITS serial board.
I’ve always wondered why some serial boards have a crystal- stabilized clock of their own since they could just count down from the CPU’s clock signal which is on the bus. That’s what the old serial board did — it counted down. PROBLEM! The clock rate on the Poly system is a few percent slower than the MITS and IMSAI clock. So the old MITS serial I/O counts down to the wrong baud rate. You have to (as I did) rewire all the counter presets. You will have to calculate the proper values. Since my computer wasn’t up ’til this was fixed, I was glad to have my Model-T vintage HP-35 to do the necessary calculations. The newer MITS 88-2SI0 (a fine board in my book) does it right and has its own crystal (and works like a charm in my Poly, without modification).
And another, almost unforgivable error on the Poly: it is not quite an S-100 system. Sure, everything I tried with it worked except for that one board. But when a manufac- turer came by my place with a prototype of their new 1 6K board and plopped it into the Poly it didn’t work. A few “unimportant” outputs and inputs to the CPU were left off the bus. More importantly the WAIT signal is not on the bus. This let the memory know that the computer was in the HALT state, which the memory needs to know. This is not the place to go into that (gotta leave something for other articles). This particular device did not need the other signals but some new board might. The point is this: you are either on a standard or not on a standard. There is no in-between. Polymorphic Systems’ Poly 88 is not really an S-100 compu- ter. You have to ask first: does the board you wish to use with it require separate disables (address, status, data out)?
Does it need HLDA, INTE, WAIT? If it does, then the Poly won’t do. The disables are on the bus, in non-standard loca- tions, but they can only be disabled as a group (as required by DMA’s). One of the output signals such as WAIT can be fed through a spare buffer on the board to the bus. (For your information it’s IC 8, an 8T97. Because this is DDJ I’m sure the editor will permit the gory details.) Bus lines 22, 18, 23, 26, 19, 28 and 27 are not on the bus. To put WAIT on the bus, jumper pin 24 on the 8080 to pin 2 of IC 8, and pin 3 of IC 8 to bus line 27. I don’t know if I’ve missed something else that should have been on the bus. I called Polymorphic Sys- tems; they tell me that are putting out a list of differences, and plan to connect the WAIT as I suggest. It’s the least they can do.
Another problem with the Poly is that when a number of chassis are plugged into one another, the cooling, adequate with a single system, becomes inadequate. Not only does the system look like a toaster, My friend, Kent Strother, made a cardboard enclosure with a super quiet ROTRON fan (as per the IMSAI). Now even the regulators run cool. The secret: put the fan on top, sucking up, thus forced air aids natural convection. There are slots around the periphery at the bottom of the case, and all other openings are sealed, forcing the air to pass the boards and transformer. Not enouth attention has been given to air flow in the IMSAI or the ALTAIR 8800 which both have a lot of stagnant air spots even with the fan going. Kent also designed a cardboard case for the keyboard that used to he around naked. Call us the Cardboard Computer Com- pany. It’s cheap, in keeping with our homebrew budget, and if done carefully looks surprisingly good.
Second, a word for the Poly video board, but first, a word for the Processor Technology video board. The PT VDM is a top-notch piece of equipment. Their check-out procedure as you put in the chips is a classic of good manual writing. The VDM board I built worked perfectly. The Poly video board suffers from drawing current at the hairy edge of what the regulator can handle. What I like about it is the relatively fine graphics it allows: a 128 by 48 bit resolution. The graphics can be mixed with alphanumerics in any arbitrary way. It is a good use of that eighth bit that the ASCII code doesn’t require. Use of the Motorola MCM6571 AL character generator gives me upper and lower case Roman characters as well as the Greek alphabet and a gaggle of other special characters, including the entire official ASCII set, the square root symbol, etc.
I don’t know enough to write articles like this without some help from my friends so: thanks to Doug Wyatt, my constant colleague on microcomputers, Kent Strother for the cardboard craftsmanship, Steve Calebotta for finding the problem with the missing WAIT, and out editor Jim Warren for the phrase “Hidden Gotchas” that graces these articles [who plagerized it from Dave Wyland at Ratheon].
If you find any hidden, or just plain hanging out gotchas, send them to me. I’ll check them out and include them in a future article. You’ll get credit in the mag, and as much moola as I get. Zilch. [Ahhh but such glory and fame you get!]
