Expanding KIM-Style 6502 Single Board Computers
Part 1 of 3:
Hal Chamberlin
Undoubtedly the most successful single board computer ever has been the KIM-1 made by MOS Technology (now Commodore). When introduced it apparently had just the right combination of features and price to attract tens of thousands of users. More recently of course the SYM-1 from Synertek and AIM-65 from Rockwell have incorporated numerous additional features into the same self-contained single board computer concept. Fortunately for users, all three of these machines are quite similar in their electrical characteristics.
Sooner or later however all computers need to be expanded and these single board machines are no exception. Although the SYM-1 and AIM-65 can be expanded somewhat merely by plugging in additional memory chips, the maximum limit is only 4K bytes of programmable memory. Thus additional boards are required for substantially increased RAM, ROM, or I/O capability. Recognizing this fact, the computer manufacturers as well as a number of independent accessory manufacturers have designed and brought to market a wide variety of expansion boards for the KIM, SYM, and AIM computers.
In most cases just having expansion boards available is not enough; there must also be a motherboard offered to plug them into since these computers have no on-board bus and slot sockets of their own. To date the computer manufacturers and independents have selected no fewer than four distinctly different ways to do this. First on the scene of course was MOS Technology who offered the KIM-4 expansion motherboard which mated with their KIM-2 and KIM-3 expansion memory boards. The bus presented by the KIM-4, which is called the "KIM-4 Bus", is in many ways similar to the bus presented by the computer itself as its own expansion edge connector. The primary difference is an altered pin assignment which is basically a one pin shift from the expansion connector assignments. This apparently was done to provide additional ground connections. Since then, independent manufacturers have also offered KIM-4 style expansion motherboards although there are important differences from the original KIM-4 (see Compute issue #3).
Shortly thereafter, as soon as the KIM's popularity became known, other independent manufacturers offered expansion motherboards which presented an S-100 style bus to the expansion boards. The primary advantage of this approach is that the user is not restricted to using expansion boards designed specifically for KIM-style machines but instead can choose from hundreds of S-100 compatible boards designed for 8080 based systems. Unfortunately many of the more sophisticated S-100 boards such as large dynamic memories, graphic display interfaces, and disk controllers could not be used because of substantial timing differences between 6502 and 8080 style microprocessors.
Late in 1977 Micro Technology Unlimited introduced a motherboard and card cage for the 6502 based single-board computers. The motherboard is little more than 5 edge connectors wired in parallel with one for the computer and the other 4 for expansion boards. The bus presented is the same pinout as that of the processor's expansion connector. The main advantage of this technique is the low cost and compact packaging afforded by the elimination of bus buffers. In addition, expansion boards compatible with this bus may be easily connected directly in parallel with the expansion connector if for some reason the motherboard is not desired. The main disadvantage is that the number of expansion boards is limited to four by the small drive capability of the computer's own bus.
Recently Rockwell has introduced its expansion motherboard which essentially presents and Exorcisor bus to the expansion boards. Motorola originated this bus for use in their Exorcisor microprocessor development systems. Rockwell also uses the Exorcisor bus in their system 65 development system. The advantage of this method, at least to Rockwell, is avoiding the need to develop new expansion boards just for the AIM-65. To users the biggest drawback of the Exorcisor bus probably is the lack of reasonably priced boards to plug into it.
All four of these techniques are quite viable methods for expanding KIM-1, SYM-1, and AIM-65 single board computers and each has a broad base of dedicated users.
Mechanics
All three of the single board computers are intended to simply rest flat on a tabletop using the several quarter-inch high rubber feet provided. Although not the most beautiful thing in the world, it works well in many cases and is certainly inexpensive. In situations where better appearance is desired or small children are present, there are vaccum-forced dress covers available that simply slip over the computer board hiding everything except the display and keyboard.
The KIM-4, S-100, and Exorcisor type expansion motherboards simply extend this board-on-the-table concept. Typically the expansion motherboard is roughly the same size as the computer board and plugs straight onto its expansion connector. On the motherboard are perhaps a dozen integrated circuits for address decoding, bus buffering, and voltage regulation. The majority of the space however is taken up by 4 to 8 edge connectors which form the "slots" of the expansion bus. When plugged into these slots, the expansion boards assume a vertical orientation.
A system expanded this way uses a large amount of additional table space, and in the case of th KIM, it is useful space to the left of the computer. The assembly of interconnected boards is also rather fragile and certainly not portable unless dismantled (most people would probably bolt the computer and motherboard to a sheet of plywood or plastic to avoid this). In particular a stray elbow can do considerable damage if a board is knocked out of its slot during operation. Unfortunately the available plastic dress covers do nothing to protect the added motherboard or expansion boards.
Another approach that has been slowly gaining acceptance is to place the expansion boards underneath and parallel to the computer board. Thus the expansion motherboard, which ties all of the boards together, is vertical. In order to hold this assembly of boards together, an aluminum frame with card guides is typically supplied and the motherboard is attached to an opening in the frame. Figures 1 and 2 show the KIM and AIM versions respectively of Micro Technology's implementation of this concept.
The advantages of this configuration of course are reduced table space requirements and greatly increased protection for the expansion boards. The entire assembly of computer and boards is now one portable unit with only the power supply left over to worry about. The computer board is still exposed however. Probably the only potential disadvantage is that the computer's keyboard has been raised about 4 inches above the tabletop.
Electronics
There are electronic factors to consider as well when expanding a KIM, SYM, or AIM computer. In order to minimize cost, complexity, and power consumption, all three of these single board computers are designed without buffers between the microprocessor chip and the expansion edge connector. The KIM-1 went one step further and omitted part of the address decoding circuitry as well. The lack of buffers means that the expansion bus presented by these computers has a DC drive capability of only one standard TTL load, or equivalently, 5 low power Shottky loads. The AC drive capability depends on the desired signal risetime. For bus operation at 1mHz, a total of approximately 25 "connections" at 6pF each can be driven. A connection here is defined as a gate input, disabled tri-state output, or MOS input (which does not contribute to DC loading).
Compared to other bus-oriented computers, such as S-100 machines, this does not sound like much of a bus at all since these machines typically have a drive capability of 30 standard TTL loads (74 series) or nearly 150 low power Shottky (74LS series) or over 200 low power TTL (74L series) loads. In fact, the original advertising for the MITS Altair computer boasted an expansion capacity of "over 200 boards". While this may have seemed necessary when using MITS's 1K memory and single port I/O boards, 10 slots is ample for even the largest S-100 setup when using today's dense memory and peripheral interface boards.
Over the years, experience has shown that several factors other than sheer driver power limit the number of boards that may be connected to a bus.
The most serious of these is crosstalk noise between the bus address/data lines and the various bus control lines. This noise arises when large numbers of address and data lines change state simultaneously, which is a common occurrence. The fast voltage risetimes (around 5NS with the popular 8T97 drivers) and 50MA or greater surges of current along each changing line couple electrostatically and magnetically to other lines in the bus and on the expansion boards themselves. Longer busses and more boards plugged in gradually increase the crosstalk until noise on the control lines causes false triggering of memory and I/O boards and thus system failure. So severe is this problem that early S-100 systems would fail to operate even before the 16 board capacity of a single cabinet was reached.
A related, but much less severe problem, is signal reflection from the ends of the bus lines, which after all, act like transmission lines. This effect becomes significant when the signal transmission time exceeds about 1/2 of the signal risetime. At 1.5NS per foot with a 5NS risetime, the bus would have to be two feet long before termination was required. The apparent success of bus terminators sold for S-100 systems is probably due to their reduction of signal swings (the logic 1 level is limited to 3 volts and floating bus lines are pulled to 3 volts) which in turn reduces crosstalk noise.
From the author's experience in designing a large, fast bus oriented system (specifically the A. B. Dick Magna SL four terminal full-page word processing computer), there are three ways to solve bus noise problems. One is to thoroughly shield the bus with a full-width ground plane, or ideally, a three-layer motherboard with data/address on one side, ground in the middle, and control signals on the other side. This solves noise coupling on the bus but not on the expansion boards which in turn must be carefully designed to minimize their own crosstalk. This technique was used in the Magna SL machine because of speed requirements.
Another technique is to use filters and delays on the control signals obtained from the bus in order to reject narrow noise pulses. This technique can be extended to deal with any kind of noise problem at the expense of system speed and is the one typically used with minicomputers such as DEC PDP-11's and Data General NOVA's.
The third technique attacks the source of the noise, namely fast risetimes and large current surges, by using a low power bus. With liesurely risetimes of 50 to 100NS and drive capabilities of less than 10MA, such a bus is virtually noise-free and quite fast enough for normal microprocessor operation. This technique, coupled with some attention to groundplane shielding, is most applicable to unbuffered KIM/SYM/AIM expansion busses.
The foregoing is not meant to imply that all of the buffered expansion motherboards available for the KIM, SYM, and AIM computers are racked with noise. In fact, their bus length and number of slots is generally small enough to keep noise at tolerable levels. The major point is that high power drivers and indefinitely expandable busses do have drawbacks of their own.
The real question at this point then is: How many expansion boards can the unbuffered microprocessor bus drive before becoming overloaded? The 6502 microprocessor is rated to drive slightly more than 1 standard TTL load (equivalent to five low power shottky loads) on its address and data busses while most of the RAM's and ROM's tied to the data bus can drive two standard TTL loads. The 6520, 6522, and 6530 I/O chips have the same drive capability as the microprocessor. Thus in general the answer is at least four boards provided that the expansion boards themselves buffer the bus such that only one low power shottky load (.36MA in the zero state) is presented to the bus by the board. Many boards on the market and particularly those designed for an unbuffered bus do this. Actually, any well designed board would be expected to buffer the bus in order to provide clean signals for the remainder of the board logic. The reason that only four boards can be driven instead of five is that some of the address lines are loaded by a low power Shottky decoder IC on the computer board itself.
Next time: The Great Experiment