Introduction Most S-100 users have a few RAM boards in their system. Early systems often consisted of 4K, 8K or 16K boards. With the advent of 16K dynamic RAM chips in the early 80's 64K dynamic RAM boards became common. The most popular being the SD Systems ExpandoRAM series. When 16 bit CPU boards appeared most vendors utilized static RAM chips because of their easier interface requirements. 64K and 256K S-100 static RAM boards were popular and were very reliable. For example the BG Computer 256K RAM card was typical. By the mid 80's a few 1MB and 2MG static RAM boards were made by CompuPro and Macrotech. However these were expensive and few were actually made.
Today the scene is very different. Gigabyte dynamic RAM chips are common place and 512K X 8 bit static RAM chips are commonly available. (Higher capacity static RAM chips are also available but as far as I know not in DIP configurations).
Since I intend to build up my S-100 System to accommodate fast and memory hungry CPUs like an 80386 etc. I decide to build a high capacity RAM board. Eight 512X8, thirty two pin DIP chips should easily fit on an S-100 board and yield a capacity of 4MG. Four such boards will fully stock the S-100 bus with 16 MG of RAM. There are a few versions of such RAMs out there. I opted for the Alliance Memory AS6C4008 32 pin DIP chips. These are available from Jameco (part # 1927617). Just two chips will give you 1MG of RAM -- enough for any 8086/80286/DOS requirement!
Memory Board Logic The S-100 bus was designed around the 8080 CPU. This CPU has a bi-directional 8 bit data path, however because Ed Roberts wanted to control it via a series of front panel switches, the CPU data bus was broken out into separate 8 bit Data In and Data Out paths on the S-100 bus. This worked fine. RAM boards had their RAM chips connected to both data paths and I/O boards were likewise connected. It was a somewhat inefficient setup but it worked (all for a simpler front panel). The first diagram below shows the basic layout.
When 16 bit CPU's came along things got more complicated. All 16 bit CPU's have a 16 bit bi-directional data path. They can interface directly to RAM chips that are 16 bits wide or two 8 bit wide RAM chips. In the early 80's the latter type of RAM's were much more common. Two 8 bit wide RAM chips differing only in address line A0 (0 or 1), would be connected to the 16 bit bus. The S-100 Data In and Data out lines were utilized together as a single bi-directional 16 bit bus and were connected via buffers to the CPUs 16 data line pins.
Now if all you were going to use was such a 16 bit CPU that would be fine. But such a setup would not work with older 8 bit RAM cards or indeed with any I/O cards which also expect a split 8 bit interface.
The solution was simple and elegant and the heart of the IEEE-696 standard. For 16 bit systems the bus behaves as a 16 bit bidirectional bus. For 8 bit systems a bridge buffer on each RAM board transfers the data coming and going to the board over separate 8 bit data lines depending on whether it's address is odd or even. If we have 8 bit data coming to the board on an even address, it travels on the "Data Out" path and goes directly to the A=0 RAM bank. If instead the 8 bit data is destined for an odd address it arrives as before at the RAM board top buffer but then is dropped down to the lower A0=1 RAM bank via a bridging buffer.
If the 8 bit CPU wants to read a even address it activates this buffer in the opposite direction so the RAM A=0 bank data is shifted down to the S-100 data in lines. If the 8 bit CPU wishes to read am odd RAM address the A0=1 RAM bank data travels directly to the CPU on the Data Out bus.
The hardware logic to do this is quite tricky. You need to factor in if we have a CPU read or write, if the data is 8 bits or 16 bits wide and if the destination address is on an even or odd address line. (Fortunately no common 16 bit CPU transfers 16 bit data on an odd address line).
Now back in the mid 80's companies like CompuPro and Macrotech implemented this logic in ROM like chips called PAL's. Unfortunately they never published the code.
A First Prototype 4 MG Static RAM Board I wanted to build a S-100 RAM board that would serve me well into the future. Where I could use it with very fast and wide CPU's. I decided against Dynamic RAM because of the complexity of refreshing (particularly for Master/Slave CPU switching). Likewise I wanted to stick with DIP type chips. There are higher density RAM chips than 512K/chips we use here but they require SMT and are difficult to work with. The 32 Pin DIP Alliance AS6C4008 seemed like the way to go, (Jameco Part #1927617).
The hardest part was figuring out a PAL equivalent circuit using standard TTL chips. I came up with the following "7400" TTL layout. It takes a few chips as opposed to one PAL but it is fast and reliable.
The above logic interfaces to the high and low RAM banks as follows:-
Since this was a completely untested design I made a "pre-production" prototype board. The schematic of this board can be seen here. You will notice that the board accommodates both static RAM chips, Flash RAM chips and EPROMS (from 2716's all the way up to 27256's). Here is a picture of the board.
You will notice a "patch area" above the normal height of an S-100 board for further prototyping/repairs etc. Fortunately we did not need to add any chips.
The board works solid as a rock with an Intersystem's 4MHz Z80 CPU board (including its ability to address above 64K) and with my own IEEE-696 compatible S-100 8086 and 80286 boards -- later write-ups! I have not determined the maximum speed of the board yet but considering the chips are rated at 55ns, I expect the bus itself would be the limitation.
4 MG Static RAM Board Prototype-II With lessons learned from this board we made a second RAM only prototype board using the following schematic and diagram. We will do a separate EPROM/Flash board later. This is a somewhat simpler board since now there is no need for all the configuration jumpers for the many types of possible EPROM's. Here is a picture of this prototype RAM board:-
The Final 4 MG Static RAM S-100 Board The above board had worked without any problems for four months in two systems using CPM+ and CPM86 with verity of hardware configurations so I decided to go ahead and do a final commercial type board. The trace layout was optimized further before doing this. Here is a picture of the final 4 MG S-100 Static RAM board (with 4MG of RAM onboard).
Building The 4 MG Static RAM S-100 Board While at first the circuitry may seem complex it is really fairly simple. There are four components. The circuitry consists of:-
Select RAM board addressing Carry out 8/16 bit data transfers Wait state circuitry (if needed) The RAM chips themselves.
The above board works fine in all systems I have. This includes CPM+ and a 750K version of CPM86+. With 59 ns access times on the RAM chips clearly no wait states are currently needed. I have the circuitry there for a later 80386/80486 CPU board I will be doing.
Realizing that a number of others have expressed an interest in obtaining these bare boards and building a board themselves . I have below a step by step procedure as to help build the board. While the board is simple to build and debug, its clearly not for a "first timer". Very little hand holding can be provided, sorry this is not a commercial operation!
The trick in assembling S-100 boards like this is to assemble it in functional sections. Sure you can put everything together in one large bite and take a chance, but I have long since found in the end, a systematic step by step approach is faster.
First check add the board 5 volt regulator and two capacitors. There is an error on the silk screen on the board and in the schematic above. Both C1 and C3 should be 33uF capacitors (not .33 and .1uF). If you use tantalum caps, make sure the polarity is correct. All capacitors should be 10 volts at least but C1 should be at least 25V. Plug it into an S-100 extender card and check you have exactly 5 volts at each 5volt pin for each chip position.
Next add all sockets, filter capacitors and 470 ohm resistors (R1-R10). Insert the rows of jumpers for address selection (P49-P54). Repeat the above voltage check.
Next we will add the board address selection circuitry. This circuitry brings down the "4MG_RAM_Select" line when the correct RAM area (within the S-100 16 MG RAM address space) is selected. With 4 MG/board, clearly just 4 such boards would fill all our S-100 needs. However in many cases initial systems will be just Z80's (64K) or 8086 (1MG) RAM. So only one board is required and only a pair or RAM chips at that. This is what we will use here.
Here is a section of the Board RAM select circuitry:-
The board utilizes 74LS682's for RAM addressing. If your are unfamiliar with this technique click here.
The important point to remember is that both 74LS682's each have two sets of inputs (P0-P7 & Q0-Q7). The chips compare each P input against each Q input if all the Highs & Lows match up only then does pin 19 goes low. IC2(a) is for RAM in the first 64K. If connector P49 pins are jumpered to P53 pins and all SW3's are closed (to ground), ONLY when address lines A16 to A23 are zero will IC2 pin 19 go low. The first 64K of RAM. If we connect P50,#8 (bottom row), to P54,#8 (middle row) and SW4,#8 (rightmost switch) is closed/on (ground) then we restrict the board to only the first 64K of RAM.
It's nice to be able to debug/bring up a board like this (with no RAM initially) with another board in the bus so we can us a ROM based monitor to debug the hardware. So let's restrict the board further to a 32K - 64K RAM window. Put an old 8K RAM board at address 0000H (for a stack for your monitor software) in your system. Jumper P50,#1 to P54,#1 and set SW4,#1 (leftmost switch) open/off (high). Then only when Address line A15 is high will the board be active (7FFFH-FFFFH).
So go ahead and add IC2 & IC3 (74LS682's), U2 & U24 (74LS244's) and U25 (74LS32) to the board. Insert it in your S-100 box and display memory. Now the Z80 refreshes S-100 RAM all the time so if you just look at pin 8 of U25 it will always be pulsing low. The question is when in a bus cycle. If we put code in our memory board at 0H such as:-
C3 00, 00
and jump to the location. The Z80 will be in a tight loop jumping all the time to zero. Under these circumstances pin 8 of U25 will show high with most logic probes. There is still a low going refresh pulse but it is now a small percentage of bus time. Hit reset and it will go back to pulsing low again.
This test assumes your ROM monitor is on your Z80 board and can overlap with a RAM board in the 32-64K window. If your ROM is on another board you may need to utilize the S-100 phantom line on our the RAM board. In this case jumper P50, # 5 to P54, #5 and open SW4 #4 (high) so pin 12 of U3 is high.
If you setup the jumpers as described above, pin #8 of U25 should go low only when RAM from 7FFFH-FFFFH is being displayed. BTW make sure your monitor is not using low memory to store flags etc. Only when you get the above correct go further. Here is a picture of me checking out this step.
The above scheme is quite flexible. By adjusting A15-A12 you can create RAM "holes" down to a 4K resolution. Useful for any ROM boards you may have in your system. BTW, the SOUT, sINP, sINTA circuitry is there to make sure the board is not addressed for I/O operations -- particularly for interrupt acknowledge situations.
Next we add in the following circuitry.
The 74LS139 (U4) looks at address lines A20 and A21 to determine where within the boards bus 4MG address space each RAM chip pair will reside. There are just 4 options. At 000000H, 010000H, 020000H and 030000H. Unless you are using an 80286 or greater only the first option is used since a Z80 or 8086 is capable of only reaching up to 1MG of RAM. We will be using more RAM with later CPU's in the future but for now one board is way more than enough!
Note; the remaining two S-100 Address lines (A22 & A23) are examined by IC2 and with 4 of these boards would determine where each (4MG board) would reside within the 16MG address space, (in 4MG blocks), of the S-100 Bus. They would reside at 000000H (this case), 040000H, 080000H or 0C0000H.
See the diagram and schematic. Remember RAM chips are use in pairs here. It does not matter which pair of slots you use for your 1st MG of RAM its just RCSO* has to be jumpered to it (RAMCS0*, RAMCS1*, RAMCS2* or RAMCS3*). I happened to use the RAMCS1* ((U104 & U105) pair. Also add the chip RAM pair select LED's. Also add the 4MG Board select LED (blue in my case). Here is a picture of the board at this stage.
Next we add the rather complex circuitry that processes the data transfers to/from RAM 8/16 bits wide. If you like, you can regard it as a black box. Again step things along. See this circuit.
Insert just U9, U12 and U13. Do an 8 bit (Z80) memory read and write to RAM at 0H and check you get a low to high pulse for a read or write as shown above. Here is a picture of the process.
For now don't bother about 16 bit read/writes for now. Carefully add the rest of the RAM read/write circuitry. Everything in the above schematic which can be see better here. As I said above, the circuit has the ability to add 0-8 wait states to any RAM read/write (See below). However since that circuit is not yet added, we need to (temporally) bend out pin #4 of U10 otherwise it will put the bus in a hold state. Here is a picture.
Repeat the above RAM read and write tests. Only go forward when you are convinced the 8 bit Read and Write signals are correct. You can add the remaining LED's which (if inserted correctly) will save you using a logic probe. A common problem is people insert the LED incorrectly. First just place the LED in the holes. Giggle it around. It should light up. If not reverse the polarity. Only then solder it in place. You will not damage the LED (assuming you have the 470 ohms resistors in place). In order to have the write LED light up (very narrow pulses) fill memory from 8000H to A000H with 0's with your monitor and watch. I have found that the smaller LED's really don't light up well with this narrow pulse. I use the larger (red) LED's. I also used a 100 ohm resister instead of the usual 470 ohm to get more light.
Next we are ready for the critical test, the RAM test. Double check that you are getting (only) 5 volts on pin 32 for the chip. These particular RAM chips normally operate at 3.3 volts. They work fine at 5 volts, but 5.5 volts is the maximum. Carefully insert a RAM pair. Take the usual precautions about static electricity. Also install the RAM buffers. Indeed, everything on the board except U27 and SW2.
Check out the board. You should now see RAM at 8000H - 9FFFH with your monitor. Check it out by changing and moving bytes etc with your monitor. You can test the IC3 circuitry is working correctly by playing around with the jumpers (1-3) of P50 and SW4. Here is a RAM map (after I removed my 8K of RAM (see above) at 0000H) and jumpered p50,#1 to P54,#1 and set SW4,#1 to ground (i.e. the board is only active if A15 is low).
If you see this you are almost there. You can remove any other RAM board from your system. You then can open the board up to its full 1MG addressing capability by essentially cutting out the input from IC3A. We do this by jumpering all the pins of P52 to P54 and setting all SW4 switches to ground. This way U3 pin 19 is always low. The board is addressed directly by the RAM chips (and A20 & A21 of U4).
There are some subtle inputs/things you can do with some of the connections on P50 but this is for later boards and CPU's I will discuss them at that time. For now ignore them. However if your system utilizes the Phantom line for your ROM board/monitor jumper P50,#5 to P54,#5 and open SW4,#5.
So far we have just tested the board with 8 bit read and writes. For 16 bit systems you will need a 16 bit CPU. I have tested the board on my own IEEE-696 compatible 8086 & 80286 boards and ran it fine with CPM86+ and MSDOS. The board also works fine with the 68000 CPU Board. Here is a "Memory Map" with that CPU.
The circuitry is very clean and fast. It should work with most boards but of course no guarantee. I have run it with a number of Z80 systems, Intersystem's, Compupro etc. Absolutely rock solid -- in fact way overkill for these CPUs!
The remaining circuitry is U27, the resistor pack RR1 and switch SW2. You can bend in pin #4 of U10 at this stage also. This circuitry allows you to add 0-8 wait states to any board read or write cycle. With 50ns RAM chips --- clearly not needed by most current systems. However since I intend to use this board in the future with very fast CPU's may as well have it now. With all SW2 switches to the OFF position, no wait states will be added the bottom left LED should never come on. Thanks to
Henry Broekhuyse a parts list for the board is available below.
Here is a picture of the jumpers and switches in the boards most common configuration. (4MG RAM, 0-3FFFFFH, no wait states):-
A Production S-100 Board Realizing that a number of people might want to utilize a board like this together with a group of people on the Google Groups S100Computers Forum, "group purchases" are made from time to time. Contact and join the group if you would like to be involved in this project. See if bare boards are available and/or see if you and others may be interested in doing another board run. Please note all the above clearly applies only to people who know what they are doing and can do a little soldering and board assembly. There will be little hand holding at this stage.
Pontus Oldberg has recently reported that switching to a 74F11 for U12 & U13 and a 74F02 for U15 that this allowed him to use the board with a 68K CPU at 10MHz.
