Guys I see some of you have already gotten your boards, that’s great. I want to let you know that I am picking up a slight timing problem with the 32MB RAM daughter board. When I push it to high speeds (40MHz) I find that execution of actual 80386 code in protected mode RAM on the daughter board (e.g. 2000000H) does not reliably work. The simple 80386 monitor test program is brutal on the hardware, since in pipeline mode it’s in a tight loop reading protected mode RAM and outputting an S100 bus character to a port. I.e. it flipping back and forth between the two buses all the time. I’m not seeing a character on the screen – as I should (and do, for say RAM at F0000H). Interestingly it is alive because if you hit the normal ESC character to stop the test the monitor comes right back. BTW, this test is done with the 80386 monitor “Y” command. It will ask you for a RAM test location. The problem I think is the whole RAM chip selection cascade is too long and right on the edge in terms of timing. With the prototype boards It seemed to work using 74S32 and 74S139’s. Not now! I thought it was because I was using a Cyrix 80486 chip but is seems to be a problem with the AMD 80386’s as well. It is interesting that display RAM, Move RAM, copy RAM , compare RAM and the likes work perfectly fine. However these commands are less intense I suppose in switching between busses. BTW, to state the obvious, for real mode RAM, e.g. MSDOS, etc. absolutely no problem. The good news is there is a very simple patch that seems to take care of the problem:- Bend out pin 12 of U23 and bridge across a small wire to its pin 13. This eliminates the DB_ADS* signal in the RAM chip selection equation. With this patch the board has no difficulty (overnight test) with the above 80386 Monitor “Y” command test. The only problem is now the D1 LED stays on all the time when the board is being accessed. The other 3 LED come on fine when the appropriate address range on the board is being addressed. I don’t have an explanation for this right now. I do know that the address lines themselves (particularly in pipeline mode) on the 80386 are in a continuous state of flux, not latching them is probably why the LED display is strange. Clearly the address lines are correct by the time CPU RAM read or write signals arrive. . Also you don’t need the 74Sxx chips, the 74LS’s are fine. Anyway just wanted to alert everybody in case you spend needless time on this. I will look into it some more but currently I’m thinking it may later be better to do a V2 version of the board with a faster/shorter RAM chip selection circuit. For example I just noticed bA32 and bA30 to U14 could be brought directly into U20 since bA23 and bA24 are not needed – eliminating U14A. Will probably wait however until we do the 80486 board because that board may have even more unique requirements. I’m attaching a picture of the change. Should be obvious to an experienced group like this I’m sure. Finally, I have not built the 8MB RAM board yet, but since it incorporates the exact same circuit, I suspect it will require the same patch. John From: John Monahan [mailto:mon...@vitasoft.org] All orders went out today guys. I sent them USPO “Priority mail”. Overseas it’s ~10 days – they tell me. Enjoy! Let us know how they work out. BTW I still have:- Two 80386 S100 CPU boards Two 8MB Daughter RAM boards Five 32/64MB Daughter RAM boards. Anybody else want these please let me know. John From: John Monahan [mailto:mon...@vitasoft.org] Here is my Revise/updated list. Any errors please let me know ASAP. I now also have all your shipping addresses, see attached, the address I will use. For new arrivals, please read below plans for payments etc. I will send them via US post office, with $10 value for overseas (to avoid customs etc.). No insurance unless requested. John Name, 80386 CPU Board , 8MB SRAM Board, 32/64MB SRAM Board, Mezzanine V06c Boards. Fabio Battaglia 3,3,3,12 Gary Kaufman 2,1,1,4 Neil Breeden 1,1,1,4 Ian Douglas 3,3,3,12 Andrew Bingham 1,1,1,4 David Fry 1,0,1,6 Leon Byles 1,0,1,4 Vince Mulhollon 2,2,2,8 Josh Bensadon 1,0,2,8 Alex Jacocks 1,1,2,8 Pete Plank 1,1,1,4 Robert Greenstreet. 1,1,1,4 Curt 1,1,1,0 Ed Snider 1,0,1,4 John Monahan 3,3,3,12 Todd Goodman 3,3,3,0 Matt Turner 2,2,2,8 Andrew Lynch 1,1,1,4 Paul Birkel 1,1,1,4 Rob Doyle 1,0,1,4 Don Caprio 1,0,1,4 From: John Monahan [mailto:mon...@vitasoft.org] Well guys the order for the S100 80386 & RAM boards from PCBCart has arrived from China. One heck of a large order for me (see picture) . I’m delighted to say that at least for the 80386 CPU board and 32MB RAM board there are no errors. I assembled both boards and they work fine in both of my systems with a 40MHz oscillator. I have not yet built the 8MB RAM board as I am waiting on sockets, but don’t anticipate any problems since there was only one small change from the last prototype. Also I have not tried the 32MB board in a 64MB configuration since I don’t yet have the 4MB X 8 chips. I have written up the step by step build here:- http://s100computers.com/My%20System%20Pages/80386%20Board/80386%20CPU%20Board.htm#Build These are complicated boards, so take your time building them. I actually have three 80386 CPU boards running here currently. They appear to be very reliable. I have run the Protected mode RAM test overnight a number of times, never a problem. That said, the traces on these boards are quite tight so use little solder and a small tip. BTW for the 80386 board, I had PCBCart use 2X thick copper to distribute the current over the board better. If you build the CPU board carefully and still have a problem getting it to run after checking it out for obvious problems I will offer to try and debug it for you if you send it to me WHEN I can get some spare time. Most of you on this list are by now seasoned builders, so I don’t anticipate this should happen often. Now the bad part, I have to try and mail these boards off to everybody! I will pack the boards in padded envelopes with a cardboard backing. The average price per S100 board in the end from PCBCart worked out to be $16/board. I will be also adding Todd’s Mezzanine boards which run at $3/board. Packaging will be $2. Shipping is variable and unknown. When you get your package, note the USPO stamp cost and add it to the total. Use the formula below to calculate your total invoice. The formula below is what I intend to send. I honestly don’t have time to nickel & dime costs and keep a track of everybody and their payments. I am counting on each of you to use the honor system and forward payments due to my PayPal account at monahan AT vitasoft.org (AT=@). My current working list is as follows:- Name, 80386 CPU Board , 8MB SRAM Board, 32/64MB SRAM Board, Mezzanine V06c Boards. · Fabio 3,3,3,12 Gary 2,1,1,4 Neil 1,1,1,4 Ian D. 3,3,3,12 Andrew B. 1,1,1,4 David 1,0,1,6 Leon 1,0,1,4 Vince 2,2,2,8 Josh 1,0,2,8 Alex 1,1,2,8 Pete 1,1,1,4 Robert G. 1,1,1,4 Curt 1,1,1,0 Ed 1,0,1,4 John M. 3,3,3,12 Todd 3,3,3,0 Matt T. 2,2,2,8 Andrew L. 1,1,1,4 Paul B. 1,1,1,4 Total 30,25,31,110 If I have your order wrong, first sorry, please let me know ASAP. John |
Attachment:
Patch for 32MB RAM Board.jpg
Description: JPEG image
