---

Amiga 8meg ram expander (C)1991 Silicon Synapse Electronics Designed by John Kamchen Support BBS: (204) 589-1078 Wpg,MB Canada Nov.28 ----------------------------------------------------------------------------- This project is ShareWare. If you build it, and it works (why wouldn't it?) then passing $20 along my way would help pay for future projects as fantastic as this one. The address is at the end of this file. ----------------------------------------------------------------------------- With memory prices dropping every day, a fully expanded Amiga looks better all the time. If you shop around, you can get a 1meg simm for less than $40. That means an 8meg system will set you back $320. Not that bad, considering 3 years ago that would have got you only 2 megs. But you need more than just RAM, you also need a RAM CARD (duhh). These days, you have a big selection to choose from. GVP, IVS, ICD to name a few. Some are simply RAM cards, some have an SCSI host as well. My GrandSlam cost $300, a Meta4 about $175, DataFlyer RAM about $125. Now you can build your own 8meg expander for less than $10 ! If you shop around, you can get all the interface and control chips for less than $5 ($4.98 to be exact). The parts list shows prices from Active. Adding sockets and whatever jacks the price up a few bucks. Parts List: Cost U1 74ls138 .40 U2 - U6 74ls157 2.35 U7 - U8 74ls245 1.28 U9 74ls14 .35 U10 74hc08 .31 U11 74ls32 .29 Misc: Decoupling caps 11 x .1uf How it works: RAM ACCESS When the CPU accesses the area from $20000 to $9FFFFF, U1 decodes A21-A23 to provide 4 possible bank select signals. About 58ns after _AS goes low, _RAS will also go low. Depending on what kind of access it is (WORD BYTE READ WRITE) one or two _CAS lines will go low either 94ns (READ) or >100ns (WRITE) after _AS low. Remember that on the 68000, _UDS/_LDS go low at the same time as _AS for a READ, but are delayed for a WRITE. That is why they go thru the 2 OR gates (U11). This ensures that _CASxx will fall at the right time. When _AS goes high at the end of the cycle, _CASxx goes high first, then _RAS. REFRSH MODE The DRAM memory is being used in _CAS before _RAS refresh mode. This means that if _CAS is brought LOW about 20ns before _RAS, the DRAM automaticly selects what row to refresh. Only 64k x 4, 256k x 4 and 1meg x 1/4 support this feature. Otherwise, there would need to be counters and stuff added, maybe 5 more chips. When the CPU accesses CHIP ram ($0 to $1FFFFF), Y0 of U1 goes LOW. This is the signal that says 'Time to refresh, cuz we ain't gonna be needed on this cycle'. All _CASxx lines will come low 72ns after _AS LOW, followed by _RAS LOW 100ns after _AS LOW. So the time diff between _CAS and _RAS is about 28ns, within the 20ns min parameter. The chip is kept refreshed untill the cycle ends. CONSTRUCTION: This design uses D0-D15, A1-A23, R/W, _AS, _UDS and _LDS. All these signals can be found at the expansion port of the 500, or CPU/Zorro slots on a 2000. For 500 owners ,this board could also be placed under the 68000 itself, saving your expansion port for better things. As long as you provide a good power path, there should be no problems. My proto-type board with 2megs used around 300ma while running Memory_Doctor, and 169ma just sitting there, no programs running. It uses ALOT more than my GrandSlam (90ma) for refreshing, but I'm not using custom chips and they are. For a 500 supply, 8megs might be more than it can handle, so consider buying a 150-200w switching supply (any AT type will do since they cost half as much as a '500 specific' supply). SIMM sockets cost ALOT of bucks, so you may want to spend a $1 more and buy SIPP ram (with the pins). Those can fit into normal IC sockets (Just to give you an idea, 8megs of SIPP sockets cost $21, while 12 20pin sockets cost $2.50). The SIPPs will tend to sit in at an angle, so fashion some type of brace for them. If you use SIPP ram, the board won't fit inside a 500, but low-profile SIMM sockets just might. Get the exact same parts that I have in the parts list. Don't sub an HC for an LS. All types have been chosen for their propagation times. Best place to get the parts is Active Components. They are cheap, and always have good stock. I have heard of people being charged upto $2 a chip at some other places (Radio Shack, WES). Board layout isn't critical, but use some common sense, try and keep data and address lines together, and so forth. Wire wrapping is OK. ADDING RAM: You can use 1meg by 8 (or 9, some place charge MORE for the 8bit sipp!), or 256k x 8/9 sipps. To use the 256k, they must have only 2 chips on them (3 if it is a 9bit simm). Why? The ones with 8 or 9 chips use 256k x 1 type DRAM chips. Those are fine in other ram expanders but not here. They DON'T support _CAS before _RAS refresh. The 256k x 4 do, so that's what type you need. ___________________________ ___________________________ | || || || || || || || || | |-------- -------- ----| | || || || || || || || || | |-------- -------- ----| uuuuuuuuuuuuuuuuuuuuuuuuuuu uuuuuuuuuuuuuuuuuuuuuuuuuuu These won't work! These WILL work fine. AUTO-CONFIG: As is, the design will not configure on power up. A program called AutoAddRAM is used to patch the expansion into the free memory pool. If you have a hard drive, this can be done during the startup-sequence. For floppy users, follow the docs for the program. Unlike a PAL'd ram card, this one hard-wires the ram address area. The folowing are some examples of how to write the .arr file. 200000 3ffffe 0 Little /* 2meg, using 1meg simm, in bank 1 & 8 400000 9ffffe 0 BigBoy /* 6meg, using 1meg simm, bank 2,3,4,5,6 & 7 200000 27fffe 0 Expansion /* 256k, using 256k simm, bank 1 & 8 400000 47fffe 0 Expansion /* 256k, bank 2 & 7 600000 67fffe 0 Expansion /* 256k, bank 3 & 6 800000 87fffe 0 Expansion /* 256k, bank 4 & 5 With those last 256k areas, you just added 2megs, although it's broken up into 4 256k sections in the expansion mapped area. This is a cheaper way to get to 2megs, since those 256k simms are CHEAP when bought used. I found that APPLE service centers seems to always have the right type of simm (I assume Apple/Mac machines use _CAS before _RAS as well). You can also mix and match memory sizes, 2megs using 1meg simms, and an extra 768k using 256k simms. This list shows where your memory will show up. Bank1/8 Bank2/7 Bank3/6 Bank4/5 from $200000 $400000 $600000 $800000 256k $27fffe $27fffe $67fffe $67fffe 1meg $3ffffe $5ffffe $7ffffe $9ffffe I provided an .iff picture of where to place the parts and all. There are holes here and there for decoupling caps. There are also 2 files in BoardMaster format, which contain the actual artwork. Those can be plotted, or HP LaserJet printed using BoardMaster. A good way to mount this board is to solder 64 wires into the board, then solder your 68000 to those (Look at the AtOnce). Just an idea. ---------------------------------------------------- Also from Silicon Synapse Electronics ShareWare Division: S.A.P.E.P. (Simple Amiga Parallel Eprom Programmer) Incredible software/hardware combo. Program 2764-27128 21v/12.5v eproms thru the Pport. Fairly simple project shows just how versitile the Amiga's parallel port can be! Includes complete schematics, SAPEP program & C source. The Computer Tech Journal Issue #1: Audio Sampler Fundamentals (includes demo circuits) Issue #2: 8meg Ram for Amiga (ignore this, new one this fall) Issue #3: Eprom Programmer (Look for S.A.P.E.P., software/hardware project) Issue #4: Proto-Board for the Amiga (a must have for any Amiga hacker!) To get all the ShareWare/PD files listed in this file, just send $2 (US or CDN) and your return address to: John Kamchen c/o Silicon Synapse Electronics 41 Matheson Ave.E Winnipeg Manitoba Canada R2W-0C1 Or call Silicon Synapse Electronics BBS (204) 589-1078 12/24 24hrs or Fire & Brimstone BBS (204) 255-8824 12/24 24hrs Fido# 1:348:705.0 J.Kamchen Nov.27 1991 3:30pm -eof-

---

The views and opinions stated within this web page are those of the author or authors which wrote them and may not reflect the views and opinions of the ISP or account user which hosts the web page. The opinions may or may not be those of the Chairman of The Skeptic Tank.

Return to The Skeptic Tank's main Index page.

E-Mail Fredric L. Rice / The Skeptic Tank