ranjit@eniac.seas.upenn.edu (Ranjit Bhatnagar) (04/14/89)
I'm considering building one or more of the following hardware projects, for kicks and to save some dough. I'd like advice from those In the Know as to which are practical and which I should save for the pros, and any relevant information that you can spare. * A zorro-II card cage for the 1000 I figure I need a case, a real smooth power supply, and about a zillion bus drivers. I'd hope to be able to plug 2000-style memory and hard disk cards in it. This SEEMS like it should be easy, so what am I missing? What's a limit on the number of slots I can have? How do I avoid timing problems and such? And where can I get a QUIET fan to put in the box? * A 32-bit memory card for the Ronin Hurricane 1000. I have no idea how complex a project this is - there must be SOME reason why an unpopulated 32-bit card costs over 500 bucks, but I'd love to avoid paying that. What DOES make a 32-bit card cost so much more than a 16-bit card, after you factor out the memory chips themselves? And are 120 ns chips fast enough? Would a hand-drawn pc-board be far too noisy for such an application anyway? (Well, drawn in MacDraw anyway.) Could the LUCAS design be easily adapted? * A memory card - 32-bit or 16-bit - that would take SIMMs. Sometimes those are cheaper. Well, that's about it. If I dig up or invent any successful designs, I'll share 'em with the net... - Ranjit "Trespassers w" ranjit@eniac.seas.upenn.edu mailrus!eecae!netnews!eniac!... I'd be calling all the hotlines/I'd be missing all my deadlines I'd be standing in the breadlines/I'd be six feet underground
daveh@cbmvax.UUCP (Dave Haynie) (04/19/89)
in article <9891@netnews.upenn.edu>, ranjit@eniac.seas.upenn.edu (Ranjit Bhatnagar) says: > I'm considering building one or more of the following hardware > projects, for kicks and to save some dough. That's cool, and really shouldn't be any major design problem. > * A zorro-II card cage for the 1000 > I figure I need a case, a real smooth power supply, > and about a zillion bus drivers. You need about 2 Amps per slot, plus an additional 2 per box, if you're following the specifications. We say you get 2 Amps in any given slot, but one board (presumably the one with "all that memory" on it can really draw up to 4 and still be safe). You need bus drivers 'ala "backplane example" from the A1000 "Schematics and Expansion Specifications", or based on the German A2000 design in "A500/A2000 Technical Reference Manual", with corrections. > I'd hope to be able to plug 2000-style memory and hard disk > cards in it. This SEEMS like it should be easy, so what am > I missing? It really isn't going to be very difficult. There's no reason your Zorro II backplane connected to an A1000 or A500 shouldn't be able to run most (if not all) cards, providing your backplane design is sound. > What's a limit on the number of slots I can have? The limit you're most likely to run in to is the number of slots you can serve by your DMA arbiter. From a physical point of view, the existing specs will have no trouble with 6 or fewer slots, though the actual bus buffer implementation will set the real achieveable number. > How do I avoid timing problems and such? Look at the "B2000" design in the "A500/A2000 Technical Reference Manual" and try to emulate that. The maximum delay through the BUSTER chip is speced at 40ns for non-clocked outputs. > And where can I get a QUIET fan to put in the box? I wonder if anyone at Commodore would know this -- I certainly don't. > * A 32-bit memory card for the Ronin Hurricane 1000. > I have no idea how complex a project this is - there must > be SOME reason why an unpopulated 32-bit card costs > over 500 bucks, but I'd love to avoid paying that. > What DOES make a 32-bit card cost so much more than > a 16-bit card, after you factor out the memory chips > themselves? And are 120 ns chips fast enough? The actual interface available must be based on what Ronin presents on their memory connector. One obvious explanation for the extra cost of such a board is "we're the only ones who make it". If you're considering building a 32 bit add-on, I'd suggest you get information on the Ronin, GVP, CSA, and possibly other daughterboard hookups before designing the card. Even if you can't come up with a 1-board-fits-all design, you may be able to offer the same design with different connectors. The speed of the chips you need is completely dependent on what you're trying to do with the system. You could build a board with 120ns DRAMs that goes faster than normal A2000 RAM boards, but it's likely that someone else will build one with 100ns DRAMs, which could beat the best 120ns design by a wait state. > Would a hand-drawn pc-board be far too noisy for > such an application anyway? (Well, drawn in MacDraw > anyway.) Could the LUCAS design be easily adapted? Nothing's inherently wrong with hand-drawn boards, if you're using a CAD program of some kind, though you may have to manually follow design rules to get a reproducable board. LUCAS looks like may have problems with other add-on devices. > * A memory card - 32-bit or 16-bit - that would take SIMMs. Sometimes > those are cheaper. SIMMs are always more expensive than DIP or ZIP parts, by 20-50%. > - Ranjit -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy Amiga -- It's not just a job, it's an obsession
elg@killer.Dallas.TX.US (Eric Green) (04/23/89)
In message <6622@cbmvax.UUCP>, daveh@cbmvax.UUCP (Dave Haynie) says:
$in article <9891@netnews.upenn.edu>, ranjit@eniac.seas.upenn.edu (Ranjit Bhatnagar) says:
$> I'm considering building one or more of the following hardware
$> projects, for kicks and to save some dough.
$That's cool, and really shouldn't be any major design problem.
However, there may be some construction problems. I remember reading
about noise nightmares with wire-wrapping on ancient S-100 busses
using 2mhz 8080s, and similiarly with memory boards, so will
definitely have to be a pc board. Do you have the facilities? If not,
prototyping services can get expensive real quick (i.e., forget about
saving some dough).
$> * A zorro-II card cage for the 1000
$> I figure I need a case, a real smooth power supply,
$> and about a zillion bus drivers.
This would be a VERY good project. Current Zorro-II card cages only
have 2 or 3 slots, and are somewhat pricy (i.e., >$175). I suspect
that most of that's because of the power supply and case -- a
commercial outfit cannot rely on cheap surplus power supplies and
cases. You can.
One nice thing is that, if you design it right, it can also be
used with the Amiga 500. Just make sure that the "fingers" stick out
far enough so that components don't bump up against the "ledge" on an
Amiga 500 (means you'll see a little open air between the box and the
A-1000, but that should be no big deal).
My purchasing plans for the future include a hard disk controller,
a RAM card, and a multi-serial I/O card. I have an old IBM power
supply to supply juice to it all. Current 2-slot A-500 boxes are
simply too small, and contain a power supply and case that I do not
need.
I just asked my brother what it'd cost to get quantity three from a
prototyping service. He said it'd be probably about $40/board plus a
setup charge of maybe $100. Going quantity 100 would probably bring
the blank PCB cost below $10, but the question of what to do with the
other 99 arises ;-).
$> And where can I get a QUIET fan to put in the box?
I always wonder about people who insist on QUIET fans. The first time
I turned on an Amiga 1000, I almost freaked, because I didn't hear the
vigorous "whoosh" of most AT-clone fans. That "whoosh" is reassuring
to me, because it means that my fan is still running and my equipment
is still keeping cool.
$> * A 32-bit memory card for the Ronin Hurricane 1000.
$> I have no idea how complex a project this is - there must
$> be SOME reason why an unpopulated 32-bit card costs
$> over 500 bucks, but I'd love to avoid paying that.
The costs of low quantities extend beyond the higher cost of the PC
boards (which in itself is quite significant). It takes just as many
people to ship out a hundred per month as it takes to ship out a
thousand per month, and I suspect that Ronin is shipping in relatively
low quantities. Your development staff has to be paid whether you're
shipping hundreds or thousands, unless it's a one-man operation...
$> Would a hand-drawn pc-board be far too noisy for
$> such an application anyway? (Well, drawn in MacDraw
$Nothing's inherently wrong with hand-drawn boards, if you're using a
$CAD program of some kind, though you may have to manually follow
$design rules to get a reproducable board.
You may want to get PCLO or some other such program. Drawing up PC
boards entirely by hand is, at least for me, infuriating.
--
| // Eric Lee Green P.O. Box 92191, Lafayette, LA 70509 |
| // ..!{ames,decwrl,mit-eddie,osu-cis}!killer!elg (318)989-9849 |
| \X/ Amiga. The homestation for the blessed of us. |carlson@betelgeuse (Richard L. Carlson) (04/23/89)
In article <7916@killer.Dallas.TX.US> elg@killer.Dallas.TX.US (Eric Green) writes: >In message <6622@cbmvax.UUCP>, daveh@cbmvax.UUCP (Dave Haynie) says: >$in article <9891@netnews.upenn.edu>, ranjit@eniac.seas.upenn.edu (Ranjit Bhatnagar) says: >$> I'm considering building one or more of the following hardware >$> projects, for kicks and to save some dough. >$That's cool, and really shouldn't be any major design problem. > >However, there may be some construction problems. I remember reading >about noise nightmares with wire-wrapping on ancient S-100 busses >using 2mhz 8080s, and similiarly with memory boards, so will >definitely have to be a pc board. Whoa...not so fast! :-) I built a RAM expansion when it came across the net 2-1/2 years ago (back when comp.sys.amiga wasn't comp.sys.amiga :-) and I've been using it in my Amiga {1,2}000 ever since without any problems. It's all wire wrapped. Agreed that you need to be careful to keep signals as short as possible, make the power supplies very clean (my supplies actually used wire heavier than wire-wrap gauge), etc., but it *is* possible to wire-wrap your hardware hacks. >$> And where can I get a QUIET fan to put in the box? > >I always wonder about people who insist on QUIET fans. The first time >I turned on an Amiga 1000, I almost freaked, because I didn't hear the >vigorous "whoosh" of most AT-clone fans. That "whoosh" is reassuring >to me... Personally, when I want to be reassured, I want to have to listen carefully to hear the fan, not have it announce its presence a couple of rooms away. I recently had to replace my 2000's fan, though, and noticed something interesting: to make sure I had the fan wired right, I turned the power supply on with the new fan just dangling---it was practically silent. "Wow!" I thought, put everything back together, turned the power on, and---the new fan was as noisy as the last one (but no more periodic garbage disposal). It looks to me like the 2000's fan is noisy only because it's accoustically coupled to the rest of the machine. I know there is very little room in there, but has anyone tried to isolate the fan so that its vibrations don't get transferred to the rest of the machine? I think it would improve the noise measurably... -- Richard {tektronix,dual,sun,decvax,...}!ucbvax!ernie!carlson carlson@ernie.berkeley.edu
ranjit@leguin.cis.upenn.edu (Ranjit Bhatnagar) (04/23/89)
Thanks to Dave and Eric for their advice! With a bit more help like this, LUCES (a Little Ugly Cheap Expansion System) may show up soon. elg@killer.Dallas.TX.US (Eric Green) writes the > lines: daveh@cbmvax.UUCP (Dave Haynie) says the >$ lines: I wrote the >$> lines: >$> I'm considering building one or more of the following hardware >$> projects, for kicks and to save some dough. >$That's cool, and really shouldn't be any major design problem. As it turns out, at the time I posted the original message, I was under the impression that there wasn't much more to an amiga card cage than 100 parallel wires and a buncha sockets. I have no experience in designing bus arbitration circuitry, and the ancient white hardware manual doesn't say enough to teach me. But with someone else's electrical design (HINT!) I could probably manage the layout and mechanical stuff. > >However, there may be some construction problems. I remember reading >about noise nightmares with wire-wrapping on ancient S-100 busses >using 2mhz 8080s, and similiarly with memory boards, so will >definitely have to be a pc board. Do you have the facilities? If not, >prototyping services can get expensive real quick (i.e., forget about >saving some dough). What about a board laid out in MacDraw or some similar CAD program, and then optically printed or silk screened? I can do that in my basement... except for drilling the holes. How practical is it to print a board on perforated PC-board? Would have to make the traces extra wide so they're not interrupted by holes, but it would save tremendous amounts of time not to have to drill. Silk-screening is very cheap and fast, so if there is a way to avoid making your own holes, I think it could make it worthwhile to avoid the professionals for runs of a hundred or less... (and I was only thinking of ONE!) > >$> * A zorro-II card cage for the 1000 >$> I figure I need a case, a real smooth power supply, >$> and about a zillion bus drivers. > >This would be a VERY good project. Current Zorro-II card cages only >have 2 or 3 slots, and are somewhat pricy (i.e., >$175). I suspect >that most of that's because of the power supply and case -- a >commercial outfit cannot rely on cheap surplus power supplies and >cases. You can. Yeah - I figured about 50 bucks worth of parts altogether. > I just asked my brother what it'd cost to get quantity three from a >prototyping service. He said it'd be probably about $40/board plus a >setup charge of maybe $100. Going quantity 100 would probably bring >the blank PCB cost below $10, but the question of what to do with the >other 99 arises ;-). Well, that's pretty cheap. If something like this took off like Lucas, it would be worthwhile. >$> And where can I get a QUIET fan to put in the box? >I always wonder about people who insist on QUIET fans. The first time >I turned on an Amiga 1000, I almost freaked, because I didn't hear the >vigorous "whoosh" of most AT-clone fans. That "whoosh" is reassuring >to me, because it means that my fan is still running and my equipment >is still keeping cool. And it keeps ME awake at night. Can't wait to dump the 2000 for a nice peaceful 1000. (OR move the 2000 into another room...) >$> * A 32-bit memory card for the Ronin Hurricane 1000. >$> I have no idea how complex a project this is - there must >$> be SOME reason why an unpopulated 32-bit card costs >$> over 500 bucks, but I'd love to avoid paying that. I got nice answers about low volume and high startup costs, but what I really was hoping for was design hints. Is refresh circuitry qualitatively different for such an application? (Not that I know anything about ordinary refresh circuitry - but I learn fast.) If one wanted DMA, how would one do it? Has anyone out there reverse-engineered Ronin's memory connector, or would I have to plead with them for info? Although I'm no longer in any hurry on this project, because I decided to just buy a 2-meg SOTS box and put off the expansion and 32 bit memory til later, I'll keep at it if other people seem enthusiastic and if I can get help designing interface circuits. No reason why this shouldn't go the way of LUCAS - a nice, cheap design for all to share. Some things to think about... One can probably save money and get quicker results for quantities under 100 or so boards by silk-screening them. I've used silk-screen only for fine-arts applications, but the application to pc design is obvious. A silk-screen kit costs about $40 in an art store; a laser-printed design can be optically transferred to the screen in about 3 hours; after that, one can turn out about 30 prints an hour with registration precise to about 1/20th of an inch if one is careful. (Important for doing 2-sided boards!) Then dump all the prints in acid and there you have it. I still don't know how to handle the holes, though... automated drill presses are a bit more expensive. This technique is practical if there are no major objections to printing on prepunched PC boards. If there are people out there who are enthusiastic about such a project AND know what they're doing (I only satisfy the first condition), I wouldn't mind if they did all the work, in the form of a circuit design for the expansion controller and, optionally, a layout. I would be happy to turn the design into a layout and turn out one or two or a dozen boards (once you've got the silk mask, there's no reason to stop). How about a system that sits above the 1000 or below the 500, just by rotating it? I like the subsystem idea because it saves desk space. Of course, a design that lies the cards on their sides requires sideways-mounted daughterboards and brackets to keep the cards from sagging. Can hardcards safely be operated sideways? Four slots in the 2000 take up about 3 inches of width, implying that a four-slot sub/supersystem could be just 4 inches high or so, and perhaps 6 inches wider than the length of a card to make room for a power supply and/or a half-height floppy or hard disk. Any objections to having the power supply in a separate box? That way it can be stowed on the floor to save desk space, or on the desk to save floor space. As Eric pointed out, any old surplus power supply will do... there's always 4 or 5 choices in the $20-30 range in the Computer Shopper. What's the maximum power requirements one would expect for four 2000 cards and one 3 inch hard disk? Should the expansion case have a pass-thru for any old SOTS boxes you might have sitting around? Could this cause any electrical problems? Should the case be painted Commodore Tan (boring) or Pixar Speckled (glamorous), or maybe black with brushed-chrome highlights and a little green flourescent 7-segment display that tells you just how much money you're spending... --ranjit "Trespassers w" ranjit@eniac.seas.upenn.edu mailrus!eecae!netnews!eniac!... I'd be calling all the hotlines/I'd be missing all my deadlines I'd be standing in the breadlines/I'd be six feet underground
elg@killer.Dallas.TX.US (Eric Green) (04/24/89)
in article <10267@netnews.upenn.edu>, ranjit@leguin.cis.upenn.edu (Ranjit Bhatnagar) says: > I have no experience in designing bus arbitration circuitry, > and the ancient white hardware manual doesn't say enough to > teach me. But with someone else's electrical design (HINT!) > I could probably manage the layout and mechanical stuff. >>However, there may be some construction problems. I remember reading > What about a board laid out in MacDraw or some similar CAD program, > and then optically printed or silk screened? I can do that in my Sounds like you have most of the equipment needed to make PC boards, then. Next thing you need is some resist, cleaner, and ferric chloride, if you want to do quantity 1. Photographic floods probably would supply enough UV for exposing the board. Get the book _Electronic Prototype Construction_ by Steve Kasten for more info (your book dealer will probably have to order it direct from Howard Sams Inc.). These are all very toxic chemicals, so follow his instructions for handling and disposal. > basement... except for drilling the holes. How practical is it Drilling the holes is no big problem. Any ordinary drill-press will do. Kasten gives a way of getting around the registration problems that can occur, too.... etch one side, drill, then line up the other side's holes & shoot & etch the other side. > I got nice answers about low volume and high startup costs, but > what I really was hoping for was design hints. Is refresh circuitry > qualitatively different for such an application? Maybe someone on the net could help by digging up chip #s/manufactures for refresh controllers. I just wish to point out that doing a memory board that connects to a 25mhz 68020/68030 is very different from doing a memory board that slaps on the side of an Amiga 1000.... > How about a system that sits above the 1000 or below the 500, just > by rotating it? I like the subsystem idea because it saves desk Won't work. You'd have to do some fancy sheet-metal work, and if you are off by a micrometer, you have an unsightly gap, mechanical stress on your edge connector, or some other nastiness. Whereas, if you slap it on the side, any old box will do (as long as it's big enough for the cards!). Looking at the ET slap-on-the-side, it looks suspiciously like an ordinary external floppy-disk case. Not to mention that my keyboard is already high off the floor, without adding another 4 inches ;-). > Any objections to having the power supply in > a separate box? That way it can be stowed on the floor to save > desk space, or on the desk to save floor space. As Eric > pointed out, any old surplus power supply will do... there's You're getting aweful fancy for a hacker's project. The question of "where do I put it?" for the power supply depends on exactly what kind of power supply and what case you get. In my case, I'd probably put a spare PC-clone power supply outside of the enclosure (it is in its own box already), and run the power to both the card cage and to the Amiga 500 (which has an amazingly wimpy power supply). -- | // Eric Lee Green P.O. Box 92191, Lafayette, LA 70509 | | // ..!{ames,decwrl,mit-eddie,osu-cis}!killer!elg (318)989-9849 | | \X/ Amiga. The homestation for the blessed of us. |
FelineGrace@cup.portal.com (Dana B Bourgeois) (04/24/89)
Eric Lee Green makes the comment that he doesn't understand why people want QUIET fans. I *love* the A1000 fan because it is so quiet. I can hear the telephone ring while it is on, I can listen to music without blowing down the doors, I can...well, you get the idea. As for the quietness of a fan, I think there are two ways to reduce fan noise and both of them involve reducing the speed of the blade. One way is to use an AC fan which is derated to reduce its noise. This is a fan that may move say 55CFM that the manufacturer has slowed down to deliver only 30CFM. The other way is to buy a DC fan which can be controlled by its input voltage. In other words, get your hands on a 4 inch DC fan and dial the voltage up until the fan noise is just below what you consider 'too loud'. I've seen a few DC fans in the surplus places and in the catalogs they are about 30-50% more than AC models. Hope this helps. Dana
daveh@cbmvax.UUCP (Dave Haynie) (04/25/89)
in article <7922@killer.Dallas.TX.US>, elg@killer.Dallas.TX.US (Eric Green) says: > Maybe someone on the net could help by digging up chip #s/manufactures > for refresh controllers. I just wish to point out that doing a memory > board that connects to a 25mhz 68020/68030 is very different from > doing a memory board that slaps on the side of an Amiga 1000.... Absolutely true. While you have more concerns about noise at those speeds, the main problem is the handling of refresh. With a simple, normal speed SOTS memory card for the A1000, it's possible to interleave memory access with refresh, much as the CHIP memory works, and still run full speed with cheap (150ns) memories. Once you even get to 14.3MHz on a 68020, even 100ns parts aren't fast enough on their own to give you no wait state operation, and they way you handle refresh can slow this down even more. Refresh is basically another arbitration problem -- the 68020 wants the memory bus most of the time, but you need to supply a refresh to keep things going or you're in big trouble. Thing is, you never know when the 68020 will be using the memory, and you probably don't know just when a refresh cycle is going to come along. Most of the design cleverness, on this level, is how you create an artifical time at which both the status of the 68020 cycle (either definitely in or definitely out of a cycle) and refresh are known long enough to prevent the other from taking place. There are lots of ways to solve this problem. A controller chip will do this for you, but they can be expensive and possibly either slow or uncooperative. The other half of a fast memory system is in making it as fast as you can using the slowest memory parts you can get away with that are also the fastest memory chips you can afford. There's on-chip support of special addressing modes in most DRAM -- the fast page, nybble, or static column modes you sometimes hear about. There are also architectural decisions on memory structure that may influence speed. And it's often a tradeoff between the number of chips, the complexity of the support logic, and the speed of the memory board; eg, adding more chips and/or more, clever support logic can result in a faster board. If you need a 4 meg board you can use nybble DRAM with a 68030 and get extra speed for very little extra logic. If you need a 1 meg board, you could get the same or possibly a better speedup using static column memories, but the support logic is much more complex. Juding 32 bit memory speed is going to be interesting. For example, most of the design enhancements have some form of locality of reference which could result, for instance, in board A being faster than board B for linear access, but board B coming out on top if you access every 2nd or 4th memory location. > | // Eric Lee Green P.O. Box 92191, Lafayette, LA 70509 | > | // ..!{ames,decwrl,mit-eddie,osu-cis}!killer!elg (318)989-9849 | > | \X/ Amiga. The homestation for the blessed of us. | -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy Amiga -- It's not just a job, it's an obsession
jbh@mibte.UUCP (James Harvey) (04/25/89)
In article <17547@cup.portal.com>, FelineGrace@cup.portal.com (Dana B Bourgeois) writes: > Eric Lee Green makes the comment that he doesn't understand why people > want QUIET fans. I *love* the A1000 fan because it is so quiet. I can > hear the telephone ring while it is on, I can listen to music without > blowing down the doors, I can...well, you get the idea. > > As for the quietness of a fan, I think there are two ways to reduce > fan noise and both of them involve reducing the speed of the blade. One > way is to use an AC fan which is derated to reduce its noise. This is > a fan that may move say 55CFM that the manufacturer has slowed down to > deliver only 30CFM. The other way is to buy a DC fan which can be > controlled by its input voltage. In other words, get your hands on a > 4 inch DC fan and dial the voltage up until the fan noise is just below > what you consider 'too loud'. I've seen a few DC fans in the surplus > places and in the catalogs they are about 30-50% more than AC models. > > Hope this helps. > > Dana If you can find one, a Two Microfarad, 200 volt, NON-POLAR (this means paper, mylar etc. not electrolytic) capacitor in series with the line slows down a 115 volt AC muffin fan just perfectly. I did this for years on my Commodore 64 disk drives and I use similar arrangements to force circulate air through the steam radiators in my living room. The advantage of using a capacitor for dropping the fan voltage is that the capacitor doesn't get hot like a resistor would. -- Jim Harvey | "Ask not for whom the bell Michigan Bell Telephone | tolls and you will only pay 29777 Telegraph | Station-to-Station rates." Southfield, Mich. 48034 | ulysses!gamma!mibte!jbh
rsb@dukeac.UUCP (R. Scott Bartlett) (04/28/89)
In article <6678@cbmvax.UUCP> daveh@cbmvax.UUCP (Dave Haynie) writes: > >Absolutely true. While you have more concerns about noise at those speeds, >the main problem is the handling of refresh. With a simple, normal speed >SOTS memory card for the A1000, it's possible to interleave memory access >with refresh, much as the CHIP memory works, and still run full speed with >cheap (150ns) memories. Once you even get to 14.3MHz on a 68020, even 100ns >parts aren't fast enough on their own to give you no wait state operation, >and they way you handle refresh can slow this down even more. Refresh is >basically another arbitration problem -- the 68020 wants the memory bus most >of the time, but you need to supply a refresh to keep things going or you're >in big trouble. Thing is, you never know when the 68020 will be using the Okay, I've been considering building a memory board for my A1000 for a while now, but there is one thing that I have been unable to find out from people that i have asked and the (somewhat dated) data books i have. How often do you need to refresh each individual column? (or is it a row? sorry but it has been a while since i have messed w/ this) Do you need to wait for the full time that a normal memory cycle would take (assuming CAS is already set up)? Or can you strobe RAS (w/ the address set up already) and go on? If so, how long do you have to wait before you can either do a normal memory access or another refresh? About a year ago i calclated how often i would have to do a refresh based upon my understanding of the timing information, and it turned out that the processor would almost never be able to access the memory because it was constantly being refreshed. (these were 256kDRAMS i was designing w/) I figured out that i was doing something really wrong, and i was misinterpreting something seriously. What gives? Exactly how often does an individual cell need to be refreshed before the charge leaks away? Thanks, rsb --> can't wait for that '040 board running at 40Mhz!! Nudge, Nudge, Wink, Wink.. -- rsb@dukeac.ac.duke.edu /// "DEC and IBM: Just say NO!" rutgers!mcnc!ecsvax!dukeac!rsb /// "I'm saying eleven."-- Negativ Land I'm a HORSE, of course. \\\/// sex -n necrophilia option (if target proc is Disclaimer NOT included! \XX/ not dead, program kills it) EUNUCH prog manual
daveh@cbmvax.UUCP (Dave Haynie) (05/01/89)
in article <1376@dukeac.UUCP>, rsb@dukeac.UUCP (R. Scott Bartlett) says: > Keywords: Refresh times > Okay, I've been considering building a memory board for my A1000 for a while > now, but there is one thing that I have been unable to find out from people > that i have asked and the (somewhat dated) data books i have. How often do > you need to refresh each individual column? (or is it a row? sorry but it has > been a while since i have messed w/ this) Refresh is done by row. The exact refresh cycle will, of course, depend on the chips you're dealing with. The 256k density parts typically need a 256 row refresh every 4ms, the 1 meg density parts typically need a 512 row refresh every 8ms. At least they made it easy this way, but this can vary by part type and even a little by manufacturer. > Do you need to wait for the full > time that a normal memory cycle would take (assuming CAS is already set up)? > Or can you strobe RAS (w/ the address set up already) and go on? If so, how > long do you have to wait before you can either do a normal memory access or > another refresh? The time is probably going to be driven by RAS, no matter how you go about refreshing. You'll typically have to keep RAS low for a whole Tras time, then hold it high for a whole Trp time. If you're providing the row address, you have to worry about Tasr and Trah, if you're using the built-in refresh counters in a CAS before RAS cycle, you have a Tcsr to wait from CAS to RAS; CAS going high won't be a gating item. If you're setting up the so-called hidden refresh cycle, where a normal cycle is immediately followed by a CAS before RAS cycle, you're completely RAS driven; just keep CAS low 'till near the end of the refresh cycle (there's some kind of CAS hold from RAS time to consider, but it's no big deal). > About a year ago i calclated how often i would have to do a refresh based upon > my understanding of the timing information, and it turned out that the > processor would almost never be able to access the memory because it was > constantly being refreshed. Definitely something wrong there. You didn't, perhaps, drop a power of 10^3 on the refresh timing, did you? Refreshing every 4uS vs. every 4ms would certainly eat up most of the available time. Also, on a slow memory board (like a standard Amiga 16 bit memory system) there's enough time in a memory cycle to interleave memory and refresh cycles. > rsb > --> can't wait for that '040 board running at 40Mhz!! Nudge, Nudge, Wink, Wink.. We'll have to see about that one. I can safely say right now we're waiting on Motorola for the chips... -- Dave Haynie "The 32 Bit Guy" Commodore-Amiga "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy Amiga -- It's not just a job, it's an obsession
dillon@POSTGRES.BERKELEY.EDU (Matt Dillon) (05/02/89)
Refresh has always seemed so silly to me... nobody does the smart thing, which is to *monitor* the address bus and only supply refresh for the rows not accessed by the computer. So, the more loaded your computer gets the fewer refresh cycles need to come in and steal time. That isn't too difficult a function to add in a VLSI refresh controller. -Matt
dbk@teroach.UUCP (Dave Kinzer) (05/02/89)
Matt's good idea aside (being too difficult for SSI hackers), I've always wondered why board designers don't use the lower address bits (which cycle faster, typically) for the row address. Apart from the considerations of page/nibble/column/etc mode, using the lowest address lines available will keep the DRAM capacitors more fully charged, and thus less suceptable to changing state on an alpha hit. Seems easy enough to do... | // GOATS - Gladly Offering All Their Support Dave Kinzer (602)897-3085| | // >> In Hell you need 4Mb to Multitask! << uunet!mcdphx!teroach!dbk | | \X/ #define policy_maker(name) (name->salary > 3 * dave.salary) | In article <8905012101.AA25658@postgres.Berkeley.EDU> dillon@POSTGRES.BERKELEY.EDU (Matt Dillon) writes: > Refresh has always seemed so silly to me... nobody does the smart >thing, which is to *monitor* the address bus and only supply refresh for >the rows not accessed by the computer. So, the more loaded your computer >gets the fewer refresh cycles need to come in and steal time. > That isn't too difficult a function to add in a VLSI refresh >controller. > -Matt
tomb@hplsla.HP.COM (Tom Bruhns) (05/02/89)
> Refresh has always seemed so silly to me... nobody does the smart > thing, which is to *monitor* the address bus and only supply refresh for > the rows not accessed by the computer. So, the more loaded your computer > gets the fewer refresh cycles need to come in and steal time. I don't think it's fair to say _nobody_ does this -- but darned few do. Of course another trick is to refresh through something that has to happen to all rows anyway, like accessing for video pixel info (like I believe the old Apples did -- and I guess the C64, too). Then the refresh comes essentially free. > > That isn't too difficult a function to add in a VLSI refresh > controller. Agreed! Yet another very simple refresh technique: if your processor will be dead while awaiting the refresh completion anyway, and you can tolerate doing the once-per-whatever-millisecond refresh all at once, simply let the processor do it for you. I worked this out on a 6502 some years ago, letting a simple timer interrupt the processor (NMI, of course!) to enter the refresh routine. By very careful design you can get the time down to the number of RAS cycles required for refresh, plus perhaps a little overhead for the NMI/RTI. > > -Matt > ---------- Tom Bruhns tomb%hplsla@hplabs.hp.com