Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!sundc!pitstop!sun!amdcad!ames!hao!noao!mcdsun!nud!rover!mph
From: mph@rover.UUCP (Mark Huth)
Newsgroups: comp.sys.amiga
Subject: Re: Monitor mishmash
Message-ID: <605@rover.UUCP>
Date: Thu, 29-Oct-87 11:03:28 EST
Article-I.D.: rover.605
Posted: Thu Oct 29 11:03:28 1987
Date-Received: Wed, 4-Nov-87 04:17:02 EST
References: <3248@ccicpg.UUCP> <2616@cbmvax.UUCP>
Reply-To: mph@rover.UUCP (Mark Huth)
Organization: Motorola Microcomputer Division, Tempe, Az.
Lines: 24

In article <2616@cbmvax.UUCP- daveh@cbmvax.UUCP (Dave Haynie) writes:
-in article <3248@ccicpg.UUCP>, harald@ccicpg.UUCP ( Harald Milne) says:
-> 	I would dicker with the not cheap comment. Ever heard of interleave?
-> I also made this comment, until I realized that if you delay ram timings
-> through interleaving, You don't need fast rams.
-
-I think maybe you do.  Consider for a moment the Amiga.  At 640 across by
-4 deep, the Agnus chip is fetching data for Denise as fast as a 3.58MHz 
-DRAM can supply it, or real close.  If you want to double the scan rate
-(why you're building this thing in the first place), you're now fetching
-video information at 7.16MHz.  Or, put another way, one fetch every 158ns.
-And that's not leaving any time to put the data in there, via interleaving.
-Now you're down to about 79ns.  I think the video DRAM will be cheaper than
-the high speed static RAM you'd need for this otherwise.

I think that's where the interleave comes in.  Instead of fetching
sequential samples from the same ram chip, you fetch seqential samples
from different banks of ram.  Maybe I have four ram banks, so each ram
is used every 316ns.  With sequential addressing like you have in
video display applications the interleave technique works real well.
Just use some of the low order address bits to generate the chip
select signals.

Mark Huth