Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!rutgers!mcnc!ece-csc!ncrcae!ncr-sd!hp-sdd!hplabs!hp-pcd!hpcvlo!john
From: john@hpcvlo.HP.COM (John Eaton)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Using the DMA (direct memory access) chip
Message-ID: <1610034@hpcvlo.HP.COM>
Date: Wed, 26-Aug-87 16:28:56 EDT
Article-I.D.: hpcvlo.1610034
Posted: Wed Aug 26 16:28:56 1987
Date-Received: Sat, 29-Aug-87 11:55:18 EDT
References: <1526@ihdev.ATT.COM>
Organization: Hewlett-Packard Co., Corvallis, OR, USA
Lines: 42

<<<<
< Yeah, but the bozo asks "how can you do 512 refreshes in 2 ms if the
< interval is 15 us?" and then doesn't answer the question. I don't know
< if it was because he forgot or if he just didn't know the answer. In
< either case, I am unimpressed. 
----------
 At least the "Bozo" knew enough about the PC to realize that you can't
use the Memory-Memory transfer mode of the 8237. I have seen some articles
that claimed you could use it.

 His problem with the refresh was that he didn't realize that the PC's
rams only needed 256 refresh addresses in a 4 ms time period. The 15 us
refresh works fine.

 As far as articles about the 8237 it was better than most. He did have other
problems with:

1) Simultaneous Memory and I/O strobes. An important distinction of the 8237
   is that it drives two strobes at once. This was not mentioned in the 
   article.

2) DMA Wait states. The article mentions that you can insert wait states in
   a DMA cycle. Well you can but it doesn't do you any good. On a IO Read-
   Memory Write cycle the PC's ram will latch the data from the falling edge
   of MEMW~ irregardless of how wide the cycle is. If you can't meet the base
   system timing then adding wait states won't help.

3) DMA page registers. Contrary to popular belief the page registers for
   channels 1,2 and 3 are not at 81,82 and 83. They are at 83,81 and 82.
   This is a small price to pay for something that probably saved two inches
   of PC trace and a couple of feedthru's.

4) The author claims that a "8237-5 in my Clone digests 7.37 MHZ quite
   happily". Clocking a 8237 faster than 6 MHZ will produce a DACK pulse
   width that falls below the mimimun required for a 765 disk controller
   in verify mode. This is why FORMAT fails on many PC's that are speeded
   up.



John Eaton
!hplabs!hp-pcd!john