From gareth at dontspammeatsimtec.co.uk Sat Nov 11 20:47:00 2000 Path: pegasus.csx.cam.ac.uk!server1.netnews.ja.net!fu-berlin.de!feeder.qis.net!dispose.news.demon.net!demon!news.demon.co.uk!demon!simtec.demon.co.uk!simtec.co.uk!gareth From: Gareth Simpson Newsgroups: comp.sys.acorn.hardware Subject: Re: 2MB VRAM - Compatiable with StrongARM or not? Date: Thu, 09 Nov 2000 09:46:57 +0000 Organization: Simtec Electronics Sender: gareth at simtec.co.uk Message-ID: References: <8ubhqf$eks$1@nnrp1.deja.com> NNTP-Posting-Host: simtec.demon.co.uk X-NNTP-Posting-Host: simtec.demon.co.uk:158.152.146.181 X-Trace: news.demon.co.uk 973771347 nnrp-09:26828 NO-IDENT simtec.demon.co.uk:158.152.146.181 X-Complaints-To: abuse@demon.net X-Newsreader: Messenger v1.41d for RISC OS X-Posting-Agent: RISC OS Newsbase 0.60m X-NNTP-Poster: NewsHound v1.32 Lines: 111 Xref: pegasus.csx.cam.ac.uk comp.sys.acorn.hardware:82105 In message Paul Robinson wrote: > In article <8ubhqf$eks$1@nnrp1.deja.com>, > wrote: > > Hi! > > > > I was just wondering how you know if a 2MB VRAM is compatiable with > > StrongARM or not? (Or are they all compatiable)? > > > > Thanks! > > > > Matt > > Old Samsug VRAM causes timing problems. > There is a lot of misinformation being spread about VRAM. Here are my favourite five reasons why a VRAM sometimes doesn't work: 1) Original prototype VRAM from Acorn was fabricated on 1.6mm thick PCB with a tin plated edge connector rather than 1.27mm with gold flash. The extra thickness, besides making it difficult to insert, distorts the socket and reduced the contact reliability. If a replacement VRAM is then inserted, the over-stressed contacts on the socket don't make proper contact with the thinner VRAM upgrade. The tin/lead plating of the early boards is also prone to oxidization, and as the oxide takes up more volume than the original metal, it has a tendency to push the contacts apart, exposing more contact area which itself begins to oxidize. Oxidized contacts also increase the contact resistance so the capacitative effects of the socket, tracking and device inputs are amplified and the signals become attenuated. The lower the signal drive of the processor, the more pronounced these effects become when trying to achieve full TTL level logic swings. Oxide can be removed with IPA or a "Mars" plastic rubber. Do not use anything abrasive and *never* use switch cleaner. Switch cleaner relies on a wiping action of the switch to maintain a good electrical contact. The VRAM doesn't move (except for the slight movement of the motherbaord due to thermal expansion and vibration). Anyone who uses WD40 needs to be shot ;-) 2) Some of the early VRAM modules used Samsung 80ns parts. For correct and reliable operation, 70ns or faster parts need to be used. The fault happens not because the product is Samsung, but the fact that the access time for the device is too slow. The weaker data-bus drive of the StrongARM increases the settling time of the bus so combined with the longer access time for these slower parts, it results in read/write failures. This especially applies to DRAM modules, which require 70ns or faster FPM devices or 60ns or faster EDO to meet the required access times. Because of the differences between FPM and EDO, when used in FPM mode, EDO must be de-rated to achieve the same performance. 3) When Acorn changed the design of the RiscPC motherboard to cope with 16 bit sound and the ARM710, they introduced a track difference to the VRAM signals that caused ringing problems with the currently shipping VRAM. This was rectified by changing IC1 for another manufacturers device to alter the drive characteristics of some of the key signals to ameliorate the ringing problem. That's why you quite often see Acorn VRAM with an "odd" IC1. Reverting to the old layout or modifying the VRAM layout would have solved the problem too. 4) The VRAM sockets, because of their wiping design, do not lend themselves to long term contact reliability. Unlike the piercing high-force "point" contact used on the DRAM module sockets, the VRAM contact is done using a curved pin and relies on a light spring force to maintain light contact over a very small area. If the VRAM module or socket has contamination on the mating surfaces, then this will impair the contact between them and as explained above, can cause oxidation to the exposed surfaces especially if the gold plating is porous. In damp conditions, contaminants can then promote a galvanic reaction between the plating materials of VRAM and socket causing the contacts to corrode. 5) Lastly, a heavily loaded bus will cause problems. The impact can be minimized by reducing loading on both the data and address bus of the DRAM and VRAM. Always choose modules with the lowest device count, and if possible, try to match your DRAM modules. If the tracking of a pair of modules is completely different then it creates an impedance mismatch which increases ringing and reduces the signal integrity. Swapping modules round can help but it's better to change them for ones with a lower device count. Now for the big plug..... *sales mode on* Anyone having problems with their VRAM may be interested that Simtec have recently brought out a new 2Mb VRAM module that builds on the high reliability of its predecessor. The new module reduced the chip count to four from the tradition eight used previously on all other VRAM upgrades. This has the advantage of reducing bus loading, power consumption, and component count. The overall size reduction enabled us to produce a cleaner layout for the design which minimizes track length and impedance mismatch on the VRAM bus. It has certainly made the PC/SA systems at Simtec run more reliably ;-) *sales mode off* Gareth. --------- Gareth Simpson Design Engineer, Simtec Electronics.