Path: utzoo!attcan!utgpu!watmath!att!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!mcvax!ukc!acorn!moncam!emmo From: emmo@moncam.co.uk (Dave Emmerson) Newsgroups: comp.sys.atari.st Subject: Various topics Keywords: memory multitasking decoupling mmu Message-ID: <267@marvin.moncam.co.uk> Date: 9 Aug 89 21:18:04 GMT Organization: Monotype ADG, Cambridge, UK Lines: 63 First, an apology. Due to a misunderstanding between Harry & I, Harry recently posted that we might run a second batch of the 2/4 Mbyte expansion board. Regrettably, that's a no-go. The batch we had made were cheap as a returned favour from the PCB makers. Also, we do not wish to damage those companies who make their living from selling extras for the home computer market. Their presence is vital. As an aside, I DO think they should offer THEIR bare boards for the DIY people, who will otherwise resort to 'piggybacks' or prototyping boards. If anyone else can get a cheap batch run, and will undertake to sell at around cost price, the photography is still available. Remember, this board is for ZIGZAG drams.I've been following the multitasking debate with some interest. I disagree generally about the MMU being ESSENTIAL, Sinclair's QL (forget the grotty keyboard & microdrives) made an excellent job of it using nothing more than a 50Hz interrupt & a good OS. Sure, a rogue task can scribble all over everything, but even a MMU wont stop bad code from causing total system failures, only one particular type, and only for a limited number of tasks. Incidentally, somebody (not me, I'm off for a week [sighs of relief] ) should post a short article explaining to the less hardware-oriented types, what a memory management unit does, so's they can all join in (or at least understand what's being said). The participants in the decoupling capacitor debate have both omitted to mention that the most critical parameter when dealing with fast changing signals is the INDUCTANCE of the connections between the IC's power pins and the decoupling device. I read the results of tests made with an American product (Q-CAPs ??) which were large thin capacitors designed to fit under the chips, connecting directly to the pins with NO inductive path between the capacitor and the pin. As a result, you could reduce the capacitor by a factor of 10 and still see an improvement. I forget who actually makes them, but RS sell them in the UK, one is their ref 126-269. The point is that they can also be retro-fitted ON TOP of the chip, if you have problems. I was taught selection of the value of a damping resistor is a function of the impedance of the track and the rise time of the signal. John M Logajan is quite correct, DON'T change it unless you alter the track. Hitachi (as I recall) suggest that PCB track width and thickness be selected so that this will normally be within the range of 25 to 50 ohms, but I doubt if Atari chose 68R randomly. Piggybacking chips will not impact greatly on the track impedances, adding flying leads will. Continuing the discussion prompted by jbww@ukc.ac.uk (J.B.W.Webber) >(ever done a recursive directory listing on a network of Suns?), but, Yes, and symbolic links make it even messier but I wouldn't want to do away with them... Seems like what you need is a nonvolatile ramdisk - How about bubble memories, I haven't heard much about them for a while. Looks like you may have to be patient. Still, it didn't take long to get from a 1K ROM, 1K RAM system to where we are today... No more sounding off for a week, I promise! Dave E.