Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!mcvax!ukc!harrier.ukc.ac.uk!jbww From: jbww@ukc.ac.uk (J.B.W.Webber) Newsgroups: comp.sys.atari.st Subject: Re: Slots for the (Mega) ST - Transputer Links Summary: Try Transputer Links, for connection to apparatus Keywords: apparatus, experiments, control, equipment, Transputers Message-ID: <1963@harrier.ukc.ac.uk> Date: 4 Aug 89 14:22:23 GMT References: <8908010836.AA07638@nh.cs.wm.edu> Reply-To: jbww@ukc.ac.uk (J.B.W.Webber) Organization: Physics Lab, University of Kent at Canterbury, UK. Lines: 103 In article <8908010836.AA07638@nh.cs.wm.edu> fjmora@CS.WM.EDU (Fredric Mora) writes: >In Info-Atari16 Digest V89 #344, Larry Rimal writes: >>As a side note, hardware hackers love the ST. Easy to modify, tough >>as nails, and quite versatile, it is a joy to play around with. All that >>is missing are slots! > ^^^^^^^^^^^^^^^^^ > >I definitely second that. In fact, I sent a mail to John Townsend at Atari >about the possibility of Atari making an expansion box for the Mega ST. Such >a box would have a slot plugged in the Mega internal connector and reproduce >the connector (say, 4 to 6 slots) in a Mega-like box. >I think that such a peripheral is now needed by ST users .... > >John answered briefly: >>It has been suggested before.. however, I think it is harder than it >>sounds.. > >Certainly, I agree. > >First, the box would have to include a power supply (say, 65W. Costs $39 >in retail PC-compatible shops). Careful connection of the 2 grounds (the Mega's >and the box's) is required. That can be done, at least by dealers. > >Second, we are talking about plugging a connector into the internal Mega slot >and sending the signals on a backplane. This requires buffering the >68000 bus and shielding the box. The designers would probably need to alternate >each signal with a grounded trace on the printed circuit board for better >noise immunity. Also, mechanical shielding problems may appear and need >to be addressed carefully because of the FCC standards. I don't think that >this is a major problem..... > > O.K., can't resist putting my ha'porth in, here. First I must say that I am as keen an enthusiast for rack systems as you will find. I have been designing and building laboratory instrumentation hardware/software modules into 19" racks ever since the 6800 and 8008 first appeared, using a number of processors (including the 68000) and maths processors. (Yes, Virginia, before the 68000, came the 6800, 6802, 6809 ...) However, on my desks there are now STs - how come ? I guess the answer is that a rack costs - not just money, but also time, documentation, care of edge connectors, construction of multi-way cabling to instrumentation, etc. Does one site the computer at the experiment (which part of it), or take the signals to the computer ? ( I have seen some people's experiments end up with half a dozen computers scatterd over a room.) So, is a rack system such as is proposed for the ST possible ? Sure. Is it optimal ? Perhaps, five years ago, not now. (See above, for reasons.) For a start, the 68000 is dirt cheap (a few pounds); by the time one has payed for all the hardware, one might as well add a processor. Now one is talking about another computer, so how are the two to communicate ? I use serial port/midi port quite frequently; these are often enough for sending commands and occaisional data. For good data rates, one wants something more; the ST DMA port is a natural. I have looked closely (logic analysers, etc) at the ST DMA port, but was not able to get solid enough documentation to spend time on it. I may yet return to using this, but one of its real limitations is that it only allows for one master. Thus a SCSI bus interface will probably make more sense, if one is after a high performance multi-wire systems interface. (I do have IEEE488 available from my racks, but don't fancy that I now have the time to write the ST end of a multi-tasking driver.) A SCSI approach is still quite complex, and pricey; I have only just upgraded to a SCSI hard disk on the ST at work. So, what path am I following ? Transputer Links. For those who have not met them, Transputers are a range of processors, (16bit to 32Bit), with 4 communications links integrated on the chip, together with, in particular versions, other resources - floating point processors, disk handlers, varing amount of memory, etc. Links run at 5, 10, 20 Mbit.s-1 (selectable), and an 18 pin DIL chip is available to provide a link-to-8bits-parallel interface. This makes them cheap enough to put with single chip controllers, to interconnect instrumentation dispersed around the experiment. The link connection is just two twisted pairs, one for each direction. Most importantly, a cartridge plug-in for the ST can be done for well under 100 pounds. O.K., the ST can't support 10 MBits in/out from the cartridge, but it is good enough for most things (all transfers are acknowledged, so data is not lost). The original version of the ATW (the Mega supported ABAQ) provided a link adaptor on the Mega internal 68000 bus connector; this is undoubtedly a very good way of doing it. For high performance interconnection and greatly increased processing power, one can add transputers at nodes; each link may only have two ends - this follows from the fact that transfers are acknowledged. I guess what I really like is just the two twisted wires, and cheap interface. 10 Mbit.sec-1 is about 1MByte.sec-1, and I don't try and send data at 1MByte.sec-1, down more than a few meters of multiwire cabling. The rest is all bonus. Hope this gives people some ideas; sorry for rabbiting on for so long, cheers, beau webber jbww@ukc.ac.uk University of Kent, Physics Lab.