Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!usc!apple!coherent!dplatt From: dplatt@coherent.com (Dave Platt) Newsgroups: comp.sys.mac.hardware Subject: Re: IIfx upgrade and SCSI termination for int. hard disks Message-ID: <66928@coherent.coherent.com> Date: 12 Aug 90 05:21:23 GMT References: <3086@dftsrv.gsfc.nasa.gov> <1990Aug11.042741.22709@Neon.Stanford.EDU> Reply-To: dplatt@coherent.com (Dave Platt) Distribution: na Organization: Coherent Thought Inc., Palo Alto CA Lines: 77 In article <1990Aug11.042741.22709@Neon.Stanford.EDU> kaufman@Neon.Stanford.EDU (Marc T. Kaufman) writes: > I have been confused by this from the beginning, and now that the terminator > is documented (TN 273) I am even more confused. > > It appears that the 'fx' termination differs from a standard external > termination in that it has extra capacitance from +5 to ground, to prevent > induced crosstalk between the signal lines (by stiffening the power supply > with the 2.2uf cap and bypassing high frequency spikes with the .01uf cap). > Now an internal terminator should be relatively close to a LOW impedance > source of +5, so the problem should never arise. Why, then, must one > remove the internal terminators and replace them with external ones? Well, as I read it, the IIfx SCSI interface is so very fast that a TERMPWR bounce can occur unless the decoupling capacitance is _very_ close to the common TERMPWR point. Most internal terminators I've observed are mounted to sockets on the PC board, and do not necessarily have decoupling caps located directly by the sockets. There is usually several inches of PC-board trace lying between the terminator sockets and the power-supply jack, and several more inches of wire cabling leading back to the power supply itself. This amount of trace and wire may introduce enough of a time-delay so that the power-supply filters can't prevent the TERMPWR line from glitching down significantly in a worst-case SCSI-bus state transition. At these high speeds, even the power-supply and ground lines must be considered as transmission lines! Adding capacitance at the common TERMPWR tie-point is a surer way of keeping the TERMPWR level stable. Now... as I read Tech Note 273, it is NOT necessary to remove the termination resistors on an INTERNAL hard disk in a Mac IIfx. The IIfx does come with an "internal terminator" block... one which consists of a set of termination resistors, and a separate capacitive filter block. If you don't have a hard disk in your Mac IIfx, you need both of these parts... they ensure that the SCSI bus is terminated properly at the Mac IIfx end. If you have an internal hard disk in your IIfx, you should continue to use the termination resistors already installed on the hard disk... and you should plug the capacitive filter block in between your hard disk and the Mac's internal SCSI port. This will add some extra capacitance to your hard disk's TERMPWR line, and prevent glitches from occurring within the Mac. This block "can and should" be installed even if your internal hard disk has capacitors buffering the TERMPWR line... it won't hurt, and might help. A similar amount of capacitive filtering must be added at the far end of the SCSI bus, to make sure that the far-end terminator doesn't suffer from surge nausea. The problem is probably most acute if you're using an external terminator (non-filtered) which is getting its power from the Mac II over a long cable, and isn't receiving power from other SCSI devices on the bus (not everybody provides terminator power!). The problem might also be apparent if your bus ends with a device that has an internal terminator, but which doesn't have an adequately stiff or fast power-supply connection to the terminator. In the latter case, you _could_ get around the problem by adding capacitance within the device at the end of the bus... but this would require soldering an extra capacitor or two onto the PC board, and would almost certainly violate the warranty, damage the device, or get Senators Helms or Proxmire upset. It's safer to simply remove the internal terminators, and install a filtered external terminator (graciously provided by Apple). It's a bit of a shame, though, that Apple decided to stick with the 220/330-ohm parallel termination scheme, and simply add a capacitor to it to provide some voltage stability. A more sophisticated termination scheme has been developed, using serial termination and a regulated voltage derived from the TERMPWR line. This newer circuit provides a better match between the cable impedence and the terminator, consumes less power, and is equally resistant to termination-voltage sag (perhaps moreso). It's also quite compatible with existing SCSI devices and drivers, and is recommended in the SCSI-2 standard. Of course, it's more expensive, as it requires several active components... so I can't blame Apple for sticking with a simpler solution. I might buy or build one of these, though.