I am looking for an instrument that somehow measures SCSI cable
*quality*. Beyond measuring continuity on each pin, I need a way to define and measure the quality of the cable while it is actively in use backing up data. Our application is tape drive backup in a computer to tape drive configuration. We run SCSI HVD, sometimes over very long distances, and we have seen situations in which a marginal cable quickly becomes the source of problems. I hate having to discover this by trial and error (e.g., replacing cables, buying a different vendors' cables, etc).
Because we are doing this in a commercial setting, I don't mind investing a few thousand dollars in a nice piece of test equipment. The problem is that I lack sufficient understanding of concepts and what we should be trying to measure. Any advice on concepts or specific pieces of test equipment that would help, is understood.
It would be highly desirable to have the equipment be something that can be inserted into the SCSI cable chain and report back statistics in a clear enough form that a technician who is not an EE would understand the data.
It has been a while since I used SCSI seriously, but, from what I recall: SCSI comes in differential (high voltage and low voltage) and single ended.
For single ended, the biggest problem was external noise. I can't, off hand, think of a good meaningful way of you testing for cable noise immunity, so it is really down to buying cable with good, two layer, shielding.
For both cables, the other problems are to do with cable geometry - things like the inner core cable pairs being used for control signals and outer for data. (The ECC of most systems is better able to compensate for data corruption than control signal corruption). Again, how you test for that is difficult to say - you just have to buy the correctly laid-up cables and use the correct cores. You could chop the end few inches off a reel, strip it apart and make sure that the right pairs are in the right place. But, IME, they will be - unless you buy very cheap cable, incorrectly sold and not to ANSI standard.
You can test things like the characteristic impedance of the cable - shove a terminator on one end and measure the impedance that you see at 10MHz at the other. You can get SCSI testers that measure the reactance of unterminated cable, which is much the same test. That can show that a cable hasn't been laid up properly, but I have never come across an incorrectly laid up cable (as far as I know...). A fixed-frequency impedance test set isn't going to be expensive.
But, IME, cables companies are pretty good and the figures they quote for their cable resistance and capacitance are pretty accurate.
Cross-talk between pairs is, IME, not so much a problem and generally simply illustrates that the end connectors have (both) been wired up incorrectly. A pin-pin tester won't show that identical errors have been made at both ends, so say, a data cable has been twisted with another one and not its own return. An impedance or reactance tester will light up like a Christmas tree under such circumstances..
Now I don't know of any equipment that is actually designed to insert in a SCSI chain, purely for measuring the quality of the signals. However, there are loads of kit that do it incidently. Most SCSI controllers capture oodles of information about incomplete control signal sequences, frame faults, retry counts, bus contention and so on. So, all your technician may need is the diagnostic test software that is intended to monitor SCSI controllers. Take a reference set of figures with a "good" cable and then compare them to the cables you want to test. As I mentioned, control signal problems are the ones to watch for becuase they really screw things up. A few lost data packets here and there are easily recovered by the controllers. That can be very cheap (eg free) to quite expensive, depending on the controller and software that you use - many controller manufacturers throw such software in for free, either with the controller or on their web site.
Hope that helps a little. Generally, if you cut the end off a bit of SCSI cable and examine the core lay-up, a quality cable shows. The easier it falls to pieces as you strip it apart, the cheaper and nastier it is..
Me thinks you could be walking off into the darkness. I do that frequently. I am unaware of any "quality measurements" for ribbon cable. Good concept and question. I see the merit of it.
A Time Domain Reflectometer? That is generally a decent test of insulation quality and integrity of conductor, at least for power cables. Might it not also work for a SCSI cable?
It would certainly show up any dramatic faults in the cable (opens and shorts), "bad" connectors and so on. Depending on the TDR used, it would also show up any significant variations in characteristic impedance - provided the cable was correctly (passively) terminated and also allow the impedance to be measured.
I am not sure how you use one to test insulation quality - other than detecting that the insulation has actually failed at one or more points!
Just sticking the cable, with and without a terminator, on an LCR bridge excited at 10MHz or so should do the trick.. Both that and using a TDR would need a lot of patience and technician's time swinging through each and every pair on each cable though.
We used to derive equivalent three terminal models for multi-way cables by doing open and short cicuit tests - the short circuit was most easily achieved for a multi-way cable by shoving the cut end of the cable in a pot of mercury... I somehow don't think H&S rules allow that any more...
PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.