INMOS TN70 - Connecting 100MBaud T9000 Transputer Links > 자유게시판

As long as the errors are infrequent (one each a number of thousand years), this is fully satisfactory. The problem is just not in the attenuation within the cable, but within the losses (and consequent costs) in converting from electricity to gentle at one finish and from light to electricity at the other end. Copper wire has a finite resistance: 28AWG wire is likely one of the smallest cross sections in widespread use and has a resistance of 0.23Ohm/m, 1Ohm in 4.3m. If the characteristic impedance of the cable is 100Ohm, a resistance of 10 ohms is not going to have an effect on the sign very much, so this cable should actually be usable at 43m. The problem is that at high frequencies, the sign doesn't movement evenly throughout the conductor but concentrates at the surface of the conductor - the skin impact. For PCB connections up to 20cm, the characteristic impedance of the PCB track does not matter.

Longer distance connections, with the quantity of amplification required for the optical sign, is such that the connection must be considered as much less reliable than regular quick connections on a PCB. In truth the indications are that it may be possible to achieve worst case error charges of the order of 10−20, much better than is achieved by normal communications. DS links will probably be specified, therefore, in order that they offer such infrequent errors that the hardware can be thought-about reliable, and that the likelihood of a single bit error is so low that the chance of an undetected double error is negligible. Components with wavelengths of 820 or 850nm are in many respects more suitable for 100MBaud transputer links. Remember, when studying these texts on communications, low voltage twisted pair cable that (while the rules involved have to be understood) the distances required and the error rates obtained make transputer links a lot simpler than telecomms. I hope you guys are clamping your voltages when reading your sensors because minute mV modifications can happen attributable to RF and induced flux from the ignition leading to false reads.

Posted by zip on June 17, 1999 at 14:27:19: Just reread my post and sure your proper voltage clamping is used to forestall CMOS from going over their threshold worth. Posted by NITRO on June 17, 1999 at 14:45:10: Occasionally a number of prototypes are made to be easily accessable and are subsequently somewhat "ugly" in configuration. A few instance ’scope traces are proven of the waveforms seen with completely different lengths of connection and with completely different forms of buffering. For example workplace gear such as terminals, laser printers, disks, and fax machines, every of which might use from two to 4 of the connectors, could use one sort of connector; and computer systems, which might use many extra connectors, use a distinct kind. An instance LED outputs (infra-purple at 1300nm wave size) 0.25mW of optical power when pushed by 100mA of electrical power. The D and S pair of signals ought to be approximately the identical length, but a difference in size of 50mm would only introduce a skew of 250ps, which should be totally acceptable.

Any distinction will probably be seen as noise. In magnetic recording, such differentiation occurs naturally, however it brings its own issues; any noise reminiscent of crosstalk is coupled by the differentiator, and any AC imbalance in common-mode coupling is translated into extra noise. Our present proposal for cable connection is for DC coupling with the forty one series buffers talked about earlier, with extra circuitry to enhance the (already good) common-mode characteristics of those buffers. For longer connections, up to 200 or 500m, or for electrical isolation, our present proposal is to make use of low cost optical fibre components, with a goal designed interface chip. Even with a very huge bandwidth, it is possible to make use of tuning (tweaking?) to compensate for the frequency characteristics of the cable. The dotted line represents the DC threshold of the sign, which suggests that the signal will not be received appropriately, even when there is no such thing as a noise. There are several other mounting combinations, that are already undergiong assessments as I converse.