Chuck wrote on 8-10-2007 3:12: > i am going to be messing around with the garmin gps18-lvs for use with ntpd. hi Chuck, I've just ordered two 18lvc's to play with. Seems like we have the same in mind. > i wish to modify the wiring instructions given here: > http://time.qnan.org/ Saw that page a few weeks ago. Frowned on the idea of attatching unbuffered leds to a time critical signal. Semiconductors do have a capacity and I've never heard of high frequency leds. Apart from that a pcb is overkill for two leds and a fuse. > i will eliminate the circuit board and leds Good, that's settled then. > instead opting for an inline fuse I would dump the fuse to. Even a fast blow fuse is much to slow to protect the electronics of the usb port you use as power supply, so it's useless. Fuses are to prevent fire in the home. In a power supply the fuse goes between the mains connector and the transformer, not on the output. Besides, I've never seen a fuse in an usb device, I suppose usb ports have an electronic overcurrent protection. > if i use an additional wire to extend the pulse, ground and power for no more > than 3ft in length, to power the 2 leds in their own box that can be attached > to the front of the computer to make the leds more easily visible, would this > extra length of wire create any kind of stub that would change the leading > edge pulse slew rate enough to cause it to become less accurate? or need i > only concern myself with such a thing in much higher frequencies? Ground and power doesn't matter of course. Also the PPS is no problem over short distances. To be safe you could use cat 5E UTP (that is the stuff you have probably all over the place), that's designed for the rather high frequencies of 100 Mbit ethernet, and cost nothing. UTP has eight strands, twisted together in four pairs. Pick two that belong together for PPS/ground. Again, I have my doubts on that led on the pps. On the other hand, the 18LVC has a precision of only a microsecond, maybe it works. In theory the led will delay the pulse (or even ruin it), but I can't find any datasheet specifying the capacity of a led, so just try... > also, while i am asking dumb questions, do you think adding rj45 sockets > inline to allow using a standard ethernet cable to lengthen the wire from the > gps18 to the usb/serial junction would cause any problems? i would not > consider extensions of more than 6-9 feet. RJ45 is fine, even better than the standard DB9. The same for UTP; is better than standard untwisted serial cable.
Extending RS232 is no problem at low baudrates. I've been running RS232 over more than 200 meter at 9600 baud (we didn't have ethernet yet at the office in the seventies, and two modems was quite expensive and only 300 baud). The default baudrate for the 18lvc is 4800 baud. A meter is 3 feet I believe, so just go ahead. Did you realize that once you use the pps, the rs232 is not time critical anymore. The pps defines the UTC second boundaries, ntpd needs the nmea data only to "label" the seconds. You could even forget about the rs232 and use a ntp server on the internet for that "labeling". By the way, in datacenters it's common practice to route RS232 (terminal servers etc.) over the UTC/patchpanel/RJ45 infrastructure they have all over the place. Works like a charm. There's even a smart wiring standard you may want to follow, called Yost: http://yost.com/computers/RJ45-serial/. (There is also a not so smart ISO standard that nobody uses.) Sorry for being so verbose. regards, Jan Hoevers _______________________________________________ timekeepers mailing list [email protected] https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers
