Hi The only other wrinkle worth noting is that you may have to delay the carries out of the count blocks. All of the same stuff still applies, you just wind up with two delays where you thought you would only need one.
Bob On Apr 14, 2010, at 6:11 PM, Magnus Danielson wrote: > jimlux wrote: >> Eugen Leitl wrote: >>> Hi -- a couple somewhat lunatic questions. Figured this would >>> be the best place to ask. >>> >>> Anyone aware of a time standard which compensates in regards >>> to an ideal flat-spacetime-at-rest-relatively-to-cosmic-background >>> reference clock? I realize this is not relevant for any affordable >>> clocks. >>> >>> Unrelated, anyone aware of a hardware (IC) counter which >>> counts in a (local-bitflip) Gray code, so it can track a very >>> fast (integrated?) oscillator without dropping cycles? >> Any synchronous counter won't drop cycles, right? only ripple carry >> asynchronous have the propagation delay problem. >> However, since long synchronous counters are complex, another way to do it >> is with a linear feedback shift register (LFSR). Rather than doing it as >> usually drawn, with taps all getting xored to generate the single feedback, >> you do it using the output and feeding it "into" the stages (with an Xor) >> where the taps are. > > If you want a long synchronous counter, then you can let the lower speed > parts ripple a clock-cycle after the bottom (high-speed) part and delay the > high-speed values with a DFF to bring the values into sync. Takes a bit of > thinking about the details, but you can get about any length of synchronous > result this way. Only problem comes if you reset it, in which case the reset > pulse needs to be applied in a suitably timed fashion. Repeating this trick > and you can get arbitrary lengths of synchronous length counters. > > Cheers, > Magnus > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
