[ntp:questions] How bad is USB?
David J Taylor
david-taylor at blueyonder.not-this-part.nor-this.co.uk.invalid
Sun May 10 10:45:43 UTC 2009
Uwe Klein wrote:
> David J Taylor wrote:
>> Richard B. Gilbert wrote:
>>> You can try it and see what happens! The results may be
>>> sufficiently good for your purposes. You will almost certainly not
>>> get microsecond accuracy. If you are willing to settle for +/- 10
>>> milliseconds you can almost certainly get that.
>> My first tests show within a millisecond or so:
> Do I see a (sawtoth) beat signal between "true" time and the
> nominally 1ms polling interval from the usb bus?
I don't think so. If you mean the sawtooth on the left, peaking at 1600
the first day and 0800 the second, that's pure temperature variation. PC
Narvik was not switched to USB/serial until about 1530 on day 2. There is
an increased fine-grain variation in the offset after the change, but the
jitter reported by NTP has dropped from 110-140 microseconds to about 45
microseconds after the switch from LAN-sync to USB-serial-ref-clock.
There are more performance graphs here:
More information about the questions