[ntp:questions] ESR looking for good GPS clocks
unruh at invalid.ca
Sat Mar 3 17:22:38 UTC 2012
On 2012-03-03, Terje Mathisen <"terje.mathisen at tmsw.no"> wrote:
> Michael Deutschmann wrote:
>> It's silly of him to ask only on his blog and not here, since many here
>> would want the same thing. The only difference between what he wants
>> and the NTP hobbyist's dream is that he requires a USB interface. An
>> NTP hobbyist would prefer raw RS-232 to cut conversion overhead, but ESR
>> wants to connect his reference clocks to router devices that only have
>> USB ports. He claims to know the USB conversion impact and that his
>> error budget can take it.
> USB is perfectly OK if the GPS device can use its internal 10 MHz clock
> to time the offset between the PPS position and the polling time.
> I.e. even if regular USB is limited to 1 ms resolution, it should be
> possible to improve this to a sub-us timestamp with such a protocol.
?? How does anything know what the offset is between the time the GPS
emits the PPS pulse and the time the computer receives it? The polling
of the pseudo DCD signal on the usb device is going to take time. The
encoding of the PPS pulse as a usb message is going to take time. USB
has two lines-- send and receive, over which everything goes. Thatt
pseudo DCD is simply going to be yet another usb message sent over the
usb lines-- waiting its turn (you cannot interrrupt some other message
already on the bus AFAIK) to be sent, then received, then decoded, and
then the computer polls the usb and finally timestamps it.
More information about the questions