[ntp:questions] Using NTP to calibrate sound app

unruh unruh at invalid.ca
Mon Jan 28 22:07:39 UTC 2013


On 2013-01-28, Rob <nomail at example.com> wrote:
> no-one at no-place.org <no-one at no-place.org> wrote:
>> On 28 Jan 2013 09:57:00 GMT, Rob <nomail at example.com> wrote:
>>
>>>There will be the extra complication that a UMTS connection goes into
>>>a "sleep mode" when there is little or no traffic.  To "wake it up"
>>>you need to send about 4 packets/second for some time.  This is a
>>>very typical pattern on a UMTS connection:
>>>
>>>ping -i1 172.31.32.18
>>>PING (172.31.32.18) 56(84) bytes of data.
>>>64 bytes from (172.31.32.18): icmp_seq=1 ttl=127 time=461 ms
>>>64 bytes from (172.31.32.18): icmp_seq=2 ttl=127 time=377 ms
>>>64 bytes from (172.31.32.18): icmp_seq=3 ttl=127 time=390 ms
>> ........
>>>64 bytes from (172.31.32.18): icmp_seq=9 ttl=127 time=246 ms
>>>64 bytes from (172.31.32.18): icmp_seq=10 ttl=127 time=144 ms
>>>64 bytes from (172.31.32.18): icmp_seq=11 ttl=127 time=50.8 ms
>>>64 bytes from (172.31.32.18): icmp_seq=12 ttl=127 time=52.2 ms
>>>64 bytes from (172.31.32.18): icmp_seq=13 ttl=127 time=102 ms
>>>64 bytes from (172.31.32.18): icmp_seq=14 ttl=127 time=50.4 ms
>>>64 bytes from (172.31.32.18): icmp_seq=15 ttl=127 time=51.4 ms
>>>64 bytes from (172.31.32.18): icmp_seq=16 ttl=127 time=52.4 ms
>>>64 bytes from (172.31.32.18): icmp_seq=17 ttl=127 time=47.6 ms
>>>64 bytes from (172.31.32.18): icmp_seq=18 ttl=127 time=50.1 ms
>>>64 bytes from (172.31.32.18): icmp_seq=19 ttl=127 time=51.4 ms
>>>64 bytes from (172.31.32.18): icmp_seq=20 ttl=127 time=47.4 ms
>>>
>>>As you can see, totally unsuitable for accurate time calibration.
>>
>> Would I likely get better results if I advised my users to connect to
>> broadband wi-fi if possible and I use that instead of 3G data?
>
> WiFi has the same problem on a slightly smaller scale.
> With WiFi the pingtime jitters by 50 ms or so.

But what matters is not the ping time jitter, but the assymetric travel
time jitter. On wifi that is nearer 1 -10 ms 



More information about the questions mailing list