[ntp:questions] WiFi & NTP.

Rob Kimberley time.bandit at btinternet.com
Wed Aug 29 16:47:24 UTC 2007


Have NTP 4.2.0a running on 2 Windoze machines over my wireless router to 2 
Stratum 1 GPS servers. The thing I notice is the jitter figure which can be 
several milliseconds compared to a wired ethernet connection.

Not investigated it further, but I think this is fairly common problem with 
wireless routers.

Rob Kimberley

"Guy" <gdelavil at ens.insa-rennes.fr> wrote in message 
news:1187947307.203686.312630 at m37g2000prh.googlegroups.com...
> Dear all,
>
> I am a student from a French engineering school. I am currently
> working on a synchronization over WiFi project that could apply to
> embedded devices.
>
> For the moment, I have only reached a precision of 5 ms.
>
> (Whereas I get a precision of a few hundreds of microseconds with an
> Ethernet connection.)
>
> Nowadays the WiFi tends to be more and more integrated "transparently"
> into the other networks. And it might be interesting to consider and
> optimize the synchronization over a WiFi connection using NTP!
>
> So I was wondering if some of you had already tried to optimize a
> synchronization process over a wireless connection?
>
> If so, your feedback and suggestions are welcomed ;->
>
>
> Anyway, here is what I have done so far:
>
> . When it comes to channel congestion, I have already run a series of
> tests. I launched NTP in unicast (client/server) mode and at the same
> time I streamed data between the two devices. I got an average offset
> of 65 ms (computed as the average of the offset absolute values). So
> that's where some QoS came in handy to prioritize the NTP packets and
> give them the same priority as Network Control packets. Doing so, and
> still streaming data, I got an average offset of 15 ms. Without
> traffic on the connection and with the QoS rules, that's when I reach
> a precision of 5 ms.
>
> . I have also tried to select manually the WiFi channel so to avoid
> using channel 1 like everyone but it didn't seem to have much effect.
>
> . Concerning the RTS/CTS messages, the threshold was set by default to
> 2347. The NTP packets I send and receive only are 76 bytes long. I
> changed the threshold value to 1 (just to make sure the value is
> smaller than the actual NTP packet size and because 0 might disable
> something) to see if there are any effects. And it seems to improve
> slightly the synchronization.
>
> . A comment I read concerning the RTS/CTS delay drew my attention to
> some parameters of the WiFi driver such as the size of the congestion
> window, the srl, and lrl, I should use complementarily to my QoS
> rules. But I don't know exactly how?
>
> . By the way, I wanted to point out that in my situation the RMS
> jitter hardly goes below one millisecond. But as I don't really know
> what this value means (I have nothing to compare it to) I don't know
> whether it is important or not.
>
>
> Guy
> 





More information about the questions mailing list