"Philip Herlihy" <bounceb***@you.com> wrote in message
news:OispEjF3JHA.4412@TK2MSFTNGP06.phx.gbl...
>I look after IT issues, including networking, for a couple of small
>offices.  I sometimes get a situation where, despite apparently good signal
>strength, a machine won't connect.
>
> One machine won't connect (most of the time!) despite my trying three
> different makes of PCI wireless cards plus one USB adapter, and two
> different access points.  Now there's a cable draped across the office...
> Connectivity on other machines is poor while others nearby seem fine.
>
> I've experimented over months with adjustments to various settings,
> including setting static IP addresses.  One machine burst into life when I
> switched from the Intel wireless client to the Windows one, but I couldn't
> duplicate this on another machine.
>
> My mobile runs Windows Mobile, and this allows me to run the excellent
> "sniffi", which can graph the signals on various channels, and although
> Netstumbler can find dozens of different distinct services I've set ours
> to the channels least-used locally.
>
> I've recently tried experimenting with TCP parameters like MTU and RWIN,
> using DrTCP.exe.  However, I'm working in the dark, and I'd like to be
> able to monitor the effect of what I'm doing.  I've now started using
> Wireshark to monitor at the packet level, but I don't know what I'm
> looking for.  Does anyone have some experience of this sort of thing they
> could share?
>
> Phil, London

> Hi
> To get solid Wireless in a business you might need a network of few
> Access Points, or a WDS arrangement.
> Jack (MS, MVP-Networking)
>
> "Philip Herlihy" <bounceb***@you.com> wrote in message
> news:OispEjF3JHA.4412@TK2MSFTNGP06.phx.gbl...
>> I look after IT issues, including networking, for a couple of small
>> offices.  I sometimes get a situation where, despite apparently good
>> signal strength, a machine won't connect.
>>
>> One machine won't connect (most of the time!) despite my trying three
>> different makes of PCI wireless cards plus one USB adapter, and two
>> different access points.  Now there's a cable draped across the
>> office... Connectivity on other machines is poor while others nearby
>> seem fine.
>>
>> I've experimented over months with adjustments to various settings,
>> including setting static IP addresses.  One machine burst into life
>> when I switched from the Intel wireless client to the Windows one, but
>> I couldn't duplicate this on another machine.
>>
>> My mobile runs Windows Mobile, and this allows me to run the excellent
>> "sniffi", which can graph the signals on various channels, and
>> although Netstumbler can find dozens of different distinct services
>> I've set ours to the channels least-used locally.
>>
>> I've recently tried experimenting with TCP parameters like MTU and
>> RWIN, using DrTCP.exe.  However, I'm working in the dark, and I'd like
>> to be able to monitor the effect of what I'm doing.  I've now started
>> using Wireshark to monitor at the packet level, but I don't know what
>> I'm looking for.  Does anyone have some experience of this sort of
>> thing they could share?
>>
>> Phil, London
>

"Philip Herlihy" <bounceb***@you.com> wrote in message
news:ekDW0rS3JHA.5276@TK2MSFTNGP04.phx.gbl...
> Jack [MVP-Networking] wrote:
>> Hi
>> To get solid Wireless in a business you might need a network of few
>> Access Points, or a WDS arrangement.
>> Jack (MS, MVP-Networking)
>>
>> "Philip Herlihy" <bounceb***@you.com> wrote in message
>> news:OispEjF3JHA.4412@TK2MSFTNGP06.phx.gbl...
>>> I look after IT issues, including networking, for a couple of small
>>> offices.  I sometimes get a situation where, despite apparently good
>>> signal strength, a machine won't connect.
>>>
>>> One machine won't connect (most of the time!) despite my trying three
>>> different makes of PCI wireless cards plus one USB adapter, and two
>>> different access points.  Now there's a cable draped across the
>>> office... Connectivity on other machines is poor while others nearby
>>> seem fine.
>>>
>>> I've experimented over months with adjustments to various settings,
>>> including setting static IP addresses.  One machine burst into life when
>>> I switched from the Intel wireless client to the Windows one, but I
>>> couldn't duplicate this on another machine.
>>>
>>> My mobile runs Windows Mobile, and this allows me to run the excellent
>>> "sniffi", which can graph the signals on various channels, and although
>>> Netstumbler can find dozens of different distinct services I've set ours
>>> to the channels least-used locally.
>>>
>>> I've recently tried experimenting with TCP parameters like MTU and RWIN,
>>> using DrTCP.exe.  However, I'm working in the dark, and I'd like to be
>>> able to monitor the effect of what I'm doing.  I've now started using
>>> Wireshark to monitor at the packet level, but I don't know what I'm
>>> looking for.  Does anyone have some experience of this sort of thing
>>> they could share?
>>>
>>> Phil, London
>>
>
> Thanks, Jack.  We do have two access points, and both are accessible from
> all machines, although some can only just "see" the furthest one. All
> machines have at least one of them delivering good signal strength,
> usually excellent.  Still we get these odd connection problems.  I've
> tried many things over the last few months, and would like to be able to
> measure or monitor what's happening - hence the experiments with
> Wireshark.  Are there any particular situations I should look out for?
>
> Phil

> Hi
> Set the Access Points to two different channels.
> Try temp. to put the "Lame" computers within few feet from on of the
> Access Point to figure out if it is Wireless problem, or computer setting.
> When you use WZC cleanup the preferred list and leave in only the Access
> Point that is best for the specific computer.
> http://www.ezlan.net/wireless/wzc3.jpg
> Note: I can keep going with more pointers like this (aka Trial & Error),
> but as Pavel said in his post "live it too short" ;).
> Jack (MS, MVP-Networking)
>
> "Philip Herlihy" <bounceb***@you.com> wrote in message
> news:ekDW0rS3JHA.5276@TK2MSFTNGP04.phx.gbl...
>> Jack [MVP-Networking] wrote:
>>> Hi
>>> To get solid Wireless in a business you might need a network of few
>>> Access Points, or a WDS arrangement.
>>> Jack (MS, MVP-Networking)
>>>
>>> "Philip Herlihy" <bounceb***@you.com> wrote in message
>>> news:OispEjF3JHA.4412@TK2MSFTNGP06.phx.gbl...
>>>> I look after IT issues, including networking, for a couple of small
>>>> offices.  I sometimes get a situation where, despite apparently good
>>>> signal strength, a machine won't connect.
>>>>
>>>> One machine won't connect (most of the time!) despite my trying
>>>> three different makes of PCI wireless cards plus one USB adapter,
>>>> and two different access points.  Now there's a cable draped across
>>>> the office... Connectivity on other machines is poor while others
>>>> nearby seem fine.
>>>>
>>>> I've experimented over months with adjustments to various settings,
>>>> including setting static IP addresses.  One machine burst into life
>>>> when I switched from the Intel wireless client to the Windows one,
>>>> but I couldn't duplicate this on another machine.
>>>>
>>>> My mobile runs Windows Mobile, and this allows me to run the
>>>> excellent "sniffi", which can graph the signals on various channels,
>>>> and although Netstumbler can find dozens of different distinct
>>>> services I've set ours to the channels least-used locally.
>>>>
>>>> I've recently tried experimenting with TCP parameters like MTU and
>>>> RWIN, using DrTCP.exe.  However, I'm working in the dark, and I'd
>>>> like to be able to monitor the effect of what I'm doing.  I've now
>>>> started using Wireshark to monitor at the packet level, but I don't
>>>> know what I'm looking for.  Does anyone have some experience of this
>>>> sort of thing they could share?
>>>>
>>>> Phil, London
>>>
>>
>> Thanks, Jack.  We do have two access points, and both are accessible
>> from all machines, although some can only just "see" the furthest one.
>> All machines have at least one of them delivering good signal
>> strength, usually excellent.  Still we get these odd connection
>> problems.  I've tried many things over the last few months, and would
>> like to be able to measure or monitor what's happening - hence the
>> experiments with Wireshark.  Are there any particular situations I
>> should look out for?
>>
>> Phil
>

"Philip Herlihy" <bounceb***@you.com> wrote in message
news:O2mhoAW3JHA.1512@TK2MSFTNGP05.phx.gbl...
> Jack [MVP-Networking] wrote:
>> Hi
>> Set the Access Points to two different channels.
>> Try temp. to put the "Lame" computers within few feet from on of the
>> Access Point to figure out if it is Wireless problem, or computer
>> setting.
>> When you use WZC cleanup the preferred list and leave in only the Access
>> Point that is best for the specific computer.
>> http://www.ezlan.net/wireless/wzc3.jpg
>> Note: I can keep going with more pointers like this (aka Trial & Error),
>> but as Pavel said in his post "live it too short" ;).
>> Jack (MS, MVP-Networking)
>>
>> "Philip Herlihy" <bounceb***@you.com> wrote in message
>> news:ekDW0rS3JHA.5276@TK2MSFTNGP04.phx.gbl...
>>> Jack [MVP-Networking] wrote:
>>>> Hi
>>>> To get solid Wireless in a business you might need a network of few
>>>> Access Points, or a WDS arrangement.
>>>> Jack (MS, MVP-Networking)
>>>>
>>>> "Philip Herlihy" <bounceb***@you.com> wrote in message
>>>> news:OispEjF3JHA.4412@TK2MSFTNGP06.phx.gbl...
>>>>> I look after IT issues, including networking, for a couple of small
>>>>> offices.  I sometimes get a situation where, despite apparently good
>>>>> signal strength, a machine won't connect.
>>>>>
>>>>> One machine won't connect (most of the time!) despite my trying three
>>>>> different makes of PCI wireless cards plus one USB adapter, and two
>>>>> different access points.  Now there's a cable draped across the
>>>>> office... Connectivity on other machines is poor while others nearby
>>>>> seem fine.
>>>>>
>>>>> I've experimented over months with adjustments to various settings,
>>>>> including setting static IP addresses.  One machine burst into life
>>>>> when I switched from the Intel wireless client to the Windows one, but
>>>>> I couldn't duplicate this on another machine.
>>>>>
>>>>> My mobile runs Windows Mobile, and this allows me to run the excellent
>>>>> "sniffi", which can graph the signals on various channels, and
>>>>> although Netstumbler can find dozens of different distinct services
>>>>> I've set ours to the channels least-used locally.
>>>>>
>>>>> I've recently tried experimenting with TCP parameters like MTU and
>>>>> RWIN, using DrTCP.exe.  However, I'm working in the dark, and I'd like
>>>>> to be able to monitor the effect of what I'm doing.  I've now started
>>>>> using Wireshark to monitor at the packet level, but I don't know what
>>>>> I'm looking for.  Does anyone have some experience of this sort of
>>>>> thing they could share?
>>>>>
>>>>> Phil, London
>>>>
>>>
>>> Thanks, Jack.  We do have two access points, and both are accessible
>>> from all machines, although some can only just "see" the furthest one.
>>> All machines have at least one of them delivering good signal strength,
>>> usually excellent.  Still we get these odd connection problems.  I've
>>> tried many things over the last few months, and would like to be able to
>>> measure or monitor what's happening - hence the experiments with
>>> Wireshark.  Are there any particular situations I should look out for?
>>>
>>> Phil
>>
>
>
> Well, I found Pavel's comment surprising, as it boils down to a suggestion
> not to bother trying to understand how things work at a deep level but to
> surrender responsibility to someone else.  Other groups in which I
> participate have a different outlook, and the most naive and untutored
> questions get patient, tailored answers (more like Jack's) intended to
> help the poster up the next rung, wherever on the ladder that might be.
>
> I am a "consultant", of sorts.  Coming from a niche near-real-time
> programming background I'm now making a serious (and rather exhausting)
> attempt to be able to cover all the basics, from networking to graphic
> design, from database programming to hardware repair.  Naturally, I don't
> know everything, but I've been in the industry for 28 years so far, so
> although I've no hope of keeping up with the outer envelope, I'm doing
> fairly well at keeping on top of what most customers need.
>
> In a wireless setup I've often seen a situation where the signal strength
> appears to be good or excellent, and yet the "Limited or no connectivity"
> message suggests DHPC is failing somehow.  I've checked the issues Jack
> suggests above (thanks Jack) and still sometimes see problems.
>
> Where I am now is that I'm trying to dig deeper into the details for real
> evidence of problems.  One correspondent (elsewhere) suggested that MTU
> might be an issue, so I've been experimenting with that, but the key to
> empiricism is analytic measurement, so I've been playing around with the
> Wireshark protocol analyser hoping to see patterns in the packet stream.
> I once successfully diagnosed a (wired) connectivity problem by setting
> very detailed firewall logging, and picked up the fact that ICMP 3.4
> packets were being blocked and so the sending station was unable to detect
> that it needed to reduce its packet size - that's the sort of thing I'm
> looking for in these odd wireless situations.  Rather than try and judge
> at a macro level (eg. download times) I'm looking for something
> finer-grained, giving more certainty.
>
> Maybe this just isn't a useful approach, or maybe it's one that has been
> largely overlooked.  But when you have one machine (happy with a cable)
> that can't make a wireless connection with any of three access points
> using any of four wireless adapters (one at a time) despite very thorough
> building, repairing and rebuilding of the windows network stacks,
> something's there to be found.
>
> Phil

>>>>> "Philip Herlihy" <bounceb***@you.com> wrote in message
>>>>> news:OispEjF3JHA.4412@TK2MSFTNGP06.phx.gbl...
>>>>>> I look after IT issues, including networking, for a couple of
>>>>>> small offices.  I sometimes get a situation where, despite
>>>>>> apparently good signal strength, a machine won't connect.
>>>>>>
>>>>>> One machine won't connect (most of the time!) despite my trying
>>>>>> three different makes of PCI wireless cards plus one USB adapter,
>>>>>> and two different access points.  Now there's a cable draped
>>>>>> across the office... Connectivity on other machines is poor while
>>>>>> others nearby seem fine.
>>>>>>
>>>>>> I've experimented over months with adjustments to various
>>>>>> settings, including setting static IP addresses.  One machine
>>>>>> burst into life when I switched from the Intel wireless client to
>>>>>> the Windows one, but I couldn't duplicate this on another machine.
>>>>>>
>>>>>> My mobile runs Windows Mobile, and this allows me to run the
>>>>>> excellent "sniffi", which can graph the signals on various
>>>>>> channels, and although Netstumbler can find dozens of different
>>>>>> distinct services I've set ours to the channels least-used locally.
>>>>>>
>>>>>> I've recently tried experimenting with TCP parameters like MTU and
>>>>>> RWIN, using DrTCP.exe.  However, I'm working in the dark, and I'd
>>>>>> like to be able to monitor the effect of what I'm doing.  I've now
>>>>>> started using Wireshark to monitor at the packet level, but I
>>>>>> don't know what I'm looking for.  Does anyone have some experience
>>>>>> of this sort of thing they could share?
>>>>>>
>>>>>> Phil, London
>>>>>
>>>>
>>>> Thanks, Jack.  We do have two access points, and both are accessible
>>>> from all machines, although some can only just "see" the furthest
>>>> one. All machines have at least one of them delivering good signal
>>>> strength, usually excellent.  Still we get these odd connection
>>>> problems.  I've tried many things over the last few months, and
>>>> would like to be able to measure or monitor what's happening - hence
>>>> the experiments with Wireshark.  Are there any particular situations
>>>> I should look out for?
>>>>
>>>> Phil
>>>
>>
....

> @ Philip Herlihy:
>
> Then apologies, you seem to have much more time on your hands than
> typical busy issue-oriented IT workers.
>
> But what if the problem is in RF interference?
> This can require serious tools that ordinary IT or even electric
> engineers don't own, or software that one can't download from internet
> freely, and specific skills with these tools and software.
>
> OTOH I enjoyed working with pros (when they tolerate me looking over
> their shoulder), have learned something new from them every time.
>
> Best regards,
> -- Pavel

> Ok, then, since you already have tried the advice of Jack
> (another wi-fi adapter), and if this is not MTU, and not other software
> problem on that PC - it can be RF interference.
> TCP/IP is generally resistant to interference/signal loss, but
> lower level protocols 802.1x and EAP are fragile (do less or no retries).
>
> 1. Put another (working) PC in the place of the non working PC. If it
> will work, the problem is somewhere in software.
> 2. Otherwise check for RF conditions (use different channel, move
> antennas, move the router closer ).
> 3. The last step is ... packet sniffers. Some are freely available. The
> 802.11 spec is available as well.
>
> Good luck.
> -- pa

> Philip,
>
> I use the Omnipeek sniffer, don't have experience with Wireshark for wi-fi.
> There should be some step by step guides on Wireshark, may be Mr. Jack
> has some links on his site.
> You've mentioned the Intel wireless client.
> It can be helpful to know the exact models. Intel adapters can be made
> for specific countries and computer brands, to comply to local RF
> regulations; use of some frequencies may be blocked in hardware or
> firmware. Also, some their old models are B-only and won't work with G
> routers, or don't support some security modes.
>
> Basically you need to sniff on the air level protocol (802.11)
> and it differs from what we do usually on wired ethernet.
> First, select the channel or frequency same as of your router/AP.
> Next steps depend on what exactly "does not work".
> At least, the PC must send probe requests to the AP and receive probe
> responces.
>
> Also, as alternative to wi-fi you may want to look at powerline network
> option. AFAIK powerline is popular in UK for non-portable PCs. No new
> wires and very low maintenance.
>
> Regards,
> -- pa

> Pavel A. wrote:
>> Philip,
>>
>> I use the Omnipeek sniffer, don't have experience with Wireshark for
>> wi-fi.
>> There should be some step by step guides on Wireshark, may be Mr. Jack
>> has some links on his site.
>> You've mentioned the Intel wireless client.
>> It can be helpful to know the exact models. Intel adapters can be made
>> for specific countries and computer brands, to comply to local RF
>> regulations; use of some frequencies may be blocked in hardware or
>> firmware. Also, some their old models are B-only and won't work with G
>> routers, or don't support some security modes.
>>
>> Basically you need to sniff on the air level protocol (802.11)
>> and it differs from what we do usually on wired ethernet.
>> First, select the channel or frequency same as of your router/AP.
>> Next steps depend on what exactly "does not work".
>> At least, the PC must send probe requests to the AP and receive probe
>> responces.
>>
>> Also, as alternative to wi-fi you may want to look at powerline
>> network option. AFAIK powerline is popular in UK for non-portable PCs.
>> No new wires and very low maintenance.
>>
>> Regards,
>> -- pa
>
> Thanks, Pavel - I'll look into Omnipeek.  I do use Powerline where
> possible - Netgear make some nice ones with a built-in 4-port switch at
> each node.
>
> Phil

"Philip Herlihy" <bounceb***@you.com> wrote in message
news:O8BRTtt3JHA.6004@TK2MSFTNGP02.phx.gbl...
> Philip Herlihy wrote:
>> Pavel A. wrote:
>>> Philip,
>>>
>>> I use the Omnipeek sniffer, don't have experience with Wireshark for
>>> wi-fi.
>>> There should be some step by step guides on Wireshark, may be Mr. Jack
>>> has some links on his site.
>>> You've mentioned the Intel wireless client.
>>> It can be helpful to know the exact models. Intel adapters can be made
>>> for specific countries and computer brands, to comply to local RF
>>> regulations; use of some frequencies may be blocked in hardware or
>>> firmware. Also, some their old models are B-only and won't work with G
>>> routers, or don't support some security modes.
>>>
>>> Basically you need to sniff on the air level protocol (802.11)
>>> and it differs from what we do usually on wired ethernet.
>>> First, select the channel or frequency same as of your router/AP.
>>> Next steps depend on what exactly "does not work".
>>> At least, the PC must send probe requests to the AP and receive probe
>>> responces.
>>>
>>> Also, as alternative to wi-fi you may want to look at powerline network
>>> option. AFAIK powerline is popular in UK for non-portable PCs. No new
>>> wires and very low maintenance.
>>>
>>> Regards,
>>> -- pa
>>
>> Thanks, Pavel - I'll look into Omnipeek.  I do use Powerline where
>> possible - Netgear make some nice ones with a built-in 4-port switch at
>> each node.
>>
>> Phil
>
> I did look into OmniPeek, and I'm now a bit more in tune with your earlier
> comment to the effect that life's too short...
>
> Found a download location for OmniPeek Personal 4.1, which doesn't seem to
> be widely available now (free edition seems to have been launched in 2006
> and since withdrawn).  Have asked for a price for the "Basic" edition
> (irritating when you have to ask), but noted the Enterprise one is about
> $6K.  The Personal edition does have many bells and whistles that
> Wireshark doesn't, and appears to be able to dig deep into the wireless
> transmission layer, but only if you have a supported wireless card.  The
> website is little help in figuring out which ones those might be, apart
> from a short list.
>
> I think I probably draw the line at trying to follow conversations at the
> wireless transmission layer, unless I can find a utility demanding rather
> less study than this one - spent about two hours "getting started" and
> that's it for this week at least!
>
> I have seen (and largely ignored) low-level parameters such as beacon
> interval and fragmentation threshold which some client software exposes. I
> wonder if there is any mileage in tampering with those?  My guess is that
> it's no accident they are usually hidden.
>
> Phil

"Philip Herlihy" <bounceb***@you.com> wrote in message
news:%23Q9Cr543JHA.4272@TK2MSFTNGP06.phx.gbl...
> Pavel A. wrote:
>> Jack-MVP wrote:
>>> Live is short, and Tough ;)
>>
>> or rather -  life is tough, but short :(
>>
>> -- pa
>
>
> Ok.  I'll go and walk the dog in the Forest instead.
>
> Thanks, folks.  (Just don't like being beaten!)
>
> OmniPeek Basic is just under £1000 (+ Vat, presumably), by the way.
> Wireshark is free.
>
> Phil

> Hi
> I am using Wireshark for years (use to be called Ethereal) and it is one
> of the best free tools.
> A Jaguar cost in the USA $80.000, walking is Free.
> I can easily walk 40 (or more) blocks in New York (20 streets blocks is
> a Mile), but I am not going to walk from New York to LA. ;)
> Jack (MS, MVP-Networking)
>
> "Philip Herlihy" <bounceb***@you.com> wrote in message
> news:%23Q9Cr543JHA.4272@TK2MSFTNGP06.phx.gbl...
>> Pavel A. wrote:
>>> Jack-MVP wrote:
>>>> Live is short, and Tough ;)
>>>
>>> or rather -  life is tough, but short :(
>>>
>>> -- pa
>>
>>
>> Ok.  I'll go and walk the dog in the Forest instead.
>>
>> Thanks, folks.  (Just don't like being beaten!)
>>
>> OmniPeek Basic is just under £1000 (+ Vat, presumably), by the way.
>> Wireshark is free.
>>
>> Phil
>

> Jack [MVP-Networking] wrote:
>> Hi
>> I am using Wireshark for years (use to be called Ethereal) and it is
>> one of the best free tools.
>> A Jaguar cost in the USA $80.000, walking is Free.
>> I can easily walk 40 (or more) blocks in New York (20 streets blocks
>> is a Mile), but I am not going to walk from New York to LA. ;)
>> Jack (MS, MVP-Networking)
>>
>> "Philip Herlihy" <bounceb***@you.com> wrote in message
>> news:%23Q9Cr543JHA.4272@TK2MSFTNGP06.phx.gbl...
>>> Pavel A. wrote:
>>>> Jack-MVP wrote:
>>>>> Live is short, and Tough ;)
>>>>
>>>> or rather -  life is tough, but short :(
>>>>
>>>> -- pa
>>>
>>>
>>> Ok.  I'll go and walk the dog in the Forest instead.
>>>
>>> Thanks, folks.  (Just don't like being beaten!)
>>>
>>> OmniPeek Basic is just under £1000 (+ Vat, presumably), by the way.
>>> Wireshark is free.
>>>
>>> Phil
>>
>
> Not beaten yet...
>
> Phil

> I look after IT issues, including networking, for a couple of small
> offices.  I sometimes get a situation where, despite apparently good
> signal strength, a machine won't connect.
>
> One machine won't connect (most of the time!) despite my trying three
> different makes of PCI wireless cards plus one USB adapter, and two
> different access points.  Now there's a cable draped across the
> office...  Connectivity on other machines is poor while others nearby
> seem fine.
>
> I've experimented over months with adjustments to various settings,
> including setting static IP addresses.  One machine burst into life when
> I switched from the Intel wireless client to the Windows one, but I
> couldn't duplicate this on another machine.
>
> My mobile runs Windows Mobile, and this allows me to run the excellent
> "sniffi", which can graph the signals on various channels, and although
> Netstumbler can find dozens of different distinct services I've set ours
> to the channels least-used locally.
>
> I've recently tried experimenting with TCP parameters like MTU and RWIN,
> using DrTCP.exe.  However, I'm working in the dark, and I'd like to be
> able to monitor the effect of what I'm doing.  I've now started using
> Wireshark to monitor at the packet level, but I don't know what I'm
> looking for.  Does anyone have some experience of this sort of thing
> they could share?
>
> Phil, London