UBNT XR-9 versus XR-5 performance not as good even w/ LOS

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

UBNT XR-9 versus XR-5 performance not as good even w/ LOS

Eric Malkowski
I posted this over on Ubiquiti's forum, but I'm confident a lot of you
subscribed here use UBNT's cards for various things.
So I'm posting this here to see what advice any of you w/ experience
might have to offer.

I have several UBNT radios including XR2, SR2, XR5, SR5, and XR9
All of my 2 and 5 ghz radios work fantastic w/ 2.6.29.X kernel and
Madwifi 0.9.4 release. I think it's because they work fine w/ default 20
mhz channels (up to 54 Mbps) etc.
I use Pac-wireless (LAIRD) OD24-9 and OD24-7 along w/ OD58-12 and their
Backfire dish ES58-17 and also Superpass panel antennas for AP and
backhaul setups.
I've also used LAIRD's mesh series OD58-12M and OD24-7M that attach
nicely to an N-female bulkhead connector right on the enclosure.

My XR-9's are marginal in the Boston area (Cambridge) and somewhat Ok
out in the country.
I have a couple cheaper hyperlink yagi antennas for testing HG909Y-NM.

I realize XR9 cannot have the channel width set when using regular
madwifi and it appears ubiquiti's radio driver is no longer developed.
So my question is: What is the proper way to use the XR9 with Madwifi
0.9.4 or the Madwifi current trunk (which includes the now opensource
HAL since madwifi-free was merged to the trunk).

My guess is since channels 5 and 6 support 5/10/20 mhz width (these are
912 Mhz and 917 Mhz center frequencies respectively) then just using
Madwifi 0.9.4 to set it up on channel 5 or 6 as AP, adhoc, station, etc.
with proper tx and rx antenna settings to the single MMCX connector
would get me standard 20mhz channel width? I'm fine with that and
avoiding channels 4 and 7 since the card can't do 20mhz operation on those.

Here's the conversion chart for the channels:

Channel 4 - 2427 = 907Mhz @ 5/10 Mhz bandwidth
Channel 5 - 2432 = 912Mhz @ 5/10/20Mhz bandwidth
Channel 6 - 2437 = 917Mhz @ 5/10/20Mhz bandwidth
Channel 7 - 2442 = 922Mhz @ 5/10 Mhz bandwidth

Next question: Would going to the madwifi trunk buy my anything in terms
of functionality? i.e. would I possibly be able to set channel bandwidths?

In my testing if I remember correctly using channel 5 or 6 out in the
country w/ a couple houses and dense pine trees (no line of sight) 1000
ft apart I would get -65 signal on each end and about 18 megabit rate.
In Boston w/ clear line of sight w/ the same setup about 2000 feet apart
I could only get about -75 on each end w/ only a ping or two. I don't
remember the reported noise floors.  I'm wondering if only doing the 18
megabit rate is an indicator that the card is only doing 10 mhz width
due to default settings madwifi 0.9.4 uses.  With -65 signal and better
on XR2 or XR5 cards, I can easily do higher rates (36 48 54 Mbps).  I
don't think there should be too much noise around since we were along
the Charles river in Cambridge just shooting across the water at
constant elevation and away from the very dense buildings in Boston
downtown proper.

I read on UBNT forums also in past posts I think Mike Ford had mentioned
pac wireless antennas were one of his favorites for XR9 -- they are one
of my favorites for XR2 and XR5 but I haven't tried anything but the
cheap hyperlink yagis. I also read someone likes Superpass (which I also
like their panels for 2.4 ghz stuff) and I could try those.  It's
unclear if it's an antenna issue, noise, or the card be driven in a way
it's not meant to be by stock madwifi 0.9.4

But before I try more antennas and go any further with XR9 (or XR7 which
I have been looking at and wondering about) I wanted to know what I
could expect and how to properly use the XR9 with Madwifi 0.9.4 latest
release (which is still binary HAL) or with Madwifi trunk. I'm hoping I
can just set one of the two 20mhz supporting channels and enjoy using it
-- but an XR5 w/ similar line of sight with similar gain antennas (and
not even directional on one but rather omni one end and backfire dish
the other end) can get signal of -50 or better on both ends w/ easy 48
or 54 Mbps rates on 2000' line w/ LOS in Cambridge along the charles
river (same beam path as the XR-9s).

If I cannot expect to reliably use madwifi 0.9.4 or madwifi trunk w/ the
XR9, then what should I use and where are directions to be able to set
channel widths w/ some type of known working driver (modified Madwifi or
UBNT radio driver) that will work on recent 2.6 linux kernels on x86
hardware. (I use PC engines Alix boards that are designed to tolerate
the high powered radios of UBNT -- specifically the D series like 3D2 or
2D3.

Any advice would be appreciated.

Sorry for the long winded message.  I like to include details.

-Eric Malkowski

------------------------------------------------------------------------------
The Planet: dedicated and managed hosting, cloud storage, colocation
Stay online with enterprise data centers and the best network in the business
Choose flexible plans and management services without long-term contracts
Personal 24x7 support from experience hosting pros just a phone call away.
http://p.sf.net/sfu/theplanet-com
_______________________________________________
Madwifi-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/madwifi-users
Reply | Threaded
Open this post in threaded view
|

Re: UBNT XR-9 versus XR-5 performance not as good evenw/ LOS

Brett Wright
Hi,

The XR9 will only ever use the 20MHz channels with madwifi, unless you
hack the source code. One problem with a 20MHz wide channel at 900MHz is
there are a lot of frequency hopping radios around that use that ISM
band also. They will surely degrade your throughput.

In the case of noise from frequency hoppers, you are better off with the
10 or 5MHz channels, because a lot more of the hopping frequencies will
be outside of your band.

Which brings me to another problem I have struggled very long and hard
with. The channel noise reading for madwifi does not represent the
actual background RF energy on the channel. There is really very little
publicly available information on this topic, so let me give a quick
example of what I mean. I set up a test where I monitored the 900MHz
band used by the XR9 using a spectrum alalyzer. I could clearly see RF
interference with a signal level of -60dBm, however an XR9 reported only
-92dBm channel noise.

I would really love to know for certain that it is just not possible for
the atheros hardware to provide a true reading of channel noise? This
really is a very major problem, as you are finding out, if you cannot
accurately measure the channel noise then how can you accurately asses
the viability of your radio path (other than just noting that the
throughput one achieves is miserable given the signal strength).

Can anyone out there help shed any more light on this issue?

Brett

> -----Original Message-----
> From: Eric Malkowski [mailto:[hidden email]]
> Sent: Tuesday, 9 February 2010 5:56 AM
> To: [hidden email]
> Subject: [Madwifi-users] UBNT XR-9 versus XR-5 performance not as good
> evenw/ LOS
>
> I posted this over on Ubiquiti's forum, but I'm confident a lot of you
> subscribed here use UBNT's cards for various things.
> So I'm posting this here to see what advice any of you w/ experience
> might have to offer.
>
> I have several UBNT radios including XR2, SR2, XR5, SR5, and XR9
> All of my 2 and 5 ghz radios work fantastic w/ 2.6.29.X kernel and
> Madwifi 0.9.4 release. I think it's because they work fine w/ default
20
> mhz channels (up to 54 Mbps) etc.
> I use Pac-wireless (LAIRD) OD24-9 and OD24-7 along w/ OD58-12 and
their

> Backfire dish ES58-17 and also Superpass panel antennas for AP and
> backhaul setups.
> I've also used LAIRD's mesh series OD58-12M and OD24-7M that attach
> nicely to an N-female bulkhead connector right on the enclosure.
>
> My XR-9's are marginal in the Boston area (Cambridge) and somewhat Ok
> out in the country.
> I have a couple cheaper hyperlink yagi antennas for testing HG909Y-NM.
>
> I realize XR9 cannot have the channel width set when using regular
> madwifi and it appears ubiquiti's radio driver is no longer developed.
> So my question is: What is the proper way to use the XR9 with Madwifi
> 0.9.4 or the Madwifi current trunk (which includes the now opensource
> HAL since madwifi-free was merged to the trunk).
>
> My guess is since channels 5 and 6 support 5/10/20 mhz width (these
are
> 912 Mhz and 917 Mhz center frequencies respectively) then just using
> Madwifi 0.9.4 to set it up on channel 5 or 6 as AP, adhoc, station,
etc.

> with proper tx and rx antenna settings to the single MMCX connector
> would get me standard 20mhz channel width? I'm fine with that and
> avoiding channels 4 and 7 since the card can't do 20mhz operation on
> those.
>
> Here's the conversion chart for the channels:
>
> Channel 4 - 2427 = 907Mhz @ 5/10 Mhz bandwidth
> Channel 5 - 2432 = 912Mhz @ 5/10/20Mhz bandwidth
> Channel 6 - 2437 = 917Mhz @ 5/10/20Mhz bandwidth
> Channel 7 - 2442 = 922Mhz @ 5/10 Mhz bandwidth
>
> Next question: Would going to the madwifi trunk buy my anything in
terms
> of functionality? i.e. would I possibly be able to set channel
bandwidths?
>
> In my testing if I remember correctly using channel 5 or 6 out in the
> country w/ a couple houses and dense pine trees (no line of sight)
1000
> ft apart I would get -65 signal on each end and about 18 megabit rate.
> In Boston w/ clear line of sight w/ the same setup about 2000 feet
apart
> I could only get about -75 on each end w/ only a ping or two. I don't
> remember the reported noise floors.  I'm wondering if only doing the
18
> megabit rate is an indicator that the card is only doing 10 mhz width
> due to default settings madwifi 0.9.4 uses.  With -65 signal and
better
> on XR2 or XR5 cards, I can easily do higher rates (36 48 54 Mbps).  I
> don't think there should be too much noise around since we were along
> the Charles river in Cambridge just shooting across the water at
> constant elevation and away from the very dense buildings in Boston
> downtown proper.
>
> I read on UBNT forums also in past posts I think Mike Ford had
mentioned
> pac wireless antennas were one of his favorites for XR9 -- they are
one
> of my favorites for XR2 and XR5 but I haven't tried anything but the
> cheap hyperlink yagis. I also read someone likes Superpass (which I
also
> like their panels for 2.4 ghz stuff) and I could try those.  It's
> unclear if it's an antenna issue, noise, or the card be driven in a
way
> it's not meant to be by stock madwifi 0.9.4
>
> But before I try more antennas and go any further with XR9 (or XR7
which
> I have been looking at and wondering about) I wanted to know what I
> could expect and how to properly use the XR9 with Madwifi 0.9.4 latest
> release (which is still binary HAL) or with Madwifi trunk. I'm hoping
I
> can just set one of the two 20mhz supporting channels and enjoy using
it
> -- but an XR5 w/ similar line of sight with similar gain antennas (and
> not even directional on one but rather omni one end and backfire dish
> the other end) can get signal of -50 or better on both ends w/ easy 48
> or 54 Mbps rates on 2000' line w/ LOS in Cambridge along the charles
> river (same beam path as the XR-9s).
>
> If I cannot expect to reliably use madwifi 0.9.4 or madwifi trunk w/
the
> XR9, then what should I use and where are directions to be able to set
> channel widths w/ some type of known working driver (modified Madwifi
or
> UBNT radio driver) that will work on recent 2.6 linux kernels on x86
> hardware. (I use PC engines Alix boards that are designed to tolerate
> the high powered radios of UBNT -- specifically the D series like 3D2
or
> 2D3.
>
> Any advice would be appreciated.
>
> Sorry for the long winded message.  I like to include details.
>
> -Eric Malkowski
>
>
------------------------------------------------------------------------
--
> ----
> The Planet: dedicated and managed hosting, cloud storage, colocation
> Stay online with enterprise data centers and the best network in the
> business
> Choose flexible plans and management services without long-term
contracts
> Personal 24x7 support from experience hosting pros just a phone call
away.
> http://p.sf.net/sfu/theplanet-com
> _______________________________________________
> Madwifi-users mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/madwifi-users
>
>
>


------------------------------------------------------------------------------
The Planet: dedicated and managed hosting, cloud storage, colocation
Stay online with enterprise data centers and the best network in the business
Choose flexible plans and management services without long-term contracts
Personal 24x7 support from experience hosting pros just a phone call away.
http://p.sf.net/sfu/theplanet-com
_______________________________________________
Madwifi-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/madwifi-users
Reply | Threaded
Open this post in threaded view
|

Re: UBNT XR-9 versus XR-5 performance not as good evenw/ LOS

Eric Malkowski
Brett-

Thanks a bunch for the reply.  I will stick with the 20mhz channels so I can
stick w/ 0.9.4 madwifi which has worked well for me (so far w/ XR5, XR2, SR2,
SR5 and CM-9 radios I use)

It sounds like my issues are mostly noise.  I don't have a spectrum analyzer,
but the UBNT airview 9 looks tempting.
I thought it may have been poor antennas or the way in which madwifi configures
the hardware that made my XR9 radios do poorly compared to XR5.

I will try again w/ a pair of superpass panels out in the country to see if i
can get w/ LOS some decent performance.  Then, possibly with the aid of an
airview 9, I can take another look at the Cambridge area.

Since the XR9 can work w/ 20 mhz channels, then this also means I can try the
XR7.  The thought is the 700 band will be much less crowded.  I didn't want to
keep working with the UBNT lower frequency cards if some fundemental problem w/
using them with madwifi would make them not work well, but that sounds like it's
not the case if one can use 20mhz bandwidth channels.

My problem is we have a few areas that are somewhat non-los and it would be nice
to use 900 or 700 to get through those areas.  We've got other areas with LOS or
just a small tree branch or two and XR5 radios do fantastic.  It would nice to
do a hybrid of 5 ghz and 900 or 700 mhz to get to where I want to be.  So far
the 900 stuff has been a no-go, but I didn't want to give up just yet.

-Eric






> Hi,
>
> The XR9 will only ever use the 20MHz channels with madwifi, unless you
> hack the source code. One problem with a 20MHz wide channel at 900MHz is
> there are a lot of frequency hopping radios around that use that ISM
> band also. They will surely degrade your throughput.
>
> In the case of noise from frequency hoppers, you are better off with the
> 10 or 5MHz channels, because a lot more of the hopping frequencies will
> be outside of your band.
>
> Which brings me to another problem I have struggled very long and hard
> with. The channel noise reading for madwifi does not represent the
> actual background RF energy on the channel. There is really very little
> publicly available information on this topic, so let me give a quick
> example of what I mean. I set up a test where I monitored the 900MHz
> band used by the XR9 using a spectrum alalyzer. I could clearly see RF
> interference with a signal level of -60dBm, however an XR9 reported only
> -92dBm channel noise.
>
> I would really love to know for certain that it is just not possible for
> the atheros hardware to provide a true reading of channel noise? This
> really is a very major problem, as you are finding out, if you cannot
> accurately measure the channel noise then how can you accurately asses
> the viability of your radio path (other than just noting that the
> throughput one achieves is miserable given the signal strength).
>
> Can anyone out there help shed any more light on this issue?
>
> Brett
>
>> -----Original Message-----
>> From: Eric Malkowski [mailto:[hidden email]]
>> Sent: Tuesday, 9 February 2010 5:56 AM
>> To: [hidden email]
>> Subject: [Madwifi-users] UBNT XR-9 versus XR-5 performance not as good
>> evenw/ LOS
>>
>> I posted this over on Ubiquiti's forum, but I'm confident a lot of you
>> subscribed here use UBNT's cards for various things.
>> So I'm posting this here to see what advice any of you w/ experience
>> might have to offer.
>>
>> I have several UBNT radios including XR2, SR2, XR5, SR5, and XR9
>> All of my 2 and 5 ghz radios work fantastic w/ 2.6.29.X kernel and
>> Madwifi 0.9.4 release. I think it's because they work fine w/ default
> 20
>> mhz channels (up to 54 Mbps) etc.
>> I use Pac-wireless (LAIRD) OD24-9 and OD24-7 along w/ OD58-12 and
> their
>> Backfire dish ES58-17 and also Superpass panel antennas for AP and
>> backhaul setups.
>> I've also used LAIRD's mesh series OD58-12M and OD24-7M that attach
>> nicely to an N-female bulkhead connector right on the enclosure.
>>
>> My XR-9's are marginal in the Boston area (Cambridge) and somewhat Ok
>> out in the country.
>> I have a couple cheaper hyperlink yagi antennas for testing HG909Y-NM.
>>
>> I realize XR9 cannot have the channel width set when using regular
>> madwifi and it appears ubiquiti's radio driver is no longer developed.
>> So my question is: What is the proper way to use the XR9 with Madwifi
>> 0.9.4 or the Madwifi current trunk (which includes the now opensource
>> HAL since madwifi-free was merged to the trunk).
>>
>> My guess is since channels 5 and 6 support 5/10/20 mhz width (these
> are
>> 912 Mhz and 917 Mhz center frequencies respectively) then just using
>> Madwifi 0.9.4 to set it up on channel 5 or 6 as AP, adhoc, station,
> etc.
>> with proper tx and rx antenna settings to the single MMCX connector
>> would get me standard 20mhz channel width? I'm fine with that and
>> avoiding channels 4 and 7 since the card can't do 20mhz operation on
>> those.
>>
>> Here's the conversion chart for the channels:
>>
>> Channel 4 - 2427 = 907Mhz @ 5/10 Mhz bandwidth
>> Channel 5 - 2432 = 912Mhz @ 5/10/20Mhz bandwidth
>> Channel 6 - 2437 = 917Mhz @ 5/10/20Mhz bandwidth
>> Channel 7 - 2442 = 922Mhz @ 5/10 Mhz bandwidth
>>
>> Next question: Would going to the madwifi trunk buy my anything in
> terms
>> of functionality? i.e. would I possibly be able to set channel
> bandwidths?
>>
>> In my testing if I remember correctly using channel 5 or 6 out in the
>> country w/ a couple houses and dense pine trees (no line of sight)
> 1000
>> ft apart I would get -65 signal on each end and about 18 megabit rate.
>> In Boston w/ clear line of sight w/ the same setup about 2000 feet
> apart
>> I could only get about -75 on each end w/ only a ping or two. I don't
>> remember the reported noise floors.  I'm wondering if only doing the
> 18
>> megabit rate is an indicator that the card is only doing 10 mhz width
>> due to default settings madwifi 0.9.4 uses.  With -65 signal and
> better
>> on XR2 or XR5 cards, I can easily do higher rates (36 48 54 Mbps).  I
>> don't think there should be too much noise around since we were along
>> the Charles river in Cambridge just shooting across the water at
>> constant elevation and away from the very dense buildings in Boston
>> downtown proper.
>>
>> I read on UBNT forums also in past posts I think Mike Ford had
> mentioned
>> pac wireless antennas were one of his favorites for XR9 -- they are
> one
>> of my favorites for XR2 and XR5 but I haven't tried anything but the
>> cheap hyperlink yagis. I also read someone likes Superpass (which I
> also
>> like their panels for 2.4 ghz stuff) and I could try those.  It's
>> unclear if it's an antenna issue, noise, or the card be driven in a
> way
>> it's not meant to be by stock madwifi 0.9.4
>>
>> But before I try more antennas and go any further with XR9 (or XR7
> which
>> I have been looking at and wondering about) I wanted to know what I
>> could expect and how to properly use the XR9 with Madwifi 0.9.4 latest
>> release (which is still binary HAL) or with Madwifi trunk. I'm hoping
> I
>> can just set one of the two 20mhz supporting channels and enjoy using
> it
>> -- but an XR5 w/ similar line of sight with similar gain antennas (and
>> not even directional on one but rather omni one end and backfire dish
>> the other end) can get signal of -50 or better on both ends w/ easy 48
>> or 54 Mbps rates on 2000' line w/ LOS in Cambridge along the charles
>> river (same beam path as the XR-9s).
>>
>> If I cannot expect to reliably use madwifi 0.9.4 or madwifi trunk w/
> the
>> XR9, then what should I use and where are directions to be able to set
>> channel widths w/ some type of known working driver (modified Madwifi
> or
>> UBNT radio driver) that will work on recent 2.6 linux kernels on x86
>> hardware. (I use PC engines Alix boards that are designed to tolerate
>> the high powered radios of UBNT -- specifically the D series like 3D2
> or
>> 2D3.
>>
>> Any advice would be appreciated.
>>
>> Sorry for the long winded message.  I like to include details.
>>
>> -Eric Malkowski
>>
>>
> ------------------------------------------------------------------------
> --
>> ----
>> The Planet: dedicated and managed hosting, cloud storage, colocation
>> Stay online with enterprise data centers and the best network in the
>> business
>> Choose flexible plans and management services without long-term
> contracts
>> Personal 24x7 support from experience hosting pros just a phone call
> away.
>> http://p.sf.net/sfu/theplanet-com
>> _______________________________________________
>> Madwifi-users mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/madwifi-users
>>
>>
>>
>


------------------------------------------------------------------------------
The Planet: dedicated and managed hosting, cloud storage, colocation
Stay online with enterprise data centers and the best network in the business
Choose flexible plans and management services without long-term contracts
Personal 24x7 support from experience hosting pros just a phone call away.
http://p.sf.net/sfu/theplanet-com
_______________________________________________
Madwifi-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/madwifi-users