Multipath - Common on Long Distance Links? - 802.11a vs. b

Is multipath a common problem on long-distance 802.11b
point-to-point links?

Would going to 802.11a solve a multipath problem in b ?
(assuming that antennas of sufficient gain are chosen)

Are there any new worms in the 802.11a can compared with b
(besides the extra 6.7 dB of path loss though 32%
less required Fresnel zone radius, for the same distance)?

On 1 Jul 2004 09:42:53 -0700, (E-Mail Removed) (c hore) wrote:

>Is multipath a common problem on long-distance 802.11b
>point-to-point links?

Yep. How long is your distance? At 8-10 miles, you hit timing
problems with 802.11b.

>Would going to 802.11a solve a multipath problem in b ?
>(assuming that antennas of sufficient gain are chosen)

Maybe. 802.11a OFDM modulation is more resitant to multipath issues
than 802.11b. However, 802.11g uses exactly the same OFDM modulation
as 802.11a and offers slightly better range, cheaper hardware, etc.

>Are there any new worms in the 802.11a can compared with b
>(besides the extra 6.7 dB of path loss though 32%
>less required Fresnel zone radius, for the same distance)?

The extra path loss of 5.6GHz vs 2.4 GHz is offset largely by the
reduced effects of folliage attenuation and building material
penetration. Antenna gains for identical physical antennas also
equalize the comparison. My experience is that multipath fades are
often minor compared to simple interference by nearby 802.11b radios.
Hard to tell even with a spectrum analyzer site survey.


--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831.336.2558 voice http://www.LearnByDestroying.com
# (E-Mail Removed)
# 831.421.6491 digital_pager (E-Mail Removed) AE6KS

Jeff Liebermann <(E-Mail Removed)> wrote:
> On 1 Jul 2004 09:42:53 -0700, (E-Mail Removed) (c hore) wrote:
> >Is multipath a common problem on long-distance 802.11b
> >point-to-point links?
>
> Yep. How long is your distance? At 8-10 miles, you hit timing
> problems with 802.11b.

Most recently trying to do a 16 km (10 mile) link with LOS,
24 dBi antennas, and 15 dBm tower-mounted bridges, but it
is not working. We still have to haul up a spectrum
analyzer to see if the problem is interference.

I understand that, unlike a spectrum analyzer, the
BVS Yellowjacket-b can measure multipath.
Anyone know how much this tool costs?

I was thinking of putting in a Teletronics bidirectional
1 W SmartAmp on one side, but if it is a multipath problem
rather than interference, I suppose amplification won't help,
will it.

Abandoning the 2.4 GHz space for 5 GHz, and going to Airaya
802.11a outdoor bridges instead of cramming cheap 802.11b
bridges and assorted paraphernalia into Pelican cases
seems to be more and more appealing...

"c hore" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed) om...
> Jeff Liebermann <(E-Mail Removed)> wrote:
> > On 1 Jul 2004 09:42:53 -0700, (E-Mail Removed) (c hore) wrote:
> > >Is multipath a common problem on long-distance 802.11b
> > >point-to-point links?
> >
> > Yep. How long is your distance? At 8-10 miles, you hit timing
> > problems with 802.11b.
>
> Most recently trying to do a 16 km (10 mile) link with LOS,
> 24 dBi antennas, and 15 dBm tower-mounted bridges, but it
> is not working. We still have to haul up a spectrum
> analyzer to see if the problem is interference.
>
> I understand that, unlike a spectrum analyzer, the
> BVS Yellowjacket-b can measure multipath.
> Anyone know how much this tool costs?
>
> I was thinking of putting in a Teletronics bidirectional
> 1 W SmartAmp on one side, but if it is a multipath problem
> rather than interference, I suppose amplification won't help,
> will it.
>
> Abandoning the 2.4 GHz space for 5 GHz, and going to Airaya
> 802.11a outdoor bridges instead of cramming cheap 802.11b
> bridges and assorted paraphernalia into Pelican cases
> seems to be more and more appealing...

BVS Yellowjacket-b ... errmm that looks interesting thanks for that

AndiC

Jeff Liebermann <(E-Mail Removed)> wrote in message news:<(E-Mail Removed)>. ..
> The extra path loss of 5.6GHz vs 2.4 GHz is offset largely by the
> reduced effects of folliage attenuation and building material
> penetration. Antenna gains for identical physical antennas also
> equalize the comparison. My experience is that multipath fades are
> often minor compared to simple interference by nearby 802.11b radios.
> Hard to tell even with a spectrum analyzer site survey.

Hi Jeff;

This bit is ambiguous. Clearly you do not intend to say that
absorbtion at 5.6 is less than at 2.4, but it comes across that way.
For the record then, I am certain, having learned much from Jeff's
posts over the years, he did not intend to suggest that 5.6 is
absorbed less by vegetation or construction materials.

Absorbtion at 5.6 GHz in vegetation is about twice as great as it is
at 2.4 GHz or approximately 2 dB per meter of vegetation. This is
consistent with the fact that absorbtion at 900 MHz is roughly 1/3 as
much as it is at 2.4 GHz, or about 10 dB per 100 meters.

Reflections ( multipath ) at 5.6 are about 8 times as frequent as they
are at 2.4. This is consistent with the "Erskine's Multipath
Principle" which follows:

"For any given link the number of re-radiators which can cause
multipath propagation is inversly proportional to cube of the wave
length."

By this I mean, for any given path there will be a number of objects
which can cause reflections. Those objects will be more or less
efficient re-radiators depending upon their resonant frequencies and
orientation to the signal path. As frequency increases, wave length
decreases in a linear fashion. Therefore any object which is near
resonance at frequency X will be FOUR TIMES as likely to be near
resonance at frequency 2X. This is an immutable physical fact.

The effect of these reflections on link reliability is a function of
DATA RATE and not modulation scheme. Higher data rates are ALWAYS
more suceptible to "inter-symbol smashing". This occurs when the
direct and reflected components of the signal arrive with roughly the
same signal strength and are out of phase by approximately 1/2 the
length of a "symbol". For our puroses, think of a symbol as a binary
digit. It takes a finite time to signal a binary digit across the
link. If the two copies of the signal (the reflected bit and the
directly received bit) are 1/2 the length of a bit apart, it becomes
impossible for the receiving equipment to distinguish the current bit
from the next bit in the stream and you get a very high bit error
rate. The faster you signal your bits, the shorter time there is
between bits. This implies another "LAW": "The number of points
along a signaling path where a reradiator can be placed which also
causes inter-symbol smashing is proportional to the square of the
signaling rate." Put differently, double the signaling rate and you
are four times as likely to see inter-symbol smashing due to a
multipath condition.

All that said, multipath is not always your enemy. Just as you are
four times as likely to see inter-symbol smashing, you are also four
times as likely to see two reflections reinforce by arriving in phase
and polarization. It follows from the information in the paragraph
above that: "In any given volume the number of places where multipath
reinforces the desired signal is proportional to the square of the
signaling rate (time diversity effect)." and it follows from
"Erskine's Multipath Principle" this corollary that: "In any given
path the probability of multipath reflections arriving in phase and
thus reinforcing the signal ((spatial) diversity effect) is directly
proportional to the factorial of the number of reradiators in the
path."

All of this can be summed up as follows: "All other things being
equal, in an environment with reflectors, the higher the frequency and
the lower data signaling rate the greater the probability that all the
reflections will combine at any given point to yeild a useable
signal." This is absolute physical truth.

On 2 Jul 2004 00:06:47 -0700, (E-Mail Removed) (c hore) wrote:

>Jeff Liebermann <(E-Mail Removed)> wrote:
>> On 1 Jul 2004 09:42:53 -0700, (E-Mail Removed) (c hore) wrote:
>> >Is multipath a common problem on long-distance 802.11b
>> >point-to-point links?
>>
>> Yep. How long is your distance? At 8-10 miles, you hit timing
>> problems with 802.11b.

>Most recently trying to do a 16 km (10 mile) link with LOS,
>24 dBi antennas, and 15 dBm tower-mounted bridges, but it
>is not working. We still have to haul up a spectrum
>analyzer to see if the problem is interference.

Unless you have some serious experience with a spectrum analyzer,
you're gonna have some problems interpreting what you see. The first
problem is that since you're looking at the signal BEFORE
de-spreading, you have an automatic sensitivity loss equal to the
processing gain (about -10dB). This makes signals difficult to see
and well into the noise. You also run the risk of SA front end
overload from nearby transmitters. I can't see anything useful near a
TV station thanks to the garbage generated in the front end of the
spectrum analyzer by the overpowering TV signal. The answer is a
really good (i.e. cavity) 2.4Ghz band pass filter in front of the
spectrum analyzer and a high 3rd order intermod 2.4GHz RF amplifier.

What you'll probably see (after you deal with the aformentioned
issues) is garbage that doesn't look like the typical sinx/x lumpy
802.11 direct sequence spread spectrum. Instead, you'll see crud that
covers the entire band or exactly half the band. Those are Proxim
Tsunami and Lynx wireless point to point bridges commonly used by
railroads, utilities, telcos, WISP backhauls, and such. They are on
the air even if they're not moving data. The good news is that they
tend to use very directional antennas limiting their area of damage.
The bad news is that if you're on a mountain top, those antennas are
probably pointed directly at you.

The problem has been sufficiently bad that voluntary coordination
groups are emerging:
http://www.wbanc.com

Also, at 10 miles, you WILL run into timing problems with your common
802.11b radios. They're not made of that kind of distance and some of
the timeouts will need to be tweaked. Your fade margin is allegedly
about 18dB:
http://www.ydi.com/calculation/som.php
(assuming -3dB loss in connectors and pigtails). My experience is
that few of the radios actually deliver +15dBm and the typical
barbeque grill 24dBi antenna is more like 20dBi gain. Incidentally,
at 10miles, you WILL have antenna alignment problems. I found it best
to start inline, about 1-2 miles away, and align each hilltop antenna
individually with a strong signal. That eliminates boresight errors
and lots of swearing and cursing. Anyway, 18dB is tolerably
sufficient if you can get it and keep it once the rain hits. I don't
think you can with commodity hardware.

>I understand that, unlike a spectrum analyzer, the
>BVS Yellowjacket-b can measure multipath.
>Anyone know how much this tool costs?

I think (i.e. not sure) that it's about $1500.
http://www.bvsystems.com/Products/WL...YJ802.11bg.htm
It's really a protocol analyzer, not a spectrum analyzer.

http://www.mail-archive.com/isp-wire.../msg30641.html

If you wanna go cheap on the spectrum analyzer, the Proxim Rangelan2
7400 card and the utility software make a dandy (low sensitivity)
spectrum analyzer. You can't easily tell what you're looking at, but
it's good for gross interference and direction finding. I haven't
tried it and don't know how well it will work on a mountain top.

http://www.e3.com.au/modules.php?op=...article&sid=22

>I was thinking of putting in a Teletronics bidirectional
>1 W SmartAmp on one side, but if it is a multipath problem
>rather than interference, I suppose amplification won't help,
>will it.

1 watt (+30dBm) and a 24dBi antenna are over power and illegal. Your
maximum legal for an omni is +36dBm (1 watt plus a 6dBi omni). For
point to point, the formula gets complex in that you're allowed 2dB of
gain for every 1dB drop in tx power.

>Abandoning the 2.4 GHz space for 5 GHz, and going to Airaya
>802.11a outdoor bridges instead of cramming cheap 802.11b
>bridges and assorted paraphernalia into Pelican cases
>seems to be more and more appealing...

Good, fast, cheap. Pick two.


--
Jeff Liebermann (E-Mail Removed)
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 AE6KS 831-336-2558

On 2 Jul 2004 09:21:10 -0700, (E-Mail Removed) (Michael Erskine)
wrote:

>Jeff Liebermann <(E-Mail Removed)> wrote in message news:<(E-Mail Removed)>. ..
>> The extra path loss of 5.6GHz vs 2.4 GHz is offset largely by the
>> reduced effects of folliage attenuation and building material
>> penetration. Antenna gains for identical physical antennas also
>> equalize the comparison. My experience is that multipath fades are
>> often minor compared to simple interference by nearby 802.11b radios.
>> Hard to tell even with a spectrum analyzer site survey.

>Hi Jeff;

Oh-oh.

>This bit is ambiguous. Clearly you do not intend to say that
>absorbtion at 5.6 is less than at 2.4, but it comes across that way.

Actually, I did, but managed to forget a few key words. My experience
with 2.4GHz vs 5.6Ghz is from trying to brute force RF through the
redwood and fir trees of the Santa Cruz mountains. This is your basic
very dense forest where foliage loss is the major consideration. I've
found that 5.6 tends to penetrate better than 2.4 because I'm shooting
through holes in the leaves and gaps between the leaves. This is also
the multipath nightmare from hell.

>For the record then, I am certain, having learned much from Jeff's
>posts over the years, he did not intend to suggest that 5.6 is
>absorbed less by vegetation or construction materials.

Thanks. I also like your corner reflector idea.
http://www.freeantennas.com/projects/Ez-10/
Suggestion: Put a door hinge at the back of the reflector and make it
portable.

Well, for a given type of material, loss at 5.6Ghz is much higher.
However, for a typical obstruction, the holes appear larger at 5.6Ghz
than at 2.4Ghz. For example, common 1" chicken wire used in stucco
construction allows 5.6Ghz to pass quite nicely, while seriously
obstructing 2.4GHz. Same with cracks in walls, gaps in tree branches,
venetian blinds, and other potential obstructions. While the material
itself attenuates more at 5.6Ghz, the increased relative gap size in
wavelengths makes propogation easier at 5.6Ghz. In addition, the
increased gain of point to point antennas at 5.6GHz improves the
situation even more. My one line conclusion was that these factors
tend to counter balance the increased free space loss and material
attenuation at 5.6GHz.

>Absorbtion at 5.6 GHz in vegetation is about twice as great as it is
>at 2.4 GHz or approximately 2 dB per meter of vegetation. This is
>consistent with the fact that absorbtion at 900 MHz is roughly 1/3 as
>much as it is at 2.4 GHz, or about 10 dB per 100 meters.

True. However, let's make a simple assumption. RF at 2.4 and 5.6
does not go through water, trees, or leaves... period. Assume a brick
wall. What's left are the gaps between the water, trees, and leaves.
My contention is that the increased gap (in wavelengths) allows more
propogation at 5.6 than at 2.4.

>Reflections ( multipath ) at 5.6 are about 8 times as frequent as they
>are at 2.4. This is consistent with the "Erskine's Multipath
>Principle" which follows:
>
>"For any given link the number of re-radiators which can cause
>multipath propagation is inversly proportional to cube of the wave
>length."

Agreed. Based upon an urban canyon model, the signal arriving might
be almost all reflections. The shorter wavelength allows more objects
to participate in becoming reflectors. Yet, many real world objects
are also absorbers. The ratio is determined by the environment. For
rural, forest, and suburban residential areas, my guess(tm) is that
it's mostly folliage and act as absorbers with few reflections. For
urban and high rise, it's mostly reflectors. You can't generalize
without considering the real environment.

>By this I mean, for any given path there will be a number of objects
>which can cause reflections. Those objects will be more or less
>efficient re-radiators depending upon their resonant frequencies and
>orientation to the signal path. As frequency increases, wave length
>decreases in a linear fashion. Therefore any object which is near
>resonance at frequency X will be FOUR TIMES as likely to be near
>resonance at frequency 2X. This is an immutable physical fact.

Well, yes. I can create an obstacle course that would result in fatal
reflections and that are much worse at 5.6 than at 2.4. However, I
can also create a similar path full of absorbers that will not reflect
anything. Again, it depends on the environment.

>The effect of these reflections on link reliability is a function of
>DATA RATE and not modulation scheme. Higher data rates are ALWAYS
>more suceptible to "inter-symbol smashing". This occurs when the
>direct and reflected components of the signal arrive with roughly the
>same signal strength and are out of phase by approximately 1/2 the
>length of a "symbol". For our puroses, think of a symbol as a binary
>digit. It takes a finite time to signal a binary digit across the
>link. If the two copies of the signal (the reflected bit and the
>directly received bit) are 1/2 the length of a bit apart, it becomes
>impossible for the receiving equipment to distinguish the current bit
>from the next bit in the stream and you get a very high bit error
>rate. The faster you signal your bits, the shorter time there is
>between bits. This implies another "LAW": "The number of points
>along a signaling path where a reradiator can be placed which also
>causes inter-symbol smashing is proportional to the square of the
>signaling rate." Put differently, double the signaling rate and you
>are four times as likely to see inter-symbol smashing due to a
>multipath condition.

Yep. That's the reason that ODFM (802.11a and 802.11g) allow for a
substantial number of variable data rates. They allegedly adjust the
data rate to minimize the number of trashed symbols. For example, if
the data rate were matched to have the delayed multipath signal
exactly coincide with every other symbol, then the thruput would be
half the data rate. However, that's considerably better than having
every symbol be partially trashed by a reflection. Also, because OFDM
multiplexes the various carriers over a wide frequency range, the
reflections tend to have different reflected delays at different
frequencies. Therefore, one or more of these reflected carriers
should always arrive intact. Therefore, you're absolutely correct for
a flat plat (broadband) reflector, but methinks not for the typical
real world, frequency selective, and random surface reflectors.

>All that said, multipath is not always your enemy. Just as you are
>four times as likely to see inter-symbol smashing, you are also four
>times as likely to see two reflections reinforce by arriving in phase
>and polarization. It follows from the information in the paragraph
>above that: "In any given volume the number of places where multipath
>reinforces the desired signal is proportional to the square of the
>signaling rate (time diversity effect)." and it follows from
>"Erskine's Multipath Principle" this corollary that: "In any given
>path the probability of multipath reflections arriving in phase and
>thus reinforcing the signal ((spatial) diversity effect) is directly
>proportional to the factorial of the number of reradiators in the
>path."

I beg to differ here. The degree of phase matching between the direct
and reflected signals are far to critical to assume reinforcement.
The reason OFDM works at all is that there are HUGE intersymbol gaps,
during which the delayed signals have time to arrive without getting
clobbered by the direct signal. I don't recall seeing any
consideration for direct reinforcement in any of the ODFM specs.
Since it can be easily assumed that the direct signal MUST arrive
first, the only way that symbol reinforcement can occur is if there
were no direct signal and the reinforcing signals were equally
delayed.

>All of this can be summed up as follows: "All other things being
>equal, in an environment with reflectors, the higher the frequency and
>the lower data signaling rate the greater the probability that all the
>reflections will combine at any given point to yeild a useable
>signal." This is absolute physical truth.

No way. The probability of reflections is highly dependent upon the
environment cannot be summarily estimated solely on the basis of RF
frequency.

Gotta run.

--
Jeff Liebermann (E-Mail Removed)
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 AE6KS 831-336-2558

Jeff Liebermann <(E-Mail Removed)> wrote:
> On 2 Jul 2004 00:06:47 -0700, (E-Mail Removed) (c hore) wrote:
> >I was thinking of putting in a Teletronics bidirectional
> >1 W SmartAmp on one side, but if it is a multipath problem
> >rather than interference, I suppose amplification won't help,
> >will it.
>
> 1 watt (+30dBm) and a 24dBi antenna are over power and illegal. Your
> maximum legal for an omni is +36dBm (1 watt plus a 6dBi omni). For
> point to point, the formula gets complex in that you're allowed 2dB of
> gain for every 1dB drop in tx power.

Right. I should have mentioned that this particular installation
is in a location which does not [at least yet] have the US or
similar quantitative power limitations. Also, minor clarification,
I believe you meant to say that for fixed point-to-point in US,
you're allowed [an additional] 2 dBm of radiated power [above 36 dBm]
for every 1 dBm drop in tx power [below 30 dBm]. Or equivalently,
you're allowed an additional 3 dBi of antenna gain for every
1 dBm drop in tx power.

30 dBm + 6 dBi = 36 dBm
29 + 9 = 38
28 + 12 = 40
27 + 15 = 42
26 + 18 = 44
25 + 21 = 46
24 + 24 = 48
23 + 27 = 50
22 + 30 = 52
and so on...

Jeff Liebermann <(E-Mail Removed)> wrote in message news:<(E-Mail Removed)>. ..
> On 2 Jul 2004 09:21:10 -0700, (E-Mail Removed) (Michael Erskine)
> wrote:
>
> >Jeff Liebermann <(E-Mail Removed)> wrote in message news:<(E-Mail Removed)>. ..
> >> The extra path loss of 5.6GHz vs 2.4 GHz is offset largely by the
> >> reduced effects of folliage attenuation and building material
> >> penetration. Antenna gains for identical physical antennas also
> >> equalize the comparison. My experience is that multipath fades are
> >> often minor compared to simple interference by nearby 802.11b radios.
> >> Hard to tell even with a spectrum analyzer site survey.
>
> >Hi Jeff;
>
> Oh-oh.
>
LOL.

Good comments and thanks for the points about holes. Where I was
thinking obstructions you were thinking holes, both perspectives are
valid and I'll have to spend a bit of time playing with that concept
in my head before I can completely agree with you.

Regarding the corner reflector, yeah there is so much I need to do
with that web site. I am getting about 20k unique IP addresses
visiting per month since the Tech TV presentations. Thanks Patrick!
I am working on my personal web site right now and will be getting to
freeantennas over the next month or so. There will be a lot of
changes and hopefully better presentations of the information on the
site.

Have a good holiday.

-m-