Re: Effective Range/Throughput - Comparisons and review - (see postings 31-01-06)

Hello Jeff,
How nice of you to participate. Its a little late. Hope this is readable ...

"Jeff Liebermann" <jeffl@xxxxxxxxxxxxxxxxxxxxxx> ha scritto nel messaggio
news:ai3cv15j5j5sg1cdn8da2omb0grkngbv85@xxxxxxxxxx
"nourght" <nav0785@xxxxxxxxxxxxxxxxxxx> hath wroth:

I just hate it when someone posts an interesting question and I don't
have enough time for a proper answer. This may come in two or more parts.

Glad that at least you find this argument of some interest.

http://www.wi-fiplanet.com/tutorials/article.php/1495031

http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless/Wi-Fi#Link_Calcu
lations

ftp://download.intel.com/business/bss/infrastructure/wireless/deployment/ho
tspot.pdf

In fact John Navas justly coments :
The rather optimistic ranges listed are probably with a better than
standard omnidirectional antenna and with unobstructed line of sight.

Ummm... I wrote that part. The numbers used by Intel are rather
optimistic and assume an ideal location, no interference, no
reflections, no obstructions, and radios that actually meet their
published specifications. The tests were also performed at fairly low
and useless speeds (1 or 2 Mbit/sec association).

I do beg your pardon, I was under the impression that it was from a text
snip
from John Navas, but I was working very late and perhaps I mistook the
author.

Does anybody know of an updated Field Performance Test Comparison Table
?

No, but you can make your own quite easily. I'll do it when I have
time. Start with your favorite access point and 802.11g laptop. Start

The idea was basically that of a common pool of data.
On my own I cannot test all the combinations or even all the models
The participation of any interested Wlan technician would be appropriate
here

in close to the access point and start moving some streaming data.
Streaming video or large file copies are good. Watch the connection
speed. When it drops from 54Mbits/sec to 48Mbit/sec, stop and measure
the distance. Move VERY slowly as it takes a while for the system to
react. It's usually about 15ft when I try it. Do it again at
different speeds until you run out of walking room. Make the speed
transition points on a graph. The points will be all over the graph
but a pattern will emerge.

In fact this was more or less the method I had adopted last weekend going
from
0 - 150m ca and using a Net Performance tool to measure the corresponding
throughput. Due to lack of time I examined only distance as the pivot factor
and
noted the measured throughput and corresponding SNR. 5 distances tested over
the range each run several times. The result was really rather upsetting.
Considering the quotes promised in the Tech Specs of the components employed
and the assurances, or should I say 'lack of ', of the so called qualified
techicians on
a well known Firms Support Desk, I am concerned that something is amiss.

I am a little surprised that this type of chart has not been developed by
members
of this and other similar Forums. It would not be difficult to create a
simple Web
Form with the necessary fields and relate it to a Database which can be
interrogated
or which can be downloaded.
As a simple base meter it would be simple to read as a reference and I
believe
very useful for elementary tests.
[Of course this supposes a professionally scrupulous and honest behaviour of
the
colleagues who wish to insert the records.]

Another (better but more tedious) way to do the test is to lock the
speed at the access point to some speed (i.e. 54Mbits/sec) and move
away until the error rate goes up sufficiently to be useless. Notice
I didn't say until it disconnects because that can take quite some
time. Using the Windoze performance meter to monitor thruput for a
file copy is a good indication of error rate. Repeat at different
fixed speeds.

This is an interesting point.
Sometimes AP's (and cards) when set to a modest fixed mode and speed
transfer
data with a more stable signal and, in the long run, better throughput than
when set
to Auto which may start at G/54 Mbs and then drop to a wavering
B/11,14,8,12,
17,6 etc. I have read confirmations of this from others also. Comments?

You can also extrapolate from a few points by using the receiver
sensitivity change versus connection speed. See the first table at:

http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless/Wi-Fi#Link_Calcul
ations
Inverse square law says that -6dB change in power results from a time
two change in range. If you get 15ft at 54Mbits/sec, and the receive
sensitivity is -68dBm, then at 6dB less, you'll have twice the range.
Looking at the chart -74dBm sensitivity is 36Mbits/sec. Therefore, at
30ft, you should slow down to 36Mbits/sec.

eg : Card Name/AP Name/Mode/Distance/Signal/Throughput etc
and then testing through the various distances until the signal
is lost. Then change mode, then change X, change Y etc.

I have some results but they belong to the client. However, I can
easily run the tests again. Be prepared for huge variations from
theoretical.

Concerning this theme, I would be very interested to hear from you.

One could the compare the Named product couple with others
[entered as records] dealing with, for example, the same AP but a
different card.

The throughput should be tested using an impartial Net Tool not
the Product Control Module if it has one.
Ixia QCheck is a good Win possibility and in Linux there are
plenty of choices.

That's fine. Just remember that there needs to be a reference level
for BER (bit error rate) or thruput drop at the various distances. You
could do something sloppy like "when the file copy thruput drops in
half" or something similar. It doesn't matter what reference you use
as long as it is consistently applied.

Personally I have never witnessed a fully working efficient Wlan link
at 150m indoors going through numerous walls and other obstructions.

How many indoor buildings are there that are the size of a football
field or larger? Obviously, 150 meters is not an indoor test.

This of course was a rather caustic comment of mine :-)
It was based precisely on a couple of Spec Sheets from Linksys in which
they state that the cards cover such 'indoor ranges' (B mode.)
However it is equally true to say that I have not witnessed a fully fledged
Wlan running G Mode over an 80m Warehouse with a constant 24 Mbs.
I would be interested to know who has and how many walls etc the signal
passed through.
[Remember I proposed this query considering standard Antennae and non
boosted cards or AP's.]

What's a wall worth in dB? I collected a few charts of material
attenuation and posted them at:
http://www.thirdbreak.org/pipermail/wireless/2005-June/000804.html

Thanks for the link, tonight however it seemed dead.
Please post an update, I am interested to add your data to another table
that I have been assembling recently.

Let's pretend that a "typical" wall is good for -6dB attenuation. That
means for every wall the signal has to go through, the maximum range
is cut in half. Actually, the combination of reflections and material
attenuation result in far worse range loss.

I have read of several firms who pretend such performance and yet when
I went out and tested the stuff the result was disappointing, disastrous
would
be more appropriate.

Yep. They probably tested this with a production access point, an
insipid antenna, a highly reflective office environment, interfernce
coming in through large windows, and a laptop with its worthless
antenna. These are far from ideal, but represent the most common
wireless arrangement.

Let's play with the numbers and see what worst case produces instead
of best case. Going to the link calcs at:

http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless/Wi-Fi#Link_Calcul
ations
and plugging in some numbers. A WRT54G dribbles about +15dBm tx
power. The antenna is about 2dBi gain, but the coax and connectors
eat about -1dB in loss. The laptop uses an marginal receiver with
about -88dBm sensitivity at 6Mbits/sec connection speed, the slowest
OFDM speed. The ceramic backed antennas in the laptop lid are good
for about -2dBi gain and another -2dB in coax and connector losses.
I'll aim for 20dB fade margin (just to be consistent).

TX power = +15dBm
TX coax loss = 1dB
TX ant gain = 2dBi
Distance = unknown
RX ant gain = -2dBi
RX coax loss = -2dB
RX sens = -88dBm (at 6Mbits/sec)
Fade margin = 20dB

Plugging into:
http://www.terabeam.com/support/calculations/som.php
and trying various distances until I get about 20dB fade margin, I get
a range of 0.06 miles or 316ft. That's quite good at 6Mbits/sec.
Maybe I'm being too optimisitic somewhere.

100 m at 6 Mbs ?
Perhaps the 88dBm sensitivity at 6Mbits/sec is better than the common
cards on the market.. I guess much much better than the card I had
during one of the tests (WPC54G <--->WAP54G).
I have noted that few cards (Orinoco, Cisco, Senao, Denmarc etc)
approach or surpass the 90dBm and generally they are already down to a
couple of Mbs. Do you agree ?
Coincidently I had downloaded the Terabeam SW some weeks ago.

Now, running the same numbers with a receive sensitivity of -68dBm at
54Mbits/sec, I get 0.006 miles or 32ft. So, your range is going to
be between those numbers (for OFDM).

Exactly.

eg: Table for AP's/Card's expected field performance
Conditions : Perfect LOS, Standard Antennae and Original FW.
Standard environmental conditions and a non tweaked setup
[==> Little interference for optimal initial conditions]

Gaaak. Where am I going to find an interference and reflection free
environment the size of a football field? Over water?

:-) As you stated before, one needs a base reference.
I was merely considering typical differences in well known conditions.
eg: in the city center /in the country (obviously not under a high tension
cabling, in a nuclear power station, near industrial motors etc.
and of course during the primary tests if possibile without GSM, faulty
electrical equipment nearby, nieghbours disturbing stuff on etc
Or the difference from in the work office surrounded by offices and all
the general 'background noise' to a slightly less EMF polluted area
temporarily without the usual interferences (residential or otherwise)
Just to enable a slightly simpler interpretation of the primary results
from the tests. This does not of course pretend the unattainable silence.

On 100+m (not >400m) I have found that the signal was low
and the throughput was only usable for an Internet connection.
Little chance of considering a real Office Client/Server connection.

Well, what are you standards for an acceptable connection? The 20dB
fade margin is 100 times the signal necessary to produce the BER of
1*10^5 reference ( 1 error for every 10^5 bits). That's more than
adequate for office communications.

I was considering her basically two types of simple 'acceptable
connections'....
(for simple client setups that is.)
We are yet considering medium or large installations. The aim here is to to
be able to obtain a reasonable idea of the working real life potential of
the
components in recognisable conditions and then extrapolate from there
to have an idea on how best to work in less favourable conditions

The first is a connection stable and fast enough to support general modest
size office work (text files, spread sheets, printing, modest DB client
server
apps, and of course Internet apps. I believe from experience that people
start complaining when they go under 10 Mbs for the first part and nowadays,
with DSL around, also if they go under a constant 0,75 Mbs x Internet Apps.
This is for point to point, we are not yet considering concurrent users
and shared bandwidth.
As said, its to test what can be done as a max in basic P2P test.
It's pretty clear that for any installation that forsees a large number of
users
and/or several floors and/or certain apps a survey and a study needs to be
planned. However notwithstanding this a simlar impartial table of field
results
might be very useful to many persons. (Suppliers included !)

As an asside, I have a customer that has some goofy cash register
program that absolutely will not tolerate any packet loss. Their
database client server program literally hangs or dies if even one
packet is lost. It's not a security feature, just old crappy design.
If you're having problems maintaining database intergrity over a
wireless link, check on how well you application tolerates packet loss
before you assume it's a wireless issue. May I suggest DummyNet at:
http://info.iet.unipi.it/~luigi/ip_dummynet/
for simulating packet loss.

Interesting link, thanks.

It was down to a variable 1-peaks of 2 Mbs.
At 150m the signal/connection dropped too often]

What were you using for testing? How many walls?

As stated before Outdoors pure LOS = no walls
(WPC54G <--->WAP54G).
In fact I am concerned that the AP might be faulty in some way.

Some other experts consider that there is far too much Hype
around and that a good connection is only available at around
60m max

"Good" is how how many dB of fade margin over the receiver sensitivity
(at a BER of 1*10^5). Per my first calculation, one should be able to
do 300ft at 6Mbit/sec OFDM association. If not, either one of my
assumptions is off, my calculations suck, or some part of the puzzle
is not meeting published specifications. I really don't know which is
the case here.

Interesting. As stated above the Receive Sensitivity you mentioned for that
throughput IMHO may be better than some common AP's/Cards offer.
Apart from this Linksys does not include this value in their Spec Sheets and
after a hopeless series of encounters with their Support desk it seems that
they still refuse to give out this technical type of info.
[Other firms are far more efficient and mention plenty of values at
different
frequencies and ranges.]

In fact this is far closer to my findings than that written on the
Product Spec Sheets.

To have a Signal Connection or Throughput Peak is of little
interest here.

Take for example the often quoted' standard WAP54 series with
a standard setup, no encryption etc.
Take a standard Card (eg. WPC54 series)
What have you found ?
At 10 meters save a 10 MB file to you server,?
What was the duration/signal stability/throughput ?
Note other ranges up to unusable.
Has anybody managed to upload a 10 MB file over
a 500 meter link with a fully efficient B connection using standard
out of box 2dB antennae?

Any comments, suggestions or links would be most appreciated.
A table with the conclusions could be published here for benefit of
all, including any comments on the efficiency of the Firms Support
team to resolve any issues evolving during your experience.

Thanks

Gotta run...

Thanks for the presence.
Hope to hear from you.

___________________________________________________________

Jeff Liebermann jeffl@xxxxxxxxxxxxxxxxxxxxxx
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558


.