22 mile link screen shots

[WHT]

The actual link distance is 22 miles, to the ACK is set for 26 miles.

My signal level is as high as -68 late at night when its down to the lower 60s, and as low as -71 when over 100º.

Look at my TX/RX data rates.

You can see in the site survey I'm shooting right through another WSP's 5.8 Gig system ON THE SAME CHANNEL! Their tower is five miles north of me and about half a mile from my center beamwidth. Effectively my path is only 10 dB above the noise level then.

The Bullet 5M radios have survived several thunderstorms and power outages without any UPS or surge protection. This link has been up since the end of April.

http://whtonline.com/portal/main_page.png

http://whtonline.com/portal/site_survey.png

[danielh]

tasty :)

[drwho17]

Stupid observation, why does the Uptime say 01:23:23.

[UBNT-Mike.Ford]

Stupid observation, why does the Uptime say 01:23:23.

Firmware updates or the beta testers :)

[WHT]

I didn't install the latest firmware yet. I tried it last week and lost connection talking to the station with the previous release. Will update them both this week.

The uptime is low because I was playing with some settings and rebooted it.

[avantwireless]

Do you have any screenshots of measured TCP throughput for this link? All fine and well that it holds a connection, but what you are getting through it is the acid test. Just that the tx/rx is running at 65mb/s doesn't show data going at that speed... Specially with the other guy running the same path. Any chance you could also run this test with at 10mhz channel? This is way interesting as it's so important to ignore other N devices and to use the frequencies more efficiently...

Thanks!

[leo_cakep]

i wanna see TCP bandwith test on those link.... thanks :D

and comparison between 5mhz, 10mhz, 20mhz, 40mhz :D

[WHT]

I'm gonna set up an FTP server at the head end and download files at the tail end to test actual data load transfer. The first firmware release had an erroneous speed test, so they left it out for now.

[skyhook]

Great!

[WHT]

Just that the tx/rx is running at 65mb/s doesn't show data going at that speed.True, but that's another metric. What I'm showing is the transport speed. If moving large dumps of data thought it, the only imparted slowing down would be from the LAN to madwifi interface or LAN to Atheros engine.

I'll upload a DVD to the server tomorrow night and then download it back.

[wispwest]

I've had 54Mbps RX and TX before and only pull 1.8Mbps TCP. Then lowering the data rates of coarse brought me UP to 23Mbps on a 25mile link using XR5's.

[wispwest]

WHT, why aren't you utilizing the 5.4-5.6 spectrum? That's the "cleanest" area!

[WHT]

Because you can't legally do that in the U.S.

Edited and added....

That's because you are limited to a 6 dBm EIRP. So that isn't going to work for long distance where you need a 52 dBm EIRP.

[WHT]

Cold front is moving through the area and signal dropped from the normal -68 to -69 dB down to -71 to -72 dB.

TX/RX Rate: 65.0 Mbps / 58.5 Mbps to 65.0 Mbps
98% to 100% CCQ

[pnolan1069]

28dbi, 4degree horizontal beamwidth antennas used for this test?

[WHT]

One is 4 and the other is 6 degrees.

[netmaster]

I'm gonna set up an FTP server at the head end and download files at the tail end to test actual data load transfer. The first firmware release had an erroneous speed test, so they left it out for now.

well, did you perform actual data transfer tests. How the results was? On-Air speeds 65M or whatever do not tell anything about real transfer speeds.

[WHT]

I have to try it again. The PC I installed my FTP server on is running SQL and that really hogs the CPU. I was getting only 2 Mbps data transfers. Will try it on a free PC tomorrow, *IF* I left it turned on yesterday.

[Ben]

amazing diagram!

I'd like to see some construction pics, do you have that?

Thank you

[avantwireless]

That is what I am worried about. There is a far stretch between 65mb/s signalling, and 65 mb/s throughput...

I'm going to put my bet that you see about 15mb/s when you get your test all going...

[Dave-D]

WHT, you get some great results!

I'm noticing some big variations in over-ocean
signal strength. During mid-day, I often see -50dbm.
But at night, it deterioriates to as low as -72.

I don't think this is weather-linked, and this patch
of ocean has been fairly calm.

Background is -95, so I don't see any data drops.

Do you see big variations like these? Dave

[WHT]

I'm noticing some big variations in over-ocean
signal strength. During mid-day, I often see -50dbm.
But at night, it deterioriates to as low as -72. That's to be expected. During the hot day, there is a lot of evaporation and the air is very humid and thus not very dense.

[Edited - I got my syntax confused, what I meant to type was: At night the humidity drops and the air IS MORE DENSE.]

Its the density changes what make a difference in how the signal gets bent up or down.

[Dave-D]

'At night the humidity drops and the air isn't as dense.'

I know you meant to say, '...air is more dense.'

Your experience is that air density swing has that much
effect (~20dbi) on signal? That would suggest that winter
could be a real problem--because the lower temperatures
(coupled with very dry air) will cause a huge increase in
density too. Dave

[WHT]

Opps...let me clarify that.

Day time air is both both hot and humid, thus NOT very dense.

Night time air, while not as humid, but much cooler is MORE dense.

[WHT]

Jeeeeze...My written syntax is about as bad as UBNT's firmware naming conventions. :lol:

[Dave-D]

Well, sorta. But not....

Sun heats the ground, making air warm and moist (less dense).

Hot (invisible) air rises to the level where its moisture cools
and condenses into a cloud, and equalizes density.

Cloud (in general) stays where it's formed, unless pushed
even higher by warm air from below (such as a t-storm).

And in that case, it wants to stop being a cloud, and drops
its burden of moisture. Or, it's pushed so high and thin, it
becomes a veil of ice crystals.

So you think winter is worse than summer for 5Ghz paths?

[WHT]

We did some extensive testing three years ago that verified my predictions compared to Cisco's estimates for annual downtime versus fade margin.

Cisco assumes the environmental conditions are static, i.e. never changing from day to night, nor changing over the seasons. The only place I can think they would hold true would be on the leeward or downwind side of a mountainous Pacific island.

We found air density induced diffraction was at the peak about an hour after sunrise for two hours and around sunset for several hours. It also was during the spring and fall for a few week period. The fade followed a mid-path tower mounted hygrometer to measure the humidity about fifty feet above the ground.

[Dave-D]

Interesting!

So more dense (drier, cooler) air has more diffraction effect
on WiFi signals? And more diffraction means less signal
(or more signal absorption)? If so, why is it centered about
sunrise/sunset/spring/fall?

I can imagine this follows the humidity pattern in your area.
But air is most dense during mid-winter (cold and dry).

Is this absorption effect about density? Or maybe just humidity?

Thanks, Dave

[WHT]

The greater the difference of air densities at the interface, the greater the refraction (not diffraction as I may have miss-typed). It can be cold air underneath warm air, warm air over cold air, or dry air under humid air (or humid air over dry air).

If you have dense cool air under less dense warm (or humid) air, the radio wave will diffract at the interface towards the less dense air, or upwards. Conversely if you have more dense cool air moving over the less dense warm surface air, the wave will refract downward.

The diffraction effectively "re-aims" you antenna up or down.

During a cold winter day, the ambient air is very dense, but the sun warms the earth's surface and you have a layer of less dense warm air under the more dense cold air. Likewise during warm spring or fall morning after a cool night.

Air density is fairly static at a any given altitude, but changes as a cool front move through. But it also changes due to the humidity.

Absorption doesn't significantly enter into the equation.

[WHT]

I can imagine this follows the humidity pattern in your area. It follows the changes in humidity and changes in air temperature.

You can get some pretty chilly nights and blazing hot days in a desert or Phoenix, in both areas humidity its pretty low. It will play havoc with your signal.

[Dave-D]

So it's the situations where there's a boundary layer caused
by density layering that's the issue. And it doesn't matter
what the density of one layer is--it's the boundary that counts.

For example, on an over-ocean path, the water acts as a huge
heat sink--absorbing during the day, releasing at night. This
means in many cases you can't win: there's a boundary all
morning and all evening. (Except maybe during high wind.)

Thermals (caused by heating rock or desert) might actually
help, because they disturb the air and disrupt boundaries. Dave

[WHT]

And it doesn't matter what the density of one layer is--it's the boundary that counts.Exactly
an over-ocean path, the water acts as a huge heat sink--absorbing during the day, releasing at night. Open ocean water has enormous thermal inertia, so the air temperatures at sea level up to several hundred feet would be rather constant.

[avantwireless]

So did WHT ever get anything better than 1.5mb/s on this link? Or maybe it was 1.5MByte/sec. Still not anything near the 65mb/s negotiation speed. Time for some numbers, isn't it?

[UBNT-Mike.Ford]

Hello,

WHT said he was maxing his computer out so was switching the setup and will have more results in a bit.

Thanks,

Mike

[george]

Hello,

WHT said he was maxing his computer out so was switching the setup and will have more results in a bit.

Thanks,

Mike

Hey Mike, how's it hangin?

Interesting that no long distance test numbers are available yet. Considering a speed test is about the first thing most people do when they get a new toy to play with... The 'SQL server ate my homework' story is novel.

In addition, what the heck are you guys doing with the dancing around Airmax?

Surely this product is launched (and according to the email I got a couple weeks ago, its launched) or it isn't, in which case why are you guys teasing us at all?

As far as comments about M being the best thing since in existance for a greenfield build, I respectfully suggest you have a long way to go to prove that considering we can't even buy one yet, but I'm sure hopeful that you're right! :wink:

George - with 40 5Ms on order.

[UBNT-Mike.Ford]

Hello George,

I have seen the number first hand on 3, 6 and 17 mile links. I can say they are no joke. Our number are coming in a white paper. Our beta testers however are posting there information before we release ours and I have no control over there results.

Thanks,


Mike

[george]

Thanks Mike, that sounds encouraging.

We've done quite a bit of work with the new MikroTik N radios at extended ranges and our results are both slower and more variable than expected. Most of our results so far are using two chains though and the firmware is clearly still in beta so its not fair to critique it like you would a production release.

I really like the idea of a solid N radio using a single chain. Be interesting to see how the Ms do once we get our hands on some. I have a couple of links in mind, both with big dishes, one at 27 miles and one just over 30.

Assuming release is still on for early August, any idea when we might see the whitepaper and a statement on Airmax? As we're right in the middle of build season in the northern hemisphere more data will help to determine this autumn's activities...

George

[rconaway]

Hello George,

I have seen the number first hand on 3, 6 and 17 mile links. I can say they are no joke. Our number are coming in a white paper. Our beta testers however are posting there information before we release ours and I have no control over there results.

Thanks,


Mike

Those darn pesky data testers. Gosh darn if we didn't get the same result because "somebody" never tells about us about the latest firmware. I had to have my Uncle Louie "No Neck" Bonnano call to get it. Okay, so we saw 97Mbps instead of 104Mbps. I will have our techs beaten until they get to 104Mbps.

[WHT]

because "somebody" never tells about us about the latest firmwareBecause one of "us" didn't read Mike's email about the upgrade. :lol: Silly rabbit....

[UBNT-Mike.Ford]

You guys crack me up :)

[WHT]

Ok...tonight I'll be doing some tests. I set up a computer at the head end. 2.4 GB CPU, 3 GB RAM last Saturday. FTP worked great. Left there when the motorcycle store was closing.

Someone turned off the computer. Grrrrzzzzzzzzzzzzz.........

And they are closed on Mondays.

[UBNT-Mike.Ford]

Any update?

[avantwireless]

Supposedly mine shipped today.... So if I get mine tomorrow, I will be testing a 6 mile link the day after. Which begs the question....Updates? Is there somewhere to download updates to V? I wouldn't want to be testing with a bad version. If I do, I'll keep the results to myself....

[cameron]

Well... Results.... by Gosh Man Where are the Results!

I just got access to two towers that have LOS and 250km's apart I'm going to have a play once my M's arrive...

I am keen to see your throughput over 22miles.

[popcorrin]

I'm guessing the results weren't very good or we would have seen some by now.

[rconaway]

What results do you want? We have a few.

[WHT]

Weekend before last, I tried an FTP to a server that was running SQL and it was bogging it down.

Last weekend, they turned off the computer I intended to use minutes after I left when they closed the store.

Last night, my IT kid had his boyfriend in town so that put a damper on my testing.

Will try again today.

[popcorrin]

What results do you want? We have a few.

I'd be interested to see whatever you have. The long range numbers especially.

[rconaway]

Long range is WHT's specialty. He lives in land that that has more coyotes than people. Mine is high-bandwidth.

[WHT]

I caught this critter last week after my llamas had a round with him.

http://whtonline.com/deer_corn.jpg

[Ron]

Long range is WHT's specialty. He lives in land that that has more coyotes than people. Mine is high-bandwidth.

So, tell us, what TCP speeds did you get?

[rconaway]

We have hit 97Mbps on UDP traffic with a 40MHz channel. We hit in the low 50's with a 20MHz channel. Interestingly, the CPU overhead wasn't that much higher at 40MHz. However, firmware is still in the early stages to I would expect it to get better. Mike says he hit 104Mbps but that's just his my car is faster than your car personality.

[UBNT-Mike.Ford]

The "My Bullet M's are faster then your Bullet M's?"

:lol:

[rconaway]

No it's the "I have the programmers who keep my firmware one step ahead of the old firmware which I give you, nah, nah, nah nah syndrome. Kind of like us giving the rest of the world detuned F-16's so that if we go to war with them, we still win with better planes. I got your number buddy.

On another note, we just finished the show and got more attention from the other vendors than the cops. Not a lot of decision makers but there are a few vendors that now want to talk to us. Lots of projects got held up due to funding and the M sort of changes the game plan.

Too bad some of the mesh bigger players weren't there. That would have been fun.

[masked]

We have hit 97Mbps on UDP traffic with a 40MHz channel. We hit in the low 50's with a 20MHz channel. Interestingly, the CPU overhead wasn't that much higher at 40MHz. However, firmware is still in the early stages to I would expect it to get better. Mike says he hit 104Mbps but that's just his my car is faster than your car personality.

woh woh woh, hang on a sec, ~100Mbps UDP? The advertisement clearly says 100Mbps+ TCP/IP! I have already been fooled once by MT with their claims to 200Mbps Real World which didn't state they meant UDP!

Now can someone confirm we're going to achieve 100Mbps TCP/IP?! Surely you wouldn't have said it if it wasn't already true!? Believe me, I do trust you more than other said mob!

[rconaway]

I can't give you full TCP/IP numbers yet beyond our original ftp server test. In that test with a 20MHz channel, we hit 50+Mbps when we discovered that we hit the limit of the PCMCIA card on the laptop or the limit of the ftp server, didn't have time to start analyzing. Mike's numbers are based on a 40MHz channel which was useless with our ftp setup. We did test some other software but got conflicting information and haven't had time to go back and confirm it. Bandwidth testing is an art.

[masked]

Ok, so Mike, are you saying you've hit 104Mbps TCP/IP?

[UBNT-Mike.Ford]

Ok, so Mike, are you saying you've hit 104Mbps TCP/IP?

Yes this is correct, under ideal conditions with minimal interference.

Thanks,

Mike

[rkj]

We have hit 97Mbps on UDP traffic with a 40MHz channel. We hit in the low 50's with a 20MHz channel. Interestingly, the CPU overhead wasn't that much higher at 40MHz. However, firmware is still in the early stages to I would expect it to get better. Mike says he hit 104Mbps but that's just his my car is faster than your car personality.

woh woh woh, hang on a sec, ~100Mbps UDP? The advertisement clearly says 100Mbps+ TCP/IP! I have already been fooled once by MT with their claims to 200Mbps Real World which didn't state they meant UDP!

Now can someone confirm we're going to achieve 100Mbps TCP/IP?! Surely you wouldn't have said it if it wasn't already true!? Believe me, I do trust you more than other said mob!

The adding of input+output is a nasty habit of the market, not only Mikrotik's. Cisco switches and routers suffer this same "problem"... always test a device yourself, and tell the vendor your number. Then they can make improvements by config changes, but not by changing your test profile.

[rkj]

Ok, so Mike, are you saying you've hit 104Mbps TCP/IP?

Yes this is correct, under ideal conditions with minimal interference.

Thanks,

Mike

Could you list how much pps in both directions were flowing in this 104 Mbps test ?

[zed]

Another set of tests, showed about 94mbps burst with latencies going from 3- 350 ms

real time fdx with good stable 40-70 ms latency 50 mbps FDX


zed...

[rkj]

Another set of tests, showed about 94mbps burst with latencies going from 3- 350 ms

real time fdx with good stable 40-70 ms latency 50 mbps FDX


zed...

These latencies seem a bit high...
What is the distance and channel size of this particular test ?

[Blizz]

I notice the signal strength bar look slightly different than usual, what version of AirOS 5 used in this test, can pasted the configuration, we are going to do long range test too soon. Any timeline on the first update release of FW? thanks

[WHT]

The color bars in the earlier releases can be seen in the first post of this thread.

The latest firmware (XM.v5.0-devel.2401.090722.1132.bin) can be seen here: http://forum.ubnt.com/forum/viewtopic.php?t=13642

[kissas]

you tested with iperf?
if so, what are the results?

[WHT]

I haven't had time to do any more testing. Mebbe this weekend.