M2HP 20MHz BW ok, 10 and 5 MHz No Link - Help?!

[jlewisfl]

I have a marginal link (noise -96, signal varies from -94 to -86, with a 24dBi dish at each end), and have tried limiting RF bandwidth to improve the S/N, but no go. I can set both at 20 MHz and they link (two new M2HP bullets, same model, both with test dates of 7/24/09), If I set them both to 10 MHz, the link is immediately gone, and does not come back, Same if I set them both to 5 MHz BW. One is Ap, other is Station, running as a bridge. Theoretically, I would expect a more robust link, but with lower max thruput, with 10 or 5 MHz RF Bandwidth, but neither 10 nor 5 will link. We had one person running the Bullet M2 AirOS console at each end and coordination by cellphone to make sure we were running same BW. Any suggestions? Not running in WDS mode, but using WPA2. That does work - keys match up ok, and I can use MAC ACL or not, no effect on the link.

[WHT]

How far is the link distance? Did you properly set the ACK interval?

[jlewisfl]

I have the link distance set for about 50% more than the actual distance, which is roughly 3 KM. If this was set too short I would not expect to have any link at all. The problem seems to be they won't work at less than 20MHz RF Bandwidth. One other point - we also tried the Channel Shifting function, when we enabled it, that also broke the link. We changed at both ends, same time, and the link never re-established. If I recall, we were using channel 3 at the time, which seemed to have the best S/N at both ends of all 11 available channels. I was really hoping to use the 10 or 5 MHz RF bandwidth settings plus the Channel Shift, to get a better S/N. But each of those options, coordinated at both ends, broke the links and would not re-establish. Set them both back to 20 MHz and disable Shifting, and they had successful handshakes and showed a link within ten seconds every time.

[WHT]

I'd default the radios and get back to a known starting point. Only change one of the radios to AP and set the ACK distance to 10% greater than physical distance. Make sure your antennas are aimed. Don't change anything else for now and see what happens.

[jlewisfl]

As it stands, one is set to AP and one is set to Station. Only have two discrete points involved, just need to link between them. I aimed the antennae by first drawing the link on paper after printing a multi-page hi-def Google SatMap, to get gross aiming settings, then by setting one Bullet to AP, and aiming the Station antenna for max signal on the Station, then reversing the AP/Station config and doing the same in reverse, so I'm fairly confident on antenna pointing - we fine-tuned it with a person at each end, watching the AirOS at each end, and coordinating in real time with cellphones.

I'll reset the ack to distance + 10%, and I can certainly go back to defaults on everything else, but what I REALLY want to know is why I lose the link permanently when we drop to 10 MHZ RF BW or 5 MHz RF BW, or switch Channel Shift ON. Why, regardless of other settings, would they link and WPA2 handshake at 20 MHz RF BW, with Channel Shift OFF, but DO NOT work at 10 or 5 MHz OR with Channel Shift ON? Has anybody else seen this, or has anyone even tried narrower bandwidth settings with M2HP and AirOS 5.0?

[drwho17]

We are using Bullet2M-HP's as clients with 10mhz channel width just fine.

[jlewisfl]

But are you using linking to those M2HP clients with an M2HP set to 10MHz RF BW AP? Or are you using the M2HP's just as clients with something other than an M2HP as the AP?

[shapiros]

Seems I have read somewhere that 802.11n requires 20MHz|40MHz. If so, I would venture a guess that it applies to these devices as well. However, need Mike or someone to verify|deny this...

Sig

[jlewisfl]

I don't care which mode they actually use. All I want is for those two bullets to link and transfer data. I don't need the data throughput that n, or g for that matter, provides - I don''t need more than a few Mbits /second - 5 Mbit/sec would be max that we would really need unless something drastic happens with what we are trying to use it for - this is not a streaming video, or large-file-transfer situation. I'm not going to complain if it goes to a higher throughput, obviously. But if they WON'T operate at 10MHz or 5 MHz bandwidth, WHY are those options in their AirOS firmware? THAT is what doesn't make sense,

[shapiros]

My reply was speculation that needs Mike or someone to weigh in.

As far as, if true, why it is in the AirOS options... You obviously have never worn the "SW developer" hat. From that perspective, it would be an "undocumented feature"..."provided at no additional charge"...

Sig

[UBNT-Mike.Ford]

Hey Guys,

These units should work in 5 and 10Mhz without issue. I re-verified in the lab.

However we are investigating any issues reported. Please provide me with any pertanent info, along with support and config files to help me with my investigation.

Thanks,

Mike

[jlewisfl]

The following was posted before I saw Mike's response. I have PM'd him with more info...

-------------------------
While we wait for a definitive answer, "undocumented feature" in something you bought and are expecting to function as advertised, doesn't carry much weight. I run mostly Linux, have six or seven machines running here at the house, another half dozen at the other end of the link. At one time or another, from 1968 on, (ok, I'm a dinosaur, I started with punched cards on an 026, then an 029 keypunch, using batch FORTRAN on an IBM 360-030), then early BASIC on a GE timeshare using KSR and ASR Teletypes.... I've been at least semi-fluent with FORTRAN, BASIC, PASCAL, SAS, SPSS, and HTML at one time or another. I have written quite a lot of "undocumented features" i.e., bugs, but I've always fessed up and fixed them as soon as I was made aware of them, although usually I found and fixed them before they ended up becoming somebody else's headache.

[shapiros]

Did not mean to be deprecating. Especially to a fellow dinosaur.

Can you say Jacquard Device? I HATED it when the mechanical pins would stick and mess up an entire paper tape. Punch cards was definitely a giant leap forward, at least, until you dropped your assembler that had to be turned in in 15 minutes...

Sig

Hope Mike figures it out...

[jlewisfl]

Yeah, dropping a two foot stack of FORTRAN cards and then trying to get them back in exactly the right order was NOT fun. As a ham I had a whole bunch of 14 15 and 19 Teletype stuff (5bit) for RTTY plus some machines actually made by Western Union (model 100's) that were a Rube Goldberg contraption they designed to avoid all the Teletype Corp patents. Eventually WU gave up on them, gave away the model 100s to hams or the local dump, and went to Teletype equip.

Right now the two Bullets are linked and data is flowing, but only if 20 MHz is selected. Select 10 MHz on both of them, and neither sees the other or anything else. Set them back to 20 MHz, and the link is back in business in about 10-15 seconds. But right on the hairy edge due to low S/N, which is why I REALLY want to try them at 10 or 5 MHz. I'll send Mike a cfg dump of the two of them tomorrow. He says they should work, So something in cfg must be fouled up, even though they work ok at 20 MHz.

[WHT]

FORTRAN, BASIC, PASCAL, SAS, SPSS, and HTML
Where would the world be without "C"

[WHT]

"obal"
"pasal"
"basi"

[opampca]

The 10mhz channel width works perfectly for me.
I use it on a link of 3km over the city, and going thru 100 meters of trees.
B2M to B2M, 24dbi dish and 19dbi vagi, I obtain 3.5Mbps with -65dbm and CCQ of 48%.
With 10 streams on jperf.
If I get out of the trees, forward 100 meters, I get 10Mbps at -57dbm and CCQ of 99%.

[shapiros]

Just curious, what "IEEE 802.11 Mode" are you operating at? Looked back through the posts and did not see. I also looked it up and basic 802.11n requires 20|40MHz. Makes sense since the wider the BW, the better the throughput. Don't know if UBNT has done anything different with their devices.

One other item Mike, I hooked up two BM2HP's late last night and while in "N" or even "G/N", I see the same behavior as jlewisfl. If I switch to "B/G" and change to 5|10MHz, they work fine. Does this verify anything, don't know but is another observation...

Sig

[ljenkins]

Seems I have read somewhere that 802.11n requires 20MHz|40MHz. If so, I would venture a guess that it applies to these devices as well. However, need Mike or someone to verify|deny this...

Sig

I certainly hope that's not the case. Mike?

[ljenkins]

I have the link distance set for about 50% more than the actual distance, which is roughly 3 KM. If this was set too short I would not expect to have any link at all. The problem seems to be they won't work at less than 20MHz RF Bandwidth. One other point - we also tried the Channel Shifting function, when we enabled it, that also broke the link. We changed at both ends, same time, and the link never re-established. If I recall, we were using channel 3 at the time, which seemed to have the best S/N at both ends of all 11 available channels. I was really hoping to use the 10 or 5 MHz RF bandwidth settings plus the Channel Shift, to get a better S/N. But each of those options, coordinated at both ends, broke the links and would not re-establish. Set them both back to 20 MHz and disable Shifting, and they had successful handshakes and showed a link within ten seconds every time.

I had this very same problem with nano2's. Never resolved either. I have to use 20Mhz on those units or in 5/10mhz mode with the units at about 20 feet from each other at auto/manual ACK. After 20ft the link dies and never comes back. haven't tried the bullets though, I am still waiting on mine.

If you can find the time get the units close together (20-30ft) and see if they link up like my experiment did.

[shapiros]

Going to have to "back out" of my previous statement about verifying "N" REQUIRES 20|40MHz. Took the intiative to scower ieee802.org's website and could find no such affirmation. Without such, people's individual interpretation doesn't mean "squat"! I should have done such before making such a statement. Sorry folks.

However, still trying to interpret several of our observations...

Sig

[jlewisfl]

On my M2HP Bullets, the AirOS wireless mode choice is " Ng ". If I click the select tab, to change modes, " Ng " is the only choice. There really is no option, in other words.

The units are on tall push-up steel masts (Rohn) and a pain in the butt to raise and lower, and I also taped the cat5 to the pole every few feet, with slack in it, to avoid stress on the line or the connector. So lowering one to take the unit off, transporting it and the antenna to the other site, lowering that one so that I can put the two a few feet apart, is not going to happen. My back is still sore from putting all that stuff up and getting things pointed.

Reducing the bandwidth from 20 to 10 or to 5 breaks the link. So does channel shifting. And obviously, this means the same change was made at the same time to both ends.

Right now the link is working with 20MHz but with very slim fade margin - best signal seen was -87, worst to still function at all was -94, with a floor of -96. I was anticipating at least 6 dB, or better, S/N improvement at a 5MHz bandwidth, but obviously with a lower max thruput...

If it rains, we lose the link until foliage dries out.

[UBNT-Mike.Ford]

I certainly hope that's not the case. Mike?


Hello,

The N protocol (at least with our chip and drivers) works fine in 5/10/20 and 40Mhz.

However, I believ it may have something to do with very low RSSI's (less then 16)when operatingin 10/5Mhz. I will let you know more when I know more.

Thanks,

[jlewisfl]

Thanks Mike. If I interpret that correctly, it means that at present you can't use 10 or 5 to improve a lousy S/N on a link because they don't work if the signal is lousy. Kind of defeats the purpose of a reduced BW. Hope you find a way around it, or perhaps build a fix into the next firmware.

Joe

[UBNT-Mike.Ford]

Thanks Mike. If I interpret that correctly, it means that at present you can't use 10 or 5 to improve a lousy S/N on a link because they don't work if the signal is lousy. Kind of defeats the purpose of a reduced BW. Hope you find a way around it, or perhaps build a fix into the next firmware.

Joe


Hey Joe,

Im thinking of really lousy signals (around -82 or less percieved signal strength while in 10Mhz. This would be a -86 signal in 20Mhz). And its a just a software fix we are looking into.


Thanks,

[jlewisfl]

Thanks Mike. This current S/N qualifies as a really, REALLY lousy signal, -86 is about as high as the received signal gets, with an occasional giddy foray to -84 when the planets and the moon are properly aligned and the wind is calm, with a noise floor that stays at -96. Most of the problem, I'm pretty sure, is co-channel interference, and that also should drop down since the interference sources will stay the same spread but we won't. Theoretically, I would expect the effective punch of the spectrum slice of ours us to go up by 6 dB, while at the same time the effective interference potential from a given co-channel source will drop to 1/4 what it was, or something along those lines. Being able to half-channel slide will probably be additional help after the 5 MHZ BW works as it should, and THAT does not work NOW even at 20MHz BW - the link disappears and does not return until that feature is turned off (and obviously we are doing this simultaneously at both ends).
Believe it or not, though, it is currently a usable data link about 75-90 percent of the time even at these levels and 20MHz, with a throughput of (rarely) up to 11 Mb/s. But you can see, I think, why I REALLY want to be able to narrow it down. It'll be slow, (probably not, on average much slower than it is now) but it should be MUCH more robust.

[UBNT-Mike.Ford]

Thanks Mike. This current S/N qualifies as a really, REALLY lousy signal, -86 is about as high as the received signal gets, with an occasional giddy foray to -84 when the planets and the moon are properly aligned and the wind is calm, with a noise floor that stays at -96. Most of the problem, I'm pretty sure, is co-channel interference, and that also should drop down since the interference sources will stay the same spread but we won't. Theoretically, I would expect the effective punch of the spectrum slice of ours us to go up by 6 dB, while at the same time the effective interference potential from a given co-channel source will drop to 1/4 what it was, or something along those lines. Being able to half-channel slide will probably be additional help after the 5 MHZ BW works as it should, and THAT does not work NOW even at 20MHz BW - the link disappears and does not return until that feature is turned off (and obviously we are doing this simultaneously at both ends).
Believe it or not, though, it is currently a usable data link about 75-90 percent of the time even at these levels and 20MHz, with a throughput of (rarely) up to 11 Mb/s. But you can see, I think, why I REALLY want to be able to narrow it down. It'll be slow, (probably not, on average much slower than it is now) but it should be MUCH more robust.

Hello,

Please update to the new RC2 firmware I released today at the top of the forum. Once update, disable the noise immunity option on the advanced tab and your problem should be resolved.

Thanks,

Mike

[drwho17]

But are you using linking to those M2HP clients with an M2HP set to 10MHz RF BW AP? Or are you using the M2HP's just as clients with something other than an M2HP as the AP?
Bullet2HP's are the AP's.

[jlewisfl]

Mike, I downloaded build 2732 RC2 090918.1605, and successfully uploaded it to both and installed, and rebooted both and they show they are running it, but there is no "noise immunity tab anywhere that I can see. I do notice that the link is a little more robust, immediately. However, I notice that the "Country Code" select on the M2HP I am using as the far end "Station" is blank, and there are no countries shown as options when I click that tab. The M2HP AP unit shows the countries and the select tab works. Should I try to upload again to the "Station" Bullet?
Joe
UPDATE: I re-uploaded the new firmware to the Station Bullet and now Country Select works. Then I took a deep breath and tried 10MHz. IT WORKS, and improved the signal levels and CCQs. Then I went to 5 MHz, and have even more improvement, as I had hoped. CCQs now running 80-100%. THANK YOU!!! Signal levels now in the -87 to -81 dbM level, Huge improvement in link. Still no "noise immunity tab anywhere, though. Running MCS1 - MCS3 modes now.

[UBNT-Mike.Ford]

Mike, I downloaded build 2732 RC2 090918.1605, and successfully uploaded it to both and installed, and rebooted both and they show they are running it, but there is no "noise immunity tab anywhere that I can see. I do notice that the link is a little more robust, immediately. However, I notice that the "Country Code" select on the M2HP I am using as the far end "Station" is blank, and there are no countries shown as options when I click that tab. The M2HP AP unit shows the countries and the select tab works. Should I try to upload again to the "Station" Bullet?
Joe
UPDATE: I re-uploaded the new firmware to the Station Bullet and now Country Select works. Then I took a deep breath and tried 10MHz. IT WORKS, and improved the signal levels and CCQs. Then I went to 5 MHz, and have even more improvement, as I had hoped. CCQs now running 80-100%. THANK YOU!!! Signal levels now in the -87 to -81 dbM level, Huge improvement in link. Still no "noise immunity tab anywhere, though. Running MCS1 - MCS3 modes now.


Thanks for the update. THe noise Immunity button should be on the advanced tab of the Web GUI.

Thanks,

Mike

[WHT]

The noise Immunity button should be on the advanced tab of the Web GUI.One of my techs said he wished his wife had a Noise Immunity button.

[Full Power]

I have a marginal link (noise -96, signal varies from -94 to -86, with a 24dBi dish at each end)..
Seems like everything after that statement should have come to a screeching halt. With a pair of 24 dB grids at a range of 3 KM you should have a signal in the -30 dB range. Google up a link calculator, and then check your antennas for water.

[jlewisfl]

Mike, no noise tab or button on Advanced or anywhere else on either Bullet. And on the AP now the distance slider is an empty box, and the two RTS and fragmentation boxes say "off" rather than having a numerical value. If I go to change any value on that page now, it complains that RTS and frag must have numerical values. But when I input them and Apply, they go back to "off". Station Bullet does not have this problem. I have both open in the same browser, different windows, via the link.

Joe

[jlewisfl]

To Full Power - I know. I am only up 50' at one end and 40' at the other, Rohn push-ups. Several heavy forested areas are the culprit, I am going to try to give it another 6' at each end and see what happens. If I had a couple of 100' towers I could use two tin cans and a string.

[Full Power]

Several heavy forested areas are the culprit..Forgive my presumption, I should have asked about it first.

As coincidence would have it I just finished a 2 km link through a forest, with a 24 grid at one end. I have a -63 but cannot hold a link rate any faster than 6 Mbps with 1 Mbps TCP throughput. Maybe the trees are blurring the RF sufficiently to cause a loss of data purity. This was my first shot through trees.

[rconaway]

Trees are weird. I have seen stuff through trees that would make your hair fall out. If you go through trees, all bets are off. Trees I can see through can't get a signal 600'. Trees that look like the great wall of China act as though they aren't even there. The best test would be to change polarity and see what happens.

[WHT]

"Blurring" is exactly what is happening. Its called the "grove effect" where a thick grove of trees will have a similar effect of Fresnel distortion. As you have a plenty strong -65 signal level, intentionally misaligning the antennas won't hurt that much; but may get a more clean of path.

I have University of Texas and U.S. Dept. of Commerce citations if you'd like...

...for anyone that might think I don't know what I'm talking about.

BTW...
http://whtonline.com/Team_Ubiquiti/5.8_tree_blockage1.jpg
http://whtonline.com/Team_Ubiquiti/5.8_tree_blockage2.jpg
http://whtonline.com/Team_Ubiquiti/5.8_tree_blockage3.jpg

[rconaway]

Shapiros says you are good so at least you have one fan. However, in this case you are right. The water content in the tree wigs out the Fresnel zone.

[jlewisfl]

and if you live where it rains frequently (like I do), you can typically just turn everything off until it gets at least somewhat dry again. Truly wet foliage, or heavy fog, heavy rain, can turn a usable path into a linear dummy load. I don't know whether circular polarization might help or not. You lose some total gain but it may be less blurred overall. Didn't look at those hyperlinks, so dunno if they investigated that or not. Left-or-right-hand circular polarization is usually used for satellite up/downlinks in relatively LEO sats b/c of faraday rotation. Antenna hardware and phasing harnesses are more complex than simple linear polarization antennae. And cost more, of course, per dB of gain.

[WHT]

Shapiros says you are good so at least you have one fan. However, in this case you are right. The water content in the tree wigs out the Fresnel zone.
I worked out a model for tree grazing, and you know and seen how I tend to get very elaborate at times in my dissertations...to the point you need a dictionary.

I couldn't see anything that resembled knife-edge Fresnel distortion, rather a multi-point scatter seem to be the right fit. Nevertheless, it would have the same effect as a crappy Fresnel clearance, but with a different process.

[rconaway]

Does that scatter cancel out the signal by changing phase?

[WHT]

Yup, even a ripple tank simulator showed that.

[ljenkins]

WHT always makes monday morning a little more bearable.

[shapiros]

I worked out a model for tree grazing

Doesn't that refer to the eating habits of giraffe's?

Sig

[WHT]

Donne make me pwn yer azz........
$ rm -r shapiros?.*

[shapiros]

harsh...

At least it was not rm -rf....

[jlewisfl]

For Mike - Any idea at this point why the noise handling button is not showing on Advanced (or anywhere else)??

[UBNT-Mike.Ford]

For Mike - Any idea at this point why the noise handling button is not showing on Advanced (or anywhere else)??


Hello,

I am awaiting an answer from my software guys.
Thanks,

[rasa]

For Mike - Any idea at this point why the noise handling button is not showing on Advanced (or anywhere else)??

Hi,
Noise Immunity is switched OFF by default as it fixes connection issues on low signals. We just left ability to turn this ON in configuration. This option was not planned to be on WEB UI but we''ll think about this later. As release notes states it is available only in config:
"Ability to switch Noise immunity on/off in config, by default off "
Next time we'll be more accurate in release notes, sorry!

[jlewisfl]

OK, so if I download the cfg it will be there as a switch, but I want to leave it OFF if I am running weak signals, right? It was ON in the original release, which is why 10 MHz and 5MHz BW and channel shifting would not work?
(In other words, leave it as is, or "if it ain't broke, don't fix it!!")

Joe

[UBNT-Mike.Ford]

OK, so if I download the cfg it will be there as a switch, but I want to leave it OFF if I am running weak signals, right? It was ON in the original release, which is why 10 MHz and 5MHz BW and channel shifting would not work?
(In other words, leave it as is, or "if it ain't broke, don't fix it!!")

Joe


Correct.

Thanks,

Mike