I am using RB433ah's with a maximum of 2 R52h cards for PTMP

Here is my problem / question.

When I set card 1 to 5760 @ 10mhz channels spacing... why when I scan with the other card do I see that card using 5760 and 5800??

When I set card 2 to 5810 10Mhz spacing and I scan with card 1 it see's card 2 @ 5810 and 5770....?? I don't understand... is this normal... I don't know why I have never caught this before...... I am using NSTREM

This has been a long standing issue that drives me insane.. On 10mhz you get 'ghost' ap's at 40 mhz above/below the real frequency. When using 5mhz channels you get the ghost ap's at 20mhz above and below.. It also happens on 2.4 and 5.8 and doesn't have anything to do with nstream..

One of the craziest things about it is that clients will actually connect to that 'ghost' frequency if they have a strong enough signal..

I haven't tested lately to see if a non mikrotik device acting as an ap shows the same ghosts or not.. I do know that a non mikrotik cpe will NOT see the ghosts from a Mikrotik AP..

-Gerard

Hi,
i can confirm this issue, with 10 MHz channel you have +/- 40 MHz ghosts with a signal level about 40 dB lower than the original.
I think it's a Atheros/HAL bug because you can see this problem even with NS5/PS5/etc.

Regards,
Giorgio

I wish they would correct this problem... we probably create more self interference than we even know.

Thanks for the explanation.

I witnessed this strange phenomena today with one of our 5ghz back hauls. We use XR5 cards, with nstream enabled, and ROS 3.17. We tried using a 5mhz channel width and after a few days of stable operation the link when down. When we checked out the client side of the back haul, we found that it was seeing 3 duplicate SSID's for the same AP spaced 20mhz apart. The designated center channel would be strongest, and the 2 duplicates would have about a 30dB lower signal. We moved back to a 20mhz channel width and everything has been working fine again.

We use XR5 cards, with nstream enabled, and ROS 3.17. We tried using a 5mhz channel width and after a few days of stable operation the link when down. When we checked out the client side of the back haul, we found that it was seeing 3 duplicate SSID's for the same AP spaced 20mhz apart.

When it went down, did you do need to power cycle it or were you able to login without rebooting it?
I ask because we are seeing a similar problem (total lock up) on 40 MHz channels with XR5 + 3.17 + Nsteme.

it is Atheros problem and we can't fix it. Those ghost frequencies should cause any problems as they are with lower signal - clients always connect to the AP with the highest signal.
About those lockup issue, we need more info, support output file from the AP and the client at the time when you see this problem.

About those lockup issue, we need more info, support output file from the AP and the client at the time when you see this problem.

How does one obtain a "support output file" from a device which is totally frozen?

then make it after the reboot.

While I agree the client side SHOULD connect to the highest signal, I have have dozens of times where customers have called saying there internet 'feel' slow, and sure enough when I check there unit it will be connected to one of the ghost frequencies. It is rare but when you have hundreds of clients you do run into it. I've had this happen with CPE having a normal -40's getting -60's on the ghost, and with clients having normal -69's getting -89's on the ghost, forcefully disconnecting them from the AP a few times typically resolves this problem, BUT catching this can be difficult. This issue causes so many other problems on the AP as even if the CPE still registers a strong signal the amount of hardware retransmits kills the AP. It appears to occur when the RouterOS CPE scans the band and connects to a 'good' signal without checking all the frequencies.

I am sure this is possible but I do know if Mikrotik would be willing to implement a freature that RouterOS based CPE's and AP's trade info regarding the connection (they already do this for Tx signal and the Rx signal). It would be handy to have an option where a RouterOS based CPEs could recieve the APs set frequency and compare, if different then disconnent and reconnect.

MyThoughts, I have the same problems on my network. I've often wished they could/would implement something like that too.. I have to limit the scan list on some problem customers and it makes changing frequencies on the ap a nightmare.

uldis, I really think the ghost frequencies actually do cause additional interference to adjacent ap's.. Try setting up 2 ap's with 5mhz channels. one on 2412, and one on 2452. Take a third device and sniff on channel 2432.. Look at how much data you receive from the other 2 ap's..

I know the ghost channels aren't physically there on the rf side of things. But I don't think it matters since the radio is still seeing them. Since it sees them it's still acting as noise/interference..

Am I right or am I just crazy?

-Gerard

I forgot to answer RK's question. No, the slave/client unit didn't lock up. After rolling up onsite and using my laptop/winbox to access the Mikrotik box, I found out the ghost channel issues. Only after changing to a 20mhz channel width did the issue go away. I would have done more investigating but the site had already been down for almost 24 hours, and I had clients screaming at me, so I just fixed it and left and then later found this thread in the forum.

We run quite a few RB333+XR2, 2.4ghz AP's utilizing 5/10mhz channel widths with no ghost channel issues. All of the clients are Tranzeo or Ubiquiti CPE's however. We've only experienced this issue this one time with ROS 3.16/17 + RB333 + XR5.

A band past filter should cure the problem if it is in fact a true frequency rather then just a ghost reading. I have not tested this but some else said they had gotten data through the ghost link, has any one run a spectrum analyzer on the cards?. My guess is that i may be due to the way the filtering works on the cards. All radios have harmonicas at lower levels then the main signal whether they are a problem depends on the the design and filtering. A good filter will not only eliminate ghost issues it will improve performance of adjacent aps and and provide for some measure of protection from competitive gear that is close in frequency.

At the 2007 MUM in Orlando, I brought this problem to the attention of Mikrotik. Never got a good answer from them so I asked Mike from Ubiquiti about it. Mike, myself, and some other people ended up borrowing a spectrum analyzer from Carl from Streakwave to see if the signal was actually there.. It was not. I have since reconfirmed that the ghost channels are not there on a spectrum analyzer. The ghosts are only visible in software.

I wish I could solve the problem by using bandpass filters, unfortunately the signal isn't really there so it can't be blocked by hardware.

I can put an AP on 2437 with 5mhz channels. put a client on 2457 or 2417 and it will connect and pass traffic just fine while being connected to the wrong frequency. I have also had AP's running in AP bridge mode with WDS enabled and had them connect to each other on when they are on the wrong frequencies.

This problem has been happening ever since I started using 5mhz channels on the early 2.9.x versions..

I'm going to fire up a ubnt cpe/ap tomorrow and see if it has the same ghosting problem. I remember trying this a while back and I don't think I had ghosts with the ubnt cpe's..

-Gerard

Of course the signal is actually there, when you set a device to a frequency at is has 'X' output power that power doesn't go to 0 as soon as you leave that spectrum width of that device.

The reason you 'seeing' the ghost frequencies could be do to an Atheros issue, or it could how it is handled in software but it does not really matter the signal is indeed there.

I believe is my memory serves me correctly the standards for 2.4 GHz in North America (generalizing, I am aware each country has its own rules) call for a decrease of 1 dB in signal for every 1 MHz you move away from the frequency in use. This is why you see 'ghost' frequencies that are attenuated along those lines.

Either way I do believe that even if it is infact an Atheros issue it can be corrected via software, and should be as it degrades the percieved performance of said product, ie. RouterOS.

I truly hope this issue is taken seriously as I believe it to be a major issue, and as I get more and more customers this problem is showing up more often.

Cheers

I tested with a NS2 today and it does not see the ghost ap's when running in station mode. However when the NS2 is in AP mode the Mikrotik sees ghosts from the NS2.

I also had the spectrum analyzer going while I was doing my testing..

I did NOT see the ghost frequencies on the specan. I have a crappy specan that only goes down to a -100ish noise floor and I didn't see them at the frequencies reported by the Mikrotik..

Since both the UBNT NS2 and the Mikrotik (R52H in this case) are both Atheros chipset cards that leads me to believe there is a software or driver problem in the Mikrotik that is causing the ghost channels to show up.

-Gerard

Any development of the Mikrotik driver...

We still have the problem with MT station in 4.4 ...
Station don't connect to the strongest signal. It seems the result of the scan is:
xx:xx:xx:xx:xx:xx SOMESSID 2412 signal: -80
xx:xx:xx:xx:xx:xx SOMESSID 2432 signal: -60

and since the mac are the same it ignore the second result...

Any idea from the MT support,

Michael