I have one question regarding the use of R52Hn cards in 802.11a/b/g modes (i.e. non-"n") with both chains enabled (and two different antennas, of course). This is something relatively new to me for this class of radio devices as, if I understood it correctly, it allows interesting applications that goes beyond "traditional" diversity. This matter was someway already discussed in different threads of this forum without clear answers (and even with some "light" flame), so let me start again from the beginning. Moreover I have a specific question that was not discussed yet.

Just one kind request: be on topic and do not hesitate to correct me if you can prove I'm wrong. Avoid flames and technical non-sense. Thank you.

Connecting more than one antenna to a single radio card is usually called "antenna diversity". In "traditional" diversity (as implemented in previous generation of "a/b/g"-protocol radios NOT supporting the "n"-protocol, such as the R52/R52H) we have a single "rf-chain" (RX/ADC/DAC/TX) that can be switched between two different antennas. By the way RouterOS never implemented antenna diversity in "dynamic" way: we can just select which antenna connector to use (antenna-a/main, antenna-b/aux) so that in our typical outdoor setup we connect a single antenna to one of the available connectors, select it into the "wireless interface" settings (i.e. "antenna-mode=ant-a") and forget about the other one. To the best of my knowledge there is no way to use both connectors of a R52/R52H card to connect two different antennas so to use them both at the same time (apart of the unusual configuration where one antenna is used only for TX and the other one only for RX, which could make sense only in very particular setups): please, correct me if I'm wrong.

With "n"-capable cards, such as the R52Hn I'm referring to, we are in a completely different situation: there are 2 different and independent rf-chains that CAN be used at the same time by connecting 2 different antennas; in "n"-mode we know that this gives us the 2 different "data spatial streams" used in MSC8..MCS15.

What if the card is used with both chains enabled in "a/b/g" modes? As already pointed out in another thread of this forum, both antennas will transmit/receive the same informations at the same time (in "a/b/g" modes there is a single "data stream"): assuming that this is true (please, confirm) this can be used in different ways:

1) we can use two antennas with the same "footprint" (i.e. two omnis or two directional/sectorial antennas aimed at the same direction) in order to double our EIRP/sensitivity and, optionally, use different polarizations so to improve link quality against multipath and similar problems;

2) we can use two directional/sectorial antennas aimed at different directions so to solve particular coverage problems by using a single radio and a single RF channel (i.e. we build an AP with a single radio and two antennas: a sectorial-one used by near-by users and an high-gain grid aimed at a remote station located several kilometers away).

Assuming that all of the above is correct (again, please, confirm) I want to discuss the second option above: two directional antennas aimed at completely different directions. I'm reasonably confident that such configuration would work correctly with any "coordinated" protocol (i.e. nstreme or nv2) but what about traditional 802.11 CSMA/CA protocol (assuming that RTS/CTS protection is NOT used)? In other words: will the above setup be affected by the "hidden node" problem?

Let me better define the problem. Suppose that I have an AP located on the top of a high wall with two directional antennas connected to the same R52Hn with both chains enabled; the antennas are aimed to each side of the wall and the card is operating in standard 802.11g CSMA/CA (no nstreme, no nv2). There are 2 stations (LEFT and RIGHT), at ground level, one on each side of the wall. The stations do NOT support RTS/CTS protection (the AP might, at most, be configured to issue a "CTS-to-self"). Station LEFT does NOT hear transmission of station RIGHT, and vice-versa; both stations hear the AP from the corresponding antenna and the AP hears each station ONLY from the antenna aimed at it.

This might seem to be the classical "hidden node" problem, but actually isn't. Suppose that LEFT and RIGHT start transmitting at the same time: given the premises there will NOT be any "on-air" collision. Each antenna of the AP will receive ONLY the transmission of the corresponding station and each chain into the card will correctly decode such transmission: no "scrambling collision" up to this point... but what will happen after?

The answer is in the way the network controller works, and this is quite impossible to understand without a deep knowledge of the inner structure of the AR9220 chipset (by the way: is there any way to obtain Atheros' datasheets?)

I see only 3 possibilities:

a) a sort of "digital collision" happens "inside" the NIC chip and both transmissions are discarded
b) one of the transmissions will be correctly processed by NIC chip and the other one will be discarded
c) a sort of buffering is performed and both transmissions will be correctly processed by the NIC chip

Is there anybody in here able to answer?

I hope to have clearly explained this point as it appears quite interesting to me.

Thank you for your patience in reading.

Hi, Firstly you need to break apart the questions in here, you ask a question then answer it 3/4 so there is a few things, having 2 antenna's ruining does work on HN

We have run a link with one H-Pol and one V-Pol (with a virtual ap) and that worked just fine, Hidden node problem you will find with most out door ap's as even 2 grids that are on 2 properties next to each other will not be able to "hear" each other, as to the technicalities of all of this, cant give u much answers, testing in your environment is prob the best, both the links were running N and both got 135 on HT-long @ -68 and -65 respectively

CyberT,
your post is exactly the type of post I would not have read.

I did not ask question just to answer myself: the first part of my (may be too long, I agree) post was just a summary of information based on my knowledge and understanding of how "11N" cards differs from previous generations, given that... (quoting myself):

This matter was someway already discussed in different threads of this forum without clear answers (and even with some "light" flame), so let me start again from the beginning.

The aim was to create a common background and starting point based on what I (think to) know and DO NOT want to discuss (unless somebody finds something wrong in it) so to point out the matter I'm really interested in, which is the way "11N" cards work in "non-N" modes (as it was very clear even from topic's subject).

Nevertheless you are just giving me some info about your experience in "N" mode, which is exactly one of the things I'm NOT interested in.

Let me repeat again: I would like to discuss about "11N" cards working in "a/b/g" modes with both chains enabled, and proposed a particular configuration which, in my opinion, clearly shows the specific critical aspect I'm interested in.

You didn't say anything about such point.

Regards,
rock

I see only 3 possibilities:

a) a sort of "digital collision" happens "inside" the NIC chip and both transmissions are discarded
b) one of the transmissions will be correctly processed by NIC chip and the other one will be discarded
c) a sort of buffering is performed and both transmissions will be correctly processed by the NIC chip

Is there anybody in here able to answer?

I hope to have clearly explained this point as it appears quite interesting to me.

Thank you for your patience in reading.

Few points:

Transmit diversity feature would cause a over the air collision.
Your setup would have lower sensitivity and -3db power.
I don't think there would be a type of collision -inside of the nic -that you describe, after all the chipset is designed to process two chains at the same time.

Hi rmichael, and thank you for answering.

Some comments:

Transmit diversity feature would cause a over the air collision.

In my hypothetical configuration no "over-the-air" collision is possible as the coverage areas are totally insulated: each station does not hear the other one. Yet it is still possible that both one station and the AP start transmitting at the same time, but this is something that is inherent in 802.11-CSMA/CA, or any other un-coordinated shared-medium protocol: this is unavoidable and that's not the kind of collision I'm interested in (which is collision between different stations). Moreover this has nothing to do with diversity which, by the way, in the configuration above does not exists: in each of the insulated areas above (at each side of the wall) there is a SINGLE AP's transmitting antenna (no diversity).

Your setup would have lower sensitivity and -3db power.

Lower than what? Lower than the power/sensitivity declared by the cards manufacturer? Well, it depends on what the manufacturer declares. This is a completely unrelated point (that, may be, deserves a specific topic, more of "commercial" than technical matter). If your reference values are the "combined power/sensitivity" (i.e. both chains working) then yes, you have -3dB for both power/sensitivity in each of the insulated areas above. If your values are related to each single chain you have exactly them (and if you use them both into the same coverage area you gain +3dB in both EIRP/sensitivity, as I already said in my example application n.1)

I don't think there would be a type of collision -inside of the nic -that you describe, after all the chipset is designed to process two chains at the same time.

THIS IS THE POINT. The problem here is to understand HOW the NIC process the bits decoded by each RX chain. If both rx antennas were into the same coverage area they would receive a signal related to the same data transmission (roughly speaking: the same data frame), any difference being due to multipath propagation only: so it might do its job of "correlating" them in order to better reconstruct the original signal, and I think that actually this is the way the "combined" rx sensitivity improves. But what if, as in my hypothetical scenario, the bits decoded by each RX chain are totally unrelated (roughly speaking: two different data frames from two different stations)? Any attempt of "correlation" would bring a totally unmeaningful result, that's what I described as a "digital collision inside the NIC". And I'm inclined to think that this is what actually happens.

I don't think that the NIC was designed to handle two totally unrelated data streams from each chain in "a/b/g" modes (nor in "n" mode with a single spatial stream, e.g. with only MCS0..MSC7 used), but I would really be wrong!

Regards,
rock