Forum

  • With both Ruckus and Cisco beam forming technologies hidden STA (remember I differentiate between them) problem can exist and I fully divulge that during the road show presentation.

    However, properly designed, hidden STA problem won't exist. Many people are under the impression that beam forming is to get better range. Nope. It can be used for that, but that isn't its main advantage.

    The main advantage of beam forming is to increase the downstream data rate to the STA. For example, let's say you have an AP in place that has a Tx power of 30 mW. Now you deploy your STAs and they can all connect to the AP at varying data rates. Lets assume that there is no hidden STA problem. Now, exchange that AP with a Ruckus AP. Now, the hidden STA problem still doesn't exist, however, it did create another problem. What it did do is significantly increase the downstream data rate to each STA.

    The problem that it did cause is that when the Ruckus AP is transmitting, a STA on the other side of the beam may not hear the AP and may Tx over the AP. This doesn't cause a collision for the receiving STA, but it does cause the AP to not hear that Tx and of course not ACK. Both I and separately Ruckus and Cisco cannot deny that this can happen.

    The point that I have been trying to make is that the AP is transmitting so much more often than any one STA and the data rate is so much higher that this problem is not nearly as bad as many perceive.

    I'm not a marketing guy for Ruckus technology but there is proof in testing that it works and works well. It isn't without flaws, but it is quite easy to poke holes in every vendors special technology. I encourage you to get your hands on some Ruckus, Cisco and Xirrus gear to test beam forming.

    GT

    P.S. I was just thinking during this post: What if Ruckus borrow a Meru trick and sent a high duration CTS frame then sent a burst of beam formed frames. Hmm...

  • Actually you are lucky because I had the opportunity to see a test last week on 7962, if I remember well.
    1st test: 5 PC running bi-directional transfers => total aggregated throughput 148Mbps
    2nd test 10 PC running bi-directional transfers => total aggregated throughput 54Mbps. Weird no ?
    If you looked at AirPcap capture you can see a huge amount of FCS errors...

    You can test it yourself that's really basic, we are not even talking here about an enterprise setup, just a single AP imagine now the same test with 6 AP using 1, 6, 11 ...

    For your information please read their sponsored test report http://www.jpasoc.com/Documents/JPA-Test-Report-010808.pdf see figure 4 (3 APs) and figure 5 (5 AP) and look at clients placement, don't you see anything wrong ? This comes from Ruckus ;)

    As regards to Meru trick I actually wasn't aware about this one, so could be a solution but I don't think it's that easy to implement or all the vendors would use it.

    Anyway interesting topic.

  • Grant said:

    ???¡é?¡é?????¡­?¡°Here is another point that is often missed. Let's say you are on a VoIP call. How often do you think it is transmitting a data packet? Most people are under the impression that it is transmitting a frame every time it gets the chance and it just ain't so.???¡é?¡é????????

    Grant has brought up a very interesting and not-so- well-knownfact here. There are a number of causes for this, but one of them is in relation to the physical mechanics of how we, as humans, actually talk.

    First, let???¡é?¡é?????¡é???¡és imagine that we are trying to send a very long file of numbers [ say ] over the Wi-Fi system. Whenever the Tx STA gets the chance, it will try to send some data. After all, we have a whole bunch of bits and bytes all laid out nicely in a transmit buffer within the STA. Many more are waiting in another memory area for their turn to be transmitted. Each time the buffer empties, more data is ???¡é?¡é?????¡­?¡°pumped???¡é?¡é?????¡é???¡é into it from memory under control of the system microprocessor/ASIC units. In other words, at very possible Tx opportunity, that STA will try it???¡é?¡é?????¡é???¡és best to just ???¡é?¡é?????¡­?¡°keep on pumping???¡é?¡é???????? until such time as our very long file has been fully transmitted.

    Now onto humans. When someone says to you ???¡é?¡é?????¡­?¡°That woman [ sorry girls ] talks non-stop for hours???¡é?¡é???????? we have to look at things from a human???¡é?¡é?????¡é???¡és perspective and then from the STA???¡é?¡é?????¡é???¡és perspective. To our ears, we do indeed ???¡é?¡é?????¡­?¡°feel???¡é?¡é???????? that the chatty woman is actually speaking non-stop. However, if we were to look at her voice on an oscilloscope [ showing voltage versus time ] as output from a loudspeaker, we would see three important things. Let???¡é?¡é?????¡é???¡és take the expression ???¡é?¡é?????¡­?¡°Did you see what Blake on Day???¡é?¡é?????¡é???¡és of Our Lives did yesterday ????¡é?¡é???????? [ I know I???¡é?¡é?????¡é???¡ém going to get irrate phone calls over this one???¡é?¡é???????|]. To our ears, all that we hear is ???¡é?¡é?????¡­?¡°DidyouseewhatBlakeonDay???¡é?¡é?????¡é???¡ésofOurLivesdidyesterday ????¡é?¡é???????? Everything seems to be run together. However, no matter how fast she talks, on an oscilloscope we would see three things:



    1. When she says ???¡é?¡é?????¡­?¡°Did???¡é?¡é???????? that all sounds like one block, but it is not. When we wish to speak, our brain sends a signal to our vocal chords. The vocal chords vibrate and sound is emitted from the mouth. The vocal chords ???¡é?¡é????????expand???¡é?¡é???????? and ???¡é?¡é?????¡­?¡°relax???¡é?¡é????????. That takes physical time. To our ears it seems like no time at all, but to an 802.11 STA transmitting at megabits per second, it is a RELATIVELY long time. There are gaps between ???¡é?¡é?????¡­?¡°D???¡é?¡é???????????¡é?¡é???????|.???¡é?¡é????????i???¡é?¡é???????????¡é?¡é???????|???¡é?¡é???????|???¡é?¡é????????d???¡é?¡é????????. It is these gaps that are important.

    2. When she says the next word ???¡é?¡é?????¡­?¡°you???¡é?¡é???????? we may feel that the ???¡é?¡é?????¡­?¡°Did???¡é?¡é???????? and ???¡é?¡é?????¡­?¡°you???¡é?¡é???????? are all run together, but no matter how it seems to our ears, her vocal chords stopped working for a period of time, leaving another gap between the two words.

    3. There is a gap at the end of her sentence whilst she waits for a response [ Did you actually see what Blake did ? ].

    This phenemonon of gaps between letters and gaps between words and gaps whilst waiting for a response has actually been known for several decades. A system called Digital Circuit Multiplication Equipment actually uses this. It waits for these little gaps and then inserts a part of someone elses converstaion in your gap to minimize dead time. Very sophisticated control equipment keeps tabs of whose voice ???¡é?¡é?????¡­?¡°bits???¡é?¡é???????? are where. Then at the other end, it all gets ???¡é?¡é?????¡­?¡°un-interleaved???¡é?¡é????????.
    http://www.oriontelecom.com/voice_compression/voice_compression.htm

    So, even though we think that words are coming out of our mouths at high rate, to the STA, not so much. Unlike DCME [ a multiplexed system ] when there is a gap, the STA simply does not transmit as there is no data there, unlike the data file example, where it will try to transmit at every opportunity it can until that file is "all gone". I???¡é?¡é?????¡é???¡éll put a posting up some other time about why a sampling rate of 8000 samples per second was originally chosen for digital conversion of voice [ sampling, quantization and encoding ]. It has to do with the physiology and psychology of how we actually hear.

    Dave

  • Dave thx for your input. You are right if you talk about VOIP or low demanding data service. But this is far to be the case in an enterprise deployment. You can have Data, VOIP, Video on the same wireless LAN etc???¡é?¡é???????|
    Data can be many things: ftp uploads/downloads, Per2Per etc???¡é?¡é???????| I think nobody will argue against the fact that with 1 single PC running multiple downloads or uploads streams, you can easily use the full bandwidth of an AP (try with chariot or iperf).
    Let???¡é?¡é?????¡é???¡és take the example of an 11g client, let???¡é?¡é?????¡é???¡és say it doesn???¡é?¡é?????¡é???¡ét have to fight against any other client and its backoff time is always minimum = 135 ????????s then its data rate will be 20,5Mbps then the client is using the FULL AP bandwidth and send its packets the ones after the others there is no gap between them. Does it make sense?

  • Yes, when we have a "mix" of data and voice, the situation becomes much more complex.

    Dave

  • One of the interesting things about the human ear and speech is that the ear and brain are fairly tolerant of errors [ to a point ], but much less tolerant of delay and ???¡é?¡é?????¡­?¡°jitter???¡é?¡é????????.

    For example, let???¡é?¡é?????¡é???¡és say your friend is calling you on the phone and has said ???¡é?¡é?????¡­?¡°The Pittsburgh Steelers really played well tonight and may win the championship???¡é?¡é????????.

    Now along comes a bit of interference. ???¡é?¡é?????¡­?¡°zzzzzzzzz???¡é?¡é???????? . What you actually hear is:
    ???¡é?¡é?????¡­?¡°The PittZZZurgh Steelers really played well tonight and may win the chamZZonship???¡é?¡é????????.Now cultural issues come into play here, but it is fair to say that most folks would ???¡é?¡é?????¡­?¡°read between the lines???¡é?¡é???????? and understand what the other guy had said.

    However, if we actually heard [ with more interference ]: ???¡é?¡é?????¡­?¡°The PittZZZZZZ StZZZZrs really plZZZZ well toZZZZand may ZZZZZZZ chZZZZZZZZZp???¡é?¡é????????, chances are that we would ask them to repeat the whole thing.

    In fact this is what was done originally by British Telecom and ATT to determine an ???¡é?¡é?????¡­?¡°acceptable???¡é?¡é???????? bit error rate for digital voice communications. Large groups of volunteers were sat down and had to read pre-prepared scripts [ actually chapters from a book ]. People listened at the other end of the line and engineers gradually introduced more and more interference [ degrading the bit error rate or BER ]. Eventually the point was reached where most people said ???¡é?¡é?????¡­?¡°OK this is starting to get annoying???¡é?¡é????????. A similar thing is done in the development of MOS [ Mean Opinion Scores ] for VOIP encoders [ CELP, A-ACELP etc ]. It is also cultural. Some cultures ???¡é?¡é?????¡­?¡°tolerate???¡é?¡é???????? errors more than others.

    Jitter is really what gets the old brain annoyed. Imagine someone says ???¡é?¡é?????¡­?¡°Hello, how are you ????¡é?¡é????????. When the first geostationary satellites were launched, all the naysayers said ???¡é?¡é?????¡­?¡°It???¡é?¡é?????¡é???¡éll never work ???¡é?¡é?????¡é?€?? people will not tolerate the quarter second delay between you speaking and hearing the other guy reply???¡é?¡é????????. They were very surprised to find that people did adapt quickly to the delay, but the one thing they did not adapt to well was jitter. Jitter refers to the variation in arrival time of bits. If we hear ???¡é?¡é?????¡­?¡°H???¡é?¡é???????|..e???¡é?¡é???????|???¡é?¡é???????|l???¡é?¡é???????|???¡é?¡é???????|l???¡é?¡é???????|..o???¡é?¡é???????|howare???¡é?¡é???????|???¡é?¡é???????|???¡é?¡é???????|..y???¡é?¡é???????|???¡é?¡é???????|???¡é?¡é???????|???¡é?¡é???????|..ou ????¡é?¡é???????? This becomes extremely annoying to humans. It???¡é?¡é?????¡é???¡és not so much that we cannot understand what they are saying, it???¡é?¡é?????¡é???¡és that the ???¡é?¡é?????¡­?¡°slowing down???¡é?¡é???????? and ???¡é?¡é?????¡­?¡°speeding up???¡é?¡é???????? of the words gets us ???¡é?¡é?????¡­?¡°ticked off???¡é?¡é????????. In fact nearly all voice design takes into account the psychology of the brain when hearing.

    This is why we have to be so careful with many things in Wi-Fi related to VOIP e.g. DTIM intervals and PCO operation in 802.11n where we ???¡é?¡é?????¡­?¡°jump???¡é?¡é???????? between 20 Mhz and 40 Mhz operation. These can cause a lot of jitter, which we as humans just do not like, not to mention the built-in jitter tolerance of the hardware.


    Dave

  • Here???¡é?¡é?????¡é???¡és a really nice little overview of the Ruckus Dynamic beamforming system. As Marcus said, it has two main components: the actual antenna system itself as well as some very, very clever software.
    http://www.ruckuswireless.com/technology/beamflex

    Where did all this beamforming business come from ? Like many ???¡é?¡é?????¡­?¡°new techologies???¡é?¡é???????? it is based on something rather old.

    During The Cold War, the US was concerned that Russian missiles would be sent over the North pole to the U.S. Massive radars were built using ???¡é?¡é?????¡­?¡°over the horizon???¡é?¡é???????? scanning. The problem was that these huge antennas had to be mechanically driven via electric motors and had very complex control systems. These motors were prone to failure, and if this happened, those big antennas would not be able to move, and would potentially be unable to track incoming ICBM???¡é?¡é?????¡é???¡és. Maintenance was a nightmare as well. The US Air Force spearheaded development of what was called Phased Array Antenna Systems. In this case, instead of physically moving the antenna, you could ???¡é?¡é?????¡­?¡°shape???¡é?¡é???????? the beam to electronically steer to whatever direction you wanted. This system has served the nation well for many years. 802.11n beamforming technology uses this technology at it???¡é?¡é?????¡é???¡és heart.
    http://www.afspc.af.mil/library/factsheets/factsheet.asp?id=3656


    Dave

  • Roma59 Escribi?3:

    Dave thx for your input. You are right if you talk about VOIP or low demanding data service. But this is far to be the case in an enterprise deployment. You can have Data, VOIP, Video on the same wireless LAN etc???¡é?¡é???????|
    Data can be many things: ftp uploads/downloads, Per2Per etc???¡é?¡é???????| I think nobody will argue against the fact that with 1 single PC running multiple downloads or uploads streams, you can easily use the full bandwidth of an AP (try with chariot or iperf).
    Let???¡é?¡é?????¡é???¡és take the example of an 11g client, let???¡é?¡é?????¡é???¡és say it doesn???¡é?¡é?????¡é???¡ét have to fight against any other client and its backoff time is always minimum = 135 ????????s then its data rate will be 20,5Mbps then the client is using the FULL AP bandwidth and send its packets the ones after the others there is no gap between them. Does it make sense?


    I'm getting a bit confused as to where this thread is going. My original post dealt with direction of traffic and how improving data rates with beam forming will improve downstream data rates.

    We've all posted quite a lot and I want to sum up my points.

    - In the bulk of enterprise deployments, downstream traffic represents significantly more than 50% of the traffic. I still contend that on a byte by byte basis, it is still more like 75+% downstream.

    - Beam forming improves signals strength to STA's thereby improving data rate.

    - Correctly deployed, true hidden STA problem won't exist.

    - It will occur that the AP will be transmitting a beam to a STA and another STA won't hear that, thereby causing collisions. However, I contend that with 75+% of traffic heading downstream that this problem does exist, but the benefits of improved data rate far outweigh this issue.

    Roma, Marcus, Dave, others - I'd like to hear a true / false for these statements. I do agree that there are environments, mostly dealing with high STA upload, that my points don't fare as well. But, look at it from the "normal" enterprise if you would be so kind.

    For further reading on the advantages of beam forming, take a look at this article from Tom's Hardware.

    http://www.tomshardware.com/reviews/beamforming-wifi-ruckus,2390.html

    GT


  • For further reading on the advantages of beam forming, take a look at this article from Tom's Hardware.


    Come GT, not you. Let's be serious, even the less expert guys in CWNP forums is a Guru compared to tomshardware journalist. He is supposed to test the solution in an enterprise deployment but he uses only 1 AP and a couple of clients. What a joke. This just a cut and paste of a Ruckus white paper.

    Frankly I think we should stop with the assumptions and guessings, let's go back to reality take an AP, 10/15 clients that you will position in different locations, not as same as in the JPA report (nobody commented on this one see erlier post) and you'll see that the impact is not small. I saw the same kind of test last week, very basic no tricky stuff add as I said in an earlier post (nobody commented on this one) the aggregated throughput was 54Mbps... we can argue for hours but reality is reality.


    - In the bulk of enterprise deployments, downstream traffic represents significantly more than 50% of the traffic. I still contend that on a byte by byte basis, it is still more like 75+% downstream.
    ok let's assume you're rigth

    Beam forming improves signals strength to STA's thereby improving data rate.
    I'm not an antenna guy but I think it's more a question of SNR

    - Correctly deployed, true hidden STA problem won't exist.

    Disagree, real tests with multiple client and Ruckus JPA report shows the opposite

    - It will occur that the AP will be transmitting a beam to a STA and another STA won't hear that, thereby causing collisions. However, I contend that with 75+% of traffic heading downstream that this problem does exist, but the benefits of improved data rate far outweigh this issue

    Disagree, not true in enterprise high density AP / clients deployments

  • Quote:
    Beam forming improves signals strength to STA's thereby improving data rate.

    I'm not an antenna guy but I think it's more a question of SNR


    For this one, provided that the ambient noise environment has not changed during the actual beamforming process, an increase in signal strength will automatically cause an increase in signal to noise ratio.

    As Grant was saying, the main reason for using beam forming is not so much for range, but to improve data rates.

    As the signal level increases [ due to the "focusing" of the beam patterns ], the SNR automatically increases. As the SNR increases, a parameter called the Eb/No increases [ Energy per bit to noise density ratio ].

    With an increased Eb/No, we can do one of two things:

    1. We can "make better" the bit error rate for a particular data rate [ in other words same data rate as before, but improved BER ].

    2. We can increase the data rate for a fixed BER [ in other words, we have kept the BER the same as before when no beam forming was used, but have "boosted" the data rate ].

    Usually we have the latter case.

    Dave

Page 2 of 4