Blogs

11ac Beamforming Makes the Dog Rollover

By neela posted May 16, 2014 08:10 PM

  

The adoption rate of 802.11ac is outpacing 802.11n. The performance gains in 802.11ac, fueled in part by the demands of #GenMobile, are contributing to its rapid adoption.

 

802.11ac boasts theoretical gigabit speeds due to wider 80-MHz channels, faster 256-QAM and everyone’s favorite…beamforming. The first two are easy peasy, but beamforming, that’s pretty heady stuff.

 

At a high-level, beamforming is about providing a stronger signal to clients – more bars – to improve the client’s connection rate. While the concept of boosting client signal strength isn’t new, the industry finally standardized it in Wave 1 of 802.11ac and called it 11ac beamforming.

 

Aruba’s standards-based beamforming implementation involves a constant exchange of information between the client and AP to maximize the client receive signal. This is critical for improving wireless performance in dynamic RF environments with highly mobile clients.

 

All Aruba 802.11ac APs support standards-beamforming, thanks to Broadcom. Just about every 802.11ac device, including the Samsung Galaxy S4, MacBook Pro and MacBook Air, support beamforming. So we actually have a solution that works on the network and the client side. Screen Shot 2014-05-16 at 1.39.58 PM.png

 

You’d think something as standard as beamforming would be supported by everyone, but that unfortunately is not true. Too bad a few vendors picked Atheros chipset for their 802.11ac AP, which lacks beamforming. So some wireless vendors support proprietary analog version of beamforming (I’m telling you, it’s not standard beamforming). 

 

Sadly, non-standard analog beamforming has some fundamental trade-offs that make it less effective in the real world. 

 

Here’s what I mean. First, analog beamforming can’t accurately determine a client’s location so you can’t flex a beam to that client. Second, it’s a manual hit-and-miss process that involves tweaking antennas to steer RF beams at clients. Third, too much beam, too little MIMO.

 

We compared Aruba’s 802.11ac beamforming against non-standard beamforming in a simple rate-vs.-range test. The other vendor's performance at 15 feet is equal to Aruba’s performance at 120 feet.

 

Standard 802.11ac beamforming wins hands down.

 

Screen Shot 2014-05-16 at 4.26.20 PM.png

6 comments
15 views

Comments

bosborne
Jun 27, 2014 08:06 AM

I do not mean to inject myself into this debate, but our experience here does nor agree with point 5 above.

 

I think the issue is this. For example, Cisco says most clients support their propeietary extensions because mose *Enterprise* clients support this.

 

While most enterprise clients may well be Broadcom, Aruba prides itself in supporting educational customers.

 

 

In our experience here at Liberty University, we are seeing plenty of Atheros clients, especially in the lower proced clients purchased by students. In  Higher Ed setting, most of the devices are student owned, not Enterprise level devices.

 

Ejohnson
May 27, 2014 04:24 PM

Let's take this 1 by 1

 

1) First, client support in Wi-Fi devices is pretty widespread. Broadcom is the dominant chipset provider for handsets and all of their 11ac chipsets provide the necessary support for 11ac beamforming. i.e. they will respond to sounding packets.

So pretty much all top end Android devices from most vendors are supporting this. In fact the Samsung GS5 recently announced is a 2x2 11ac device. 11ac adoption is very rapid and widespread in phones tablet MACs and PCs. For a detailed list please refer to:

http://www.wi-fi.org/certified-products-results?capabilities%5B71%5D=71&items=30&start=90&curpage=4

2) Please refer to:

http://www.wi-fi.org/certified-products-results?capabilities%5B71%5D=71&items=30&start=90&curpage=4

 

3) Let's talk about link balance then.

A 14 dBm smart phone will generally also have a receiver that is 5 dB worse than a single chain on an AP. Link balance is established when you account for the difference is tx AND rx. An AP is 5 dB better per chain on the rx and Aruba's indoor and outdoor APs have 3 chains so that give a net rx diffetence of 9.8 dB (5+10*log(3)). So an AP should be transmitting (conducted power) at least 24 dBm. Add 5 dB of antenna gain and the system is balanced when the AP is tranmitting 29 dBm EIRP (conducted + antenna gain). Add 3 dB for mrgin into the AP and you want your AP to be transmitting up to 32 dBm. That is what a 2 way street looks like in the 11ac world.

 

4) At long range, 11ac beamforming is even MORE important. Whereas the analog beamforming solution can attempt to maximise power at the client, 11ac baseband (digital) beamforming does more than that. Yes it increases signal strength at the client but MORE IMPORTANTLY, it also optimizes the transmission at the AP  to maximise the ability to support multiple streams. Given that we are now seeing the introduction of multistream handsets 11ac beamformings importance is multiplied.

 

5) the doiminant provider of 11ac chipsets in handsets is Broadcom.  And all of their devices support 11ac sounding frames. The market has spoken, 11ac sounding support for 11ac beamforming is now table stakes so why would you deploy a device that will not support 11ac beamforming.

 

6) Ruckus is an anlog beamformer. It creates beams by physically switching PIN diodes on and off or changing the feedpoing on a multifeed element. BTW 4096 is 2^12. they have 12 switching positions. Worst kind of marketcture i can conceive of.

11ac beamfoming is done in the digital part of the radio chip set. Amplitude and phase adjustments are done in the digital domain. each stream gets its own pattern AND since the propagation channel is different from the bottom to the top 11ac at its full extent is capable of generating a different antenna pattern for EACH subcarrier. Think about that, in an 80 MHz channel there are 242 tones x 3 streams. taken to its full extent that is 726 different antenna patterns. In reality they are grouped in 8s or 16s since the variation on adjacent tones is not much but that is stil almost 100 different antenna patterns being generated on EACH burst.

 

Ruckus builds up a position estimate over many bursts. 11ac does this with a SINGLE sounding packet. 11ac is several orders of magnitude more precise and rapid than iterating to a guess using traffic derived packets.

 

 

 

 

DevinAkin
May 23, 2014 10:37 PM

Hi Neela.

I own a VAR who is partnered with both Aruba and Ruckus, and I understand 11ac TxBF and Ruckus’s Beamflex technologies quite well. I would like to address some of the points you made in this post, and before I get into my points, note that I strongly disapprove of bogus marketing and technical marketing.

 

1. You said “support beamforming”.  That seems a little vague. The chipsets may have support for beamforming in them, but that doesn’t mean that the driver support exists for that device or that the Wi-Fi Alliance has compatibility tested it.

 

2. You said, “Just about every 802.11ac device, including the Samsung Galaxy S4, MacBook Pro and MacBook Air, support beam forming.” << In fact, a tiny fraction (I would speculate <5%) of today’s 11ac-enabled client devices support (both in their chipsets and drivers) 11ac TxBF. Even the devices you listed do not support it. The Samsung S4 can only be a beamformee for example (i.e. it cannot do TxBF itself, but rather can only receive beamformed data streams).

  • New 802.11ac-enabled MacBooks are not Wi-Fi certified, and thus aren’t certified for 802.11ac TxBF. http://www.wi-fi.org/content/search-page?keys=macbook
  • What about iPhones, iPads, and the myriad of Samsung 11ac enabled devices out there (Galaxy Tabs, Note 3, S5, etc.)? 11ac TxBF is nowhere near “ready for primetime” and probably isn’t worth much discussion here or otherwise.

3. It’s important to keep in mind that RF communication is a two-way street. It’s great if the AP can beamform downlink, but can the client beamform uplink transmissions? Typically not, due to drivers and the lack of multiple transmitters in mobile devices (e.g. single spatial stream (1SS) devices)….hence the moniker “beamformee”

4. At close range, you don’t need TxBF. At long range, the radio environment becomes too complex for TxBF to be effective. Therefore, TxBF is only worthwhile at mid-range, and even then, what does it really give you? 1-2 data rates of improvement, best case? Hardly worth discussion.

  • When clients move too quickly in the environment, TxBF becomes less effective. Mobile client may receive a better signal, but that may very well unbalance the AP/client connection because client devices operate at such low power (due to being battery-powered), and the AP radios aren’t sensitive enough to compensate, that you have one-way communication (always a bad thing).

5. You mentioned vendors choosing Atheros for their APs and Atheros not supporting TxBF. It may seem like a good differentiator to mention this differential for an AP, but what about the clients that use Atheros, Ralink, and a long list of other chipsets that don’t support TxBF?  They are not helped by 11ac TxBF, but they are helped by Ruckus’s Beamflex (as are Broadcom client devices).

  • In a nutshell, if you have a Broadcom-enabled client with multiple transmit radios, and it’s drivers support TxBF, and it’s certified for TxBF by the Wi-Fi Alliance, then you’re good to go. That’s a tiny fraction of the 11ac clients available in the market today (easily sub-5%, but probably more like sub-2%).

6. Ruckus Beamflex is not analog nor analogous to analog, nor is 11ac TxBF a digital technology nor analogous to digital technology.

  • They are clearly different technologies, and as you said, Ruckus APs, to date, use Atheros (Qualcomm) chipsets. It’s not that Atheros isn’t going to support TxBF, but rather they are just a tad slow getting that feature into the source code that vendors can use. Atheros will get there, but truly, it doesn’t matter if they take their sweet time because TxBF is mostly worthless technology in today’s market (e.g. mobile clients don’t support it, non-Broadcom clients don’t support it, it’s useless at short or long range, etc). If I were a prospective customer, I would not buy 11ac from anyone based on 11ac TxBF capability, especially when there are far more important features to discuss that do matter.
  • Had I been Ruckus’s VP of PLM, I’d have made the same decision for the same reason(s)….and BTW, how is Ruckus supporting both Beamflex plus 11ac TxBF technology a bad thing again? Doesn’t Aruba use the same tactic (e.g. “we support both” against Aerohive when talking about controllers and controller-less technology)? #JustSayin

7. You stated:

  • "First, analog beamforming can’t accurately determine a client’s location so you can’t flex a beam to that client." << This is incorrect. Ruckus’s Beamflex technology (in the AP) can, depending on the type of AP, use in excess of 4,000 transmission patterns. Because of these patterns, the AP knows where each client device is located (vertically and horizontally).
  • "Second, it’s a manual hit-and-miss process that involves tweaking antennas to steer RF beams at clients." << This is incorrect. With Beamflex technology, client devices are tracked in near-real-time using ACKs (at a minimum). It’s equally as fast as TxBF, and in fact, in highly mobile environments tends to be far superior due to two-way communication not being required. I’ve tested it, and the proof is in the pudding.
  • "Third, too much beam, too little MIMO." << That’s absolutely wrong. In fact, I compared the results of a Ruckus 7982 (3x3:3 11n) at a distance of ~7 feet from a MacBook Air (2x2:2 11ac) to an Aruba AP-225 (3x3:3 11ac) at that same distance. There was no comparison. Ruckus’s AP was able to choose the best 2 spatial streams and steer them to the optimal points in the room to get the best throughput, and handily beat the AP-225 by a country mile. I forwarded this data to Aruba’s TME team immediately, asking for their opinion, and got no response. Again, #JustSayin

8. I think the test results posted in this blog are grossly bogus. I challenge you to reveal all of your test data, methodology, and environmental characteristics of the test, including all of the instances and scenarios where Ruckus’s R700 handily beat the AP-225. Fair is fair, right?…publish all of the results or don’t publish any of it.  Or is this just marketing hooey?

  • Further, the MacBook Pro you used in your test is not Wi-Fi certified, so it shouldn’t have been used for the test.


In closing, it would be my suggestion to delete this thread altogether (including my comments), learn from the experience, and move away from the heavy-handed technical marketing. We (Aruba and it’s channel partners, like me) need more solution support from the TMEs and alot less competitive non-sense. Believe me buddy, I’ve done it all wrong in my time too (as many of my friends would attest to), and I've learned from the experience. Focus on solving customer pain and, to borrow a phrase from a friend, skate to where the puck will be.

Devinator


neela
May 23, 2014 12:36 AM

Thanks. Appreciate the response. 

rmalaiyandi@arubanetworks.com
May 22, 2014 06:45 PM

Well written article and easy to understand.

 

thanks  ramu

neela
May 18, 2014 01:08 PM

Standards-based 11ac beamforming or explicit beamforming is synonymous to two people having a conversation. The AP (beamformer) sends a sounding frame to the client (beamformee). Based on the sounding frame received, the client sends a feedback frame, which contains the channel estimate. From the feedback received, the AP sends the beamformed frames to the client.

 

 Screen Shot 2014-05-18 at 10.07.15 AM.png

 

This frequent conversation gives an accurate representation of “everything” that happens between the baseband of the AP and the baseband of the client. “Everything” here refers to traces on the boards, variation in transmit amplifier in the AP and the LNAs in the client. These are fully accounted for in the channel calculation made by the client chipset.

 

Analog beamforming takes a longer time to locate the client and focus the beam. This is because the AP first needs to use permutations to scan through the available antenna pattern to determine which one to use. This approach takes time and it does not get an active channel estimate from the clients.

 

What happens if the client is mobile? By having a two-way conversation the 11ac beamforming algorithm is updated with channel estimate 40 time per second. This allows the system to completely track the user and keep the signal strength and MIMO maximized at the clients. Analog systems by contrast, take longer and have less information to maximize the client signal.

 

11ac beamforming maximizes MIMO. The antenna patterns created is a result of a complete solution that takes into account the environmental reflections. Building an AP that focuses a beam too much solves only half of the problem and can interfere with MIMO operation. Reduced MIMO results in reduced performance.