So who can you trust?

I know next to nothing about Frost & Sullivan, but based on the information on their web page they are “the world leader in growth consulting and the integrated areas of technology research, market research, economic research, corporate best practices, training, customer research, competitive intelligence and corporate strategy.” They seem to be a very respectable market analysis company. If one reads material produced by a Frost & Sullivan analyst, one expects that it would be credible and trustworthy. So what is one to make of reading a recent article which includes the following quote attributed to an F&S analyst:

“The company’s sustainable competitive advantage lies in this proprietary modulation method, which differs from rival technologies that use tens to hundreds of thousands of waves to convey the same bit of information,” explains Thomas. “As each additional cycle requires commensurate power output, xG’s Flash Signal technology allows dramatic efficiency gains to be achieved and results in increased RF signal range.”

and also

“… adds Thomas. “Moreover, in the context of power efficiency resulting from single-cycle signals, tests have shown that an xG base station transmitting full-motion video uses 3 million times less power than a typical 802.11 access point.”"

How can anyone with some knowledge of wireless communications say this with a straight face?

The Thomas quoted here is (based on his web page) “Luke Thomas – a Research Analyst with Frost & Sullivan’s ICT Europe practice, specializing in wireless content and applications, and wireless broadband technologies.” In an earlier post “Theater of the Absurd” I have addressed a presentation by the same analyst where xMax is described as a likely competitor to WiMax and 3G LTE (I am not making this up!).

I am also completely baffled by the “Technology Innovation Awards” given by Frost and Sullivan to Gaiacomm International in 2004 (see my post “Wild Wireless”), and to xG Technology in 2007. What kind of a technical review was conducted before deciding to give these awards? I can’t imagine how any competent communication engineer could have approved these awards. I encourage the reader to look up Gaiacomm to fully appreciate the “reality failure” involved here.

So how can we trust anything we read on the web about wireless communication technology, even if it comes from seemingly respectable sources? If anyone has an answer to this rhetorical question please let me know ….

BER vs. Eb/N0 plot – round III

So today a new “disclaimer” appeared on the xG Technology web page above the BER vs. Eb/N0 plot saying:

“Range and penetration are functions of both Eb/No and system characteristics. This chart depicts the Bit Error Rate (BER) of xMax vs. other typical modulation systems based solely on equal Eb/No, but does not depict the performance provided by xMax system gain which can add substantial advantage.”

Two communication systems differing by the modulation technique, such as xMax and a system using conventional modulation, but otherwise having the same parameters (data rate, transmit power, transmit/receive antenna gains, operating frequency, bandwidth, noise figure, etc.), will have the same range and penetration. In other words, in a fair one-to-one comparison of two communication systems, knowing that they require the same Eb/N0 to operate at a specified BER, immediately tells you that they will have the same range and penetration.

Having published a BER vs. Eb/N0 curve showing that xMax performance is very close to that of conventional communication systems, it makes no sense from a technical standpoint to argue that xMax has better range and penetration (again, assuming a fair comparison). As I tried to explain in an earlier post “Understanding the range of wireless systems”, the key issue is the maximum pathloss which the communication system can overcome. This pathloss can be calculated doing a link budget. The effect of the modulation technique enters the link budget only through the required Eb/N0.

To put this in simple and completely non technical terms: you can’t have your cake and eat it too ….

The xMax story to date

Given the various questions that keep coming from readers I want to briefly summarize where we are on the xMax issue from a technical standpoint.

xMax is a new modulation technique using a particular type of waveform. It’s performance is comparable to that of conventional techniques. As a physical layer it does not offer any advantage over the physical layers used in existing systems. Statements made on the xG ompany webpage such as “… xG Flash Signal will offer significant improvements in speed, range, and power savings over existing technologies.” and “xG Flash Signal is a breakthrough system design that improves RF signal performance at its most elemental level – the physical layer. Flash Signal uses revolutionary single cycle modulation to deliver longer range and lower power communications.” are not supported by any data and are in direct contradiction of well established principles of communication theory, as well as xG Technology’s own BER curve. All of the claimed advantages of xMax derive from the one basic claim of power efficiency which is simply false. Given xG Technology’s publication of the BER vs. Eb/N0 curve I hope that we can put to rest these claims and move on to other more interesting topics.

The fact that xMax offers no advantages as a physical layer does not mean that it can not work. Technical analysis of xMax shows that in principle it can be used as the physical layer of a wireless network, provided that xG Technology addresses the myriad of other issues involved in building such a network. It should be clearly understood, however, that the xMax based network can not perform better than a network built using conventional communication technology. While I find it puzzling that someone will go to so much trouble to re-invent the wheel, when perfectly good (or better) wheels are readily available, this is a business decision which is beyond the scope of what I want to discuss in this blog.  There may well be valid non-technical reasons for this approach.

As far as I am concerned we seem to have exhausted the issue of the xMax modulation technique – there is simply nothing very interesting there. Now a complete working system involves many issues beyond the choice of the modulation technique which has been the focus of the discussion so far. There are physical layer issues such as the multi-access method (FDMA, TDMA, etc.), handling of fading, rate adaptation, choice of coding, power control, use of multiple antennas, and so on. Then there is the MAC layer and its various protocols, scheduling of users, Quality of Service, mobility, handoff, and the list goes on. Very little information seems to be available at this time on how these topics are being addressed in the overall xMax system. The information provided by xG Technology, its patents and the media, focuses almost entirely on the modulation technique and its non-existent advantages. This is rather peculiar and potentially worrisome. I hope to discuss some of these issues if and when some meaningful technical information about them becomes available.

BER vs. Eb/N0 plot – round II

[Note added on 28 July, 2007:   The disclaimer discussed in this post seems to have been removed today from the xG Technology web page]

Well, I spoke too soon. I thought that xG Technology has finally decided to come to grips with the fact that xMax has no technical advantages as a physical layer, and move on from there. I was wrong. They have now added the following “disclaimer” to the BER vs. Eb/N0 plot:

“In response to numerous requests, following is a plot showing BER (bit-error-rate) performance against Eb/No (signal-noise ratio). It should be noted that, due to the unconventional nature of xMax technology, traditional BER plots such as these do not convey the true inherent advantages of xMax over existing technologies. These advantages will become evident in the first commercial xMax systems.”

So xMax has some unspecified “true inherent advantages over existing technologies.” These must be different from the advantages mentioned before, which relied on the claim that xMax can operate at vastly lower power levels than existing techniques, a claim clearly falsified by the BER vs. Eb/No plot (as well as by all the other issues discussed in earlier posts). Sounds rather mysterious and intriguing – we will have to wait and see what new advantages arise from the ashes of the old.

Ber vs. Eb/N0 plot

A BER vs. Eb/N0 plot has appeared on the xG Technology webpage. The plot shows the BER curve for xMax just to the right of a BER curve for BPSK. There is no text explaining the curve so one can not tell if this is a simulated or measured result. However, it does show xMax to require about 1dB higher Eb/N0 than BPSK. This is certainly plausible. This curve clearly and unequivocally shows that xMax is not more power efficient than conventional modulation techniques. In other words, it verifies what I have been saying in previous posts that the claims that xMax can operate with less power than conventional techniques are false. I am glad to see that xG Technology is finally coming to grips with this fact. I hope they will now publish a retraction of these claims, consistent with this technical information.

Given that xG Technology has been characterizing this power efficiency (which their own data shows to be nonexistent) as “the key value proposition of xMax” I assume that this is not the end of the story. Stay tuned for fast breaking developments …

A little knowledge is a dangerous thing

A reader wrote to me a rather angry message about my critical technical evaluation of xMax. He chastised me for ignoring the fact that xMax is able to reduce the received noise and interference by many orders of magnitude more than conventional systems, thanks to the Wavelet Pass Filter (WPF) which Bobier describes in several patents, especially in “Integer Cycle Event Detection using Wavelet Pass Filter System and Method“. He also pointed me to various articles such as “Going Beyond Interruptible Usage” which say:

“The Wavelet Pass Filter is the key to xMax. This device, which provides significant signal processing gain, allows the receiver to recover the weak information-bearing signal found amidst the narrowband interference and noise from legacy and neighboring users in the adjacent sidebands.”

In another article we read that:

““We’re talking about a 25 to 45 decibel advantage in an industry where 2 decibels is worth killing for,” Bobier said. “And you don’t have to throw a lot of spectrum or power at [producing a broadband signal], if you can get rid of the noise.” By virtually eliminating the typical noise floor, xMax enables the power levels for its information-bearing signal to be as much as 100,000 times below the FCC’s current regulated power limit for out-of-band emissions, which is designed to prevent systems from interfering with each other. In fact, xMax’s power levels are 10,000 times below the FCC’s power limits for ultrawideband (UWB) transmissions, according to xG officials.”

In other words, says the reader, xMax has this clever filter which virtually eliminates noise and interference. That is why it can operate with much less power. So there!

I have to confess right here and now that I did indeed ignore this claim of superior noise and interference reduction. I did so because this claim, like many other technical statements in the xMax literature and patents, is wrong and is based on ignorance of basic principles of signal processing and communications. So let me take a deep breath, get something cold to drink and maybe an aspirin or two, and try to explain briefly.

A communication system is required to detect a known signal (waveform) in the presence of thermal noise and interference. Let us ignore for the moment interference and discuss the noise. Thermal noise is known to be represented very accurately as a white Gaussian random process. The theory of detecting a known signal in white Gaussian noise is treated in standard textbooks dealing with detection or communication. The optimal detector, i.e. the detector which gives the smallest bit error rate, is known to use a matched filter whose output is used to determine whether a 0 or a 1 was transmitted. The signal-to-noise ratio (SNR) at the output of the matched filter is higher than the SNR at the output of any other filter. In other words, no other filter has a better performance than the matched filter.

Not surprisingly, all conventional communication systems use matched filters in their receivers. The only communication system I have seen which does not use a matched filter is xMax, which uses instead the WPF, which according to its description is not a matched filter. Consequently, the SNR at the output of WPF can only be worse than the corresponding SNR at the output of a matched filter. There is no doubt that the noise reduction capability of the WPF is not better, and is in fact worse than that of the matched filter used in conventional communication systems.

Next let us consider what happens in the presence of interference. Interference can be wideband or narrowband. Wideband interference acts pretty much like white Gaussian noise and therefore the matched filter is still the optimal solution. Narrowband interference has a different character and needs to be treated separately. Conventional communication systems use a variety of techniques to deal with such interference. Spread spectrum techniques (such as CDMA) convert the narrowband interference into wideband noise at the receiver, so we again can use the optimal matched filter. In multi-carrier systems such as OFDM the narrowband interference effects only one (or a few) of the frequency bins. The effect of this interference is eliminated by coding across the frequency bins. To summarize: conventional communication systems are already designed to optimally handle the effects of interference. The sub-optimal WPF offers an interference mitigation capability which is inferior to that of filters used in modern communication systems.

To conclude: the claim that the WPF can somehow reduce noise and interference more than the reduction which takes place in existing highly optimized communication systems is  false. Any competent communication engineer should recognize this claim as making no technical sense.

I really wish that the author of the WPF patents would have taken a few basic courses in signals and systems, analog/digital signal processing, and communications. Or at the very least, that he would talk and listen to some people who do have this knowledge. This would have saved me a lot of typing. A little knowledge is indeed a dangerous thing.

As a side comment, the patent “Integer Cycle Event Detection using Wavelet Pass Filter System and Method” which describes the WPF in excruciating detail reveals a lack of knowledge of well known filter design techniques. This long document has numerous technical mis-statements whose description would require a separate post.

Yes I can, no you can’t (repeat as necessary)

I keep getting e-mails saying that I am putting too much emphasis on the single claim that “xMax can use 1,000 to 100,000 times less power than comparable transmission technologies” and discussing it to the exclusion of other aspects of xMax, and in general making too big a deal of this, and what’s with this. Well, that may be so, but at least I am in good company. Looking at the xG Technology webpage I see the following statement:

“The key value proposition for xMax is that it lowers the cost of deploying broadband services. xMax accomplishes this by increasing the range of RF signals. By delivering broadband signals significantly farther than other technologies operating at the same frequency and power level, xMax reduces the amount of infrastructure required to cover a given area wirelessly by 25-50 fold depending on the terrain.”

Just so the reader understands, this is equivalent to saying that xMax can operate at 1000 – 100,000 times lower received power than “other technologies operating at the same frequency and power level”. To see this consider the following. A 25-50 fold increase in area coverage equals a 5 – 7 fold increase in range. Propagation loss in a typical urban area increases as the fourth power of distance, so 5 – 7 fold increase in range corresponds to an increase in pathloss which equals 5 to the fourth power to 7 to the fourth power. This in turn translates approximately to 1000 – 100,000 pathloss increase, or equivalently to a 1000 – 100,000 decrease in power (well a bit less, but I like nice round numbers). This is entirely consistent with statements attributed to xG Technology in many articles and documents one can find on-line.

In other words the xG Technology webpage clearly states that this claim is the key value proposition of xMax. What can I say? I happen to agree with them, and that is why I am discussing it. That it also why it should be very important to anyone interested in this technology to want to see clear unambiguous proof of the validity of this claim (such as the BER vs. Eb/N0 plot). Unfortunately, as I have explained in detail in earlier posts, this claim is in fact false. It is therefore not surprising that such proof is not available even at this late stage where the product is supposed to be ready for imminent rollout.

xMax and its deployment (part II)

A reader asked me a very interesting question. It seems that the xG Technology handsets are planned to be dual-mode and will have both xMax and WiFi. This is stated on the company web page among other places. So the question was: “How will one be able to tell if the handsets work on the xMax link or the WiFi link?” The direct answer is that the users will not know, unless the handset was designed to give an indication of which radio link is being used. Even then, most users will not pay any attention because they do not care how they are connected. However, what makes this question really interesting is what happens if a WiFi radio were to be co-located with the xG Technology basestation! The reason why this is so interesting is as follows.

In previous posts I have attempted to explain why xMax does not have a performance advantage over conventional communication systems. There is every reason to believe that the link budget of xMax is comparable to that of WiFi. Furthermore, contrary to popular opinion, a WiFi transmission can reach a very long distance under favorable propagation conditions (such as being placed on an 850 foot tower). In fact, all things being equal (transmit power, antennas, location, etc.), WiFi and xMax should have roughly the same range.

Consider the case where co-located with an xMax basestation there is a WiFi access point operating at maximum power. In this case an xMax handset may be as likely to receive the WiFi signal as it is the xMax signal. In fact, because the handset may lock on to one of the many local WiFi transmitters (hot-spots, residential access points etc.) , it may be much more likely that it will operate on the WiFi radio than on the xMax radio. This raises the intriguing possibility that a handset with the xMax radio completely turned off, may function nearly as well (or as poorly) as a handset in which the xMax radio is on! Unless the operator of the network releases statistics which report separately the performance of the xMax link and the WiFi link, it will be impossible to tell what is going on.

All of this makes me very curious as to the following: (i) Will xG Technology deploy a WiFi basestation co-located with each xMax basestation? (ii) Do the xMax handsets have an indicator showing which radio is being used? (iii) Will the operator of the network release information which will allow evaluation of the xMax performance? I do not know the answers to these questions, If any readers have some relevant information, please drop me a note.

The main point here is that it is may be possible to have an xMax deployment which produces no useful information about the performance of xMax! This it yet another compelling reason why xMax should be evaluated by standard laboratory tests, such as measuring the BER vs. Eb/N0 curve.

Let me emphasize again that I am talking specifically about evaluating the claim that “xMax can use 1,000 to 100,000 times less power than comparable transmission technologies.” What I am saying is that an xMax deployment in itself is not likely to provide a clear validation (or rather falsification) of this claim, because it involves too many unknown and uncontrolled variables. A simple laboratory test is a far more reliable way to get a definitive answer.

A technical comment to those interested.

A standard WiFi radio will not work over long distances even if the signal to noise ratio is high. This is because the transmitter expects an acknowledgment from the receiver within a time window which is set to accommodate short delays corresponding to the shorter ranges over which WiFi is normally used. However, it is straightforward to fix this by changing certain software settings, and this has been done many times to allow long range WiFi operation. The current record for WiFi is a 237 mile link! Because xG Technology designs both sides of the link, they could if they wanted to, allow the WiFi radio to operate over much longer distances than the standard WiFi you get in your local store. All it takes is a software change. I have no idea whether or not they are doing this. I certainly would expect them to do so if they intend to place a WiFi radio next to the xMax baseststion or as part of it.

Broadband networks and the unlicensed band

Deploying a broadband access network in the unlicensed band where the spectrum is free, is very tempting. So why are companies paying large amounts of money to buy licensed spectrum? The main reason is interference and its impact on the reliability of the network. The unlicensed band attracts many users with a variety of devices which all operate in the same frequency band causing potential interference. There are many misconceptions about interference, such as: because the allowed power levels are relatively low, interference is not a serious problem. To see why this is not necessarily true consider the following example.

Assume we have a basestation communicating with a user 1 mile away in an urban area in the 900 MHz unlicensed band. The pathloss between the user and the basestation (computed using the Hata model) is 140 dB. Next consider an interfering device located within line of sight of the basestation at a distance of 0.1 miles. The pathloss between the interferer and the basestation (computed using free space propagation) is 70 dB. We will say that the interferer has a “power advantage” of 70dB over the user. In other words, if the user and the device use the same power (more precisely the same effective radiated power, or ERP), the interfering signal at the basestation antenna will be 70dB stronger than the user’s signal. This is assuming the same antenna gain in the direction of the user and the interferer.

Next consider user located 10 miles from the basestation. The pathloss between that user and the basestation (computed using the Hata model) is 178 dB. In this case the interferer has a power advantage of 108dB!

We see that the interfering signal may be enormously more powerful than the desired signal because of the difference in pathloss. This difference is true regardless of the power levels used, as long as we assume that the user and interfere use the same power.

The situation can be somewhat ameliorated by the processing done at the receiver which may enhance the desired signal relative to the interfering signal. We will refer to this as the “processing gain” of the receiver. We will therefore say that the  interference to signal ratio after processing (in dB) equals the pathloss difference (in dB) minus the processing gain (in dB).

The processing gain will depend on the characteristics of the communication systems as well as of the interference. In some cases the gain will be small or non-existent. In other it may be large. For example, if the communication system has a bandwidth of 20MHz and the interference is narrowband, the receiver may offer a large process gain. If the communication system and the interferer use exactly the same kind of signals, there may be no processing gain at all. Even if the processing gain is significant, it will generally be much smaller than the pathloss differences shown in the examples above, causing the interfering signal to overwhelm the desired signal from the user.

The point of this discussion is to show that an interferer near the basestation can “jam” the basestation and make it unable to communicate with the users. Similarly, an interferer located in close proximity to the user may “jam” that user and make it unable to communicate with the basestation. Here we focused on the basestation “jammer” because its effects are potentially more severe in that it can cause shutting down a complete cell, rather than just one or a few users.

Deploying a cellular network in the unlicensed band is therefore risky business. A single interferer in the wrong place can effectively disrupt an entire cell. One can hope that this is going to happen very infrequently, but when it happens the consequences can be quite unpleasant. So if you are committed to deploying in the unlicensed band be prepared to live with the sword of Damocles hanging over your head …

Why not?

I was hoping that readers of this blog would ask more technical questions which would allow me to provide technical explanations. Unfortunately, most of the questions I get are related to xMax, and are not really technical in nature. Recently some readers asked: “How come only you are writing critical assessments of xMax?” (actually there is one other). Others asked “Why isn’t there a more widespread technical discussion of xMax?” The answer of course is that I do not know. However, I can make a reasonable guess. There are several possible reasons, including:

(i) The xMax technology is virtually unknown to people involved in wireless communications. (ii) To get knowledgeable people interested in a new modulation technique they first need to see a credible technical exposition of the technique which will motivate them to spend the time and effort to study it and write about it. No such discussion is currently available. (iii) The information that is currently available about xMax is of such poor technical quality that a competent communication engineer or researcher will not give it a second look. In particular, the false performance claims and the misconceptions evident in the patents discourage any serious examination of xMax.

Before seeing more discussions of this technology by people who are experts in the field, there will have to be an openly available credible technical document describing xMax and its performance. From what I have seen so far, and given xG Technology’s penchant for secrecy, this is not likely to happen. More fundamentally, xMax is just not that exciting from the point of view of communication theory. It is based on a waveform that is different from the waveforms used by conventional modulation techniques, but otherwise does not seem to offer anything new, and does not offer any performance advantages over existing techniques. It is well known that there is an infinite variety of communication waveforms that can be used to design a communication system. While it is certainly of interest to learn of the implementation of a system which uses a waveform which has not been used before, there is nothing particularly surprising about that fact, given that there is an infinity of waveforms to choose from.

The only truly exciting thing about xMax, which was what got me interested in this in the first place, is the claim that “xMax can use 1,000 to 100,000 times less power than comparable transmission technologies.” It is most unfortunate that this claim is false.