Ted's Corner Archives - Page 2 of 2

August 31, 2016

Is the cellular standard roadmap the answer to Low Power Wide Area Connectivity? Pt. 1

Part One: Introduction

Low Power Wide Area (LPWA) networking technologies are attempting to connect 10s (even 100s) of Billions of currently unconnected endpoints. The cellular industry is aggressively pursuing an offering for Low Power Wide Area (LPWA) through the Third Generation Partnership Project (3GPP) standards body. By doing so, the cellular industry is conceding that a new approach is required and that the current cellular systems (2G, 3G, 4G LTE) do not address the requirements of these LPWA devices. There are several flavors of the cellular LPWA roadmap including NB-IOT, LTE-M, and EC-GSM.

However, the cellular LPWA approaches are over-constrained and will disappoint customers for many of the same reasons that cellular technology to date has been disappointing to the customers.

August 1, 2016

The “WA” in LPWA is Wide Area… and Pizza

I would like to address an aspect of Low-Power, Wide-Area (LPWA) networking that is extremely important yet receives little attention: capacity. Capacity is a measure of how many devices a piece of networking infrastructure may support and/or how rich of a data model a network can provide to those devices.

Below are LPWA technology comparisons that either ignore capacity entirely, or use unjustified vendor claims that are often simply based on the amount of addressing space supported.

http://frankrayal.com/2016/06/11/comparative-look-at-lpwa-performance/

http://lora.multitech.com/white-papers/lpwa_white_paper_86003000_2016-06-13.pdf

http://www.eejournal.com/archives/articles/20150907-lpwa/

http://www.eetimes.com/document.asp?doc_id=1328365&page_number=3

These comparisons spend much of their attention on supported link data rates, thus missing the point fundamentally on what a wide area network (WAN) is supposed to do. A WAN in general is not about delivering data rates to a single endpoint, rather, it is fundamentally about providing high-capacity connectivity to a large number of endpoints.

Here’s an analogy. When you order a pizza, would you be comfortable understanding a range of size of individual pizza slices you might receive (analogous to range of data rates), but giving the pizza parlor full control of the number of slices that they will deliver as well as the sizes of those slices (as long as they are within the stated pizza slice range)? If this pizza slice scenario makes perfect sense, then these “data rate centric” frameworks are for you. But, if you’re like me, and prefer ordering a pizza based on the size of the pizza, then you would likely be more comfortable with a different measure – one that examines the “capacity” of the system. Just as the size of a pizza is indicative of the number of hungry family members that can be satiated (or often over-satiated), capacity is a measure of how many IoT devices may be served by a network based on their data needs. We should probably have colored the circles more pizza-colored for clarity. As you can imagine, insufficient capacity means not being able to build a profitable network…or analogously, whether that pizza parlor can stay in business.

If you would like to understand the details behind these calculations, please check out Appendix A through C of How RPMA Works: The Making of RPMA.

On a side-note, I was surprised to see RPMA represented as inferior coverage in this framework. We work through the link budget calculations in Section 2.3 of the same document (How RPMA Works: The Making of RPMA) and make the point that RPMA has a massive coverage advantage.

Capacity Comparison

July 22, 2016

The Economics of Receiver Sensitivity and Spectral Efficiency

We have asserted textbook style categories for Low-Power, Wide-Area (LPWA) connectivity Blog 1: Categories of LPWA Modulation Schemes. Then we went back to basics to get some framework and vocabulary in place to discuss these various schemes in Blog 2: Back to Basics – The Shannon Hartley Theorem, where we introduced the critical concepts of spectral efficiency (η), Eb/No, and how a point of diminishing returns can be reached in the quest to reach the Shannon Limit. In Blog 3: Chirp Spread Spectrum (CSS): The Jell-O of Non-Coherent M-ary Modulation, I confirmed that CSS can be treated more generally as Non-Coherent M-ary Modulation (NC-MM).

And then in Blog 4: “Spreading” – A Shannon-Hartley Loophole?, we discussed in more detail technologies that use links with very low spectral efficiency (η << 0.1) and the capacity implications with three concrete examples: the Sigfox® technology, LoRa™, and Ingenu’s RPMA®.

In this blog, we will set the stage with a description of a public network business model including how the fundamentals of coverage and capacity fit into building that business. Given that framework, I will then go onto discuss how the three technologies highlighted below (the Sigfox technology, LoRa, and RPMA) compare on the economics of coverage (building the network), capacity (monetizing the network), and capacity scalability or cell splitting (growing a successful network.)

Category

Local Area Network

Wide Area Network

Ultra Narrow Band

(UNB)

Sigfox

Telensa

N-Wave

WaveIOT

Non-Coherent M-ary Modulation (NC-MM)

Bluetooth

802.11b

LoRa (CSS)

Sensus (7-FSK)

GSM/GPRS

EC-GSM

Direct Sequence Spread Spectrum (DSSS)

802.11

Zigbee

W-CDMA

RPMA

Orthogonal Frequency Division Multiplexing

(OFDM)

802.11a/g

LTE

WiMax

NB-IOT

Setting the Stage: The Public Network Business

The public network business requires two general components: a carrier, and applications that may benefit from that network. Let’s define these roles and look at a public LPWA network from these points of view:

The carrier owns and operates the LPWA network. The carrier invests in building this network, and charges applications for the use of this network. The successful carrier business profits by the revenue (connectivity fees from the applications) exceeding the expenses of running the LPWA network (tower rental, backhaul expenses, construction costs, human resources, etc.)
Applications are benefitted by the connectivity that the LPWA network provides. For an application to participate in an LPWA network, there must be a positive return-on-investment (ROI) of this connectivity. In other words, the value provided by the LPWA connectivity must exceed the connectivity fees paid to the carrier. Moreover, if given a choice, the application will want to maximize their ROI by selecting the LPWA network that minimizes their connectivity fees.

The LPWA business begins with the carrier. Typically, the carrier must invest in building the network prior to revenue being realized. Let’s introduce a few key terms to bridge the LPWA economics to the LPWA technology:

Coverage – a metric of how much network infrastructure is required to reliably cover a region. Most of the expenses associated with running a carrier are proportional to the amount of network infrastructure required (e.g. tower rental, backhaul expenses, construction costs) particularly as the geographic extent of the network becomes large. The number of square miles (or square kilometers) covered, on average, by a piece of network infrastructure (e.g. tower) represents the initial investment a carrier must make to build this network.
Capacity – a metric of how many devices, on average, may be supported by a piece of network infrastructure. The capacity metric is relevant to the revenue side of a successful carrier business. The amount of revenue per tower to the carrier will typically be directly proportional to the number of endpoints served by that tower.

From a carrier’s point of view, investment in coverage is a slight “dip in the road” in terms of outward cash-flow; capacity represents that “road up the mountain” in turns of how profitable that network may be based on number of devices supported.

Coverage

To build a network economically means that each piece of infrastructure must cover large amounts of area with high probability. This means that the range of the link must be as high as possible, which, in turn, means the receiver sensitivity must be a good as possible.

In the figure below, we show the amount of typical reliable coverage that can be expected with the three approaches based on the analysis in Section 2.3 of How RPMA Works: The Making of RPMA. Naturally, the cost of a carrier covering a large region is far lower if each tower is capable of covering the large area shown for RPMA.

coverageimageblog5

Capacity

It is important for a technology to allow for connection of a sufficient number of endpoints per piece of networking equipment to make long-term economic sense. There are costs to delivering LPWA connectivity including the infrastructure cost, the deployment cost, and any maintenance costs. These costs need to be shared among a tremendous number of devices such that each device’s share of the burden is very low. Keep in mind, these are typically low-value devices that can only bear minimal networking expense. If you take a look at the figure below, you will see that RPMA supports a factor of 60x to 1300x more data per piece of networking infrastructure (which equate to more 60x to 1300x more devices) relative to LoRa (exactly where in that range depends whether we are talking uplink/downlink and the particular regulatory domain). The Sigfox technology numbers are similar. Not surprisingly, the 60x to 1300x reduction in the per device connectivity cost is the difference between making the economics of the network work at all, and being able to make the economics of the network work easily.

The Sigfox technology is argued to serve only very low-bandwidth devices and thus, capacity is not an important attribute for the low-end devices they serve. There are two main problems with this argument:

Low-usage devices tend to justify only low connectivity costs. If the Sigfox technology is constrained only to low-usage devices, the carrier requires more of these devices to build a business. Whether the endpoint distribution is fewer high-usage devices, or more numerous low-usage devices, capacity is being consumed, and as such, it is an important figure of merit to understand if there is any economic value to be split between carrier and application.
Even if there is some minuscule amount of economic value to be split between application and a carrier employing Sigfox technology, an RPMA carrier will always be able to undercut the connectivity costs due to the tremendously unfair capacity advantage. An RPMA carrier will be in a position to offer more link capability at a small fraction of the cost, and that RPMA carrier will remain massively profitable.

Note the number of devices supported on average per piece of network infrastructure as shown below and imagine the amount of revenue per endpoint for the Sigfox technology and LoRa. In our opinion, these numbers will not support a profitable carrier business model if you assume reasonable connectivity cost per endpoint as revenue.

An additional headwind LoRa in particular faces is the lack of selectivity (discussed in more depth in Section 2.10 of How RPMA Works: The Making of RPMA.) where endpoints that are deployed on private LoRa networks actually consume capacity on overlapping public networks – far more capacity, as a matter of fact, than had they actually been connected to the public LoRa network.

RPMA's capacity is 10s to 100s of times more than other LPWA technologies.

Capacity Scaling (Cell Splitting)

What happens when a tower is at capacity based on the number of devices shown above? As a carrier, you would probably like to add more tower locations to continue offering robust connectivity. The cellular industry has a term for this called “cell splitting” and it can be done effectively. However, not all technologies allow for offloading capacity by adding towers. All the technologies in the LAN category share this aspect including the Sigfox technology and LoRa. Due to lack of support for transmit power control (among other things), once a critical density of endpoints is reached, the system simply ceases to work robustly.

By contrast, RPMA has been built from the ground up as an LPWAN solution. It was not easy or fast to do this. It took considerable investment because a new technology had to be developed from scratch, along with a whole new type of chip to inexpensively implement the new technology with minimal cost and power. The most important thing now is that RPMA is done, commercialized, and networks and endpoints are rapidly being deployed! We believed (and we still strongly believe) that with LPWA device counts in the 10s of billions or more, an investment in a purpose-built technology was well worth it.

In future blogs, we will explore additional topics such as sunsetting, NB-IOT looming technical issues, standardization, RPMA commercial progress, and development of RPMA technology. If you would like an advanced peek into many of these topics please take a look at the document How RPMA Works: The Making of RPMA.

July 22, 2016

“Spreading” – A Shannon-Hartley Loophole?

We have asserted textbook style categories for Low-Power, Wide-Area (LPWA) connectivity Blog 1: Categories of LPWA Modulation Schemes. Then we went back to basics to get some framework and vocabulary in place to discuss these various schemes in Blog 2: Back to Basics – The Shannon-Hartley Theorem, where we introduced the critical concepts of spectral efficiency (η), Eb/No, and how a point of diminishing returns can be reached in the quest to reach the Shannon Limit. In Blog 3: Chirp Spread Spectrum (CSS): The Jell-O of Non-Coherent M-ary Modulation, I confirmed that CSS can be treated more generally as Non Coherent M-ary Modulation (NC-MM).

So, what is “spreading”? From the most generic perspective, it is increasing receive sensitivity (thus, a longer range and more robust link) by reducing the data rate and keeping the same channel bandwidth. That underlined part is very important and represents a potential looming problem: the link data rate is becoming a trickle, and yet that link takes the same amount of valuable spectrum. The result is that the capacity could become vanishingly small. But the capacity does not have to – and therein lays the loophole.

Let’s use what we’ve learned to determine the capacity of a few well-known Low-Power, Wide-Area (LPWA) “spreading-based” connectivity technologies [LoRa™ (CSS) and RPMA®] highlighted in the table below and additionally, a well-marketed Ultra-Narrow Band (UNB) approach used as the Sigfox™ technology for comparison purposes.

Category

Local Area Network

Wide Area Network

Ultra Narrow Band

(UNB)

Sigfox

Telensa

N-Wave

WaveIOT

Non-Coherent M-ary Modulation (NC-MM)

Bluetooth

802.11b

LoRa (CSS)

Sensus (7-FSK)

GSM/GPRS

EC-GSM

Direct Sequence Spread Spectrum (DSSS)

802.11

Zigbee

W-CDMA

RPMA

Orthogonal Frequency Division Multiplexing

(OFDM)

802.11a/g

LTE

WiMax

NB-IOT

Recall from Blog 3: Chirp Spread Spectrum (CSS): The Jell-O of Non-Coherent M-ary Modulation, we introduced the calculation of the individual link spectral efficiency (η) for LoRa. Let’s take that information and insert the Sigfox technology and RPMA for comparison. The Sigfox technology looks great and RPMA looks worse than even the meager spectral efficiency of LoRa. Uh oh…Looks like RPMA is off to a rough start.

But wait, this analysis is far from complete in terms of a capacity calculation. RPMA is a WAN that allows for simultaneous transmission of overlapping waveforms. That is the loophole. Recall that the Shannon-Hartley Theorem is applicable only to a single link.

The result is that for RPMA, there is no capacity benefit for an individual link being extremely spectrally inefficient. RPMA can use extreme variable spreading (or processing gain), result in an extremely sensitive modem (-145 dBm is best in the world), and not be hit with any capacity penalty.

RPMA can demodulate 1,000 signals simultaneously for a couple of fundamental reasons:

Transmit power control is essential. Signals must arrive at the access point at very similar power levels for this multiple access capability to exist.
There must be a mechanism such that the received waveforms can be individually resolved. Any individual waveform may not be correlated with any other waveform, or neither waveform will be resolvable. That is exactly what the RPMA access scheme was designed to do.

RPMA satisfies both of these requirements; LoRa only partially satisfies the second criteria. LoRa signals at different data rates have a chance of both being resolvable which results in incremental capacity benefit (<30%), not the game-changing 1000x advantage that RPMA enables. Note that RPMA satisfies these requirements without the need for protocol overhead. This is a big differentiator from cellular technologies such as LTE (and the LPWA offshoots like NB-IOT) and I will discuss this in more depth in a future blog.

As detailed in Appendix A-C of How RPMA Works: The Making of RPMA, there are other significant factors that influence the total amount of data that can be pushed through the spectrum (η_goodput)

Amount of Media Access Control (MAC) protocol overhead such as retransmissions and headers.
The effect of transmissions that are heard by multiple access points/base stations (known as other cell interference) and the corresponding reduction in capacity.

Technology	Spectral Efficiency Single Link η _link	Spectral Efficiency Multiple Access η _MA	Spectral Efficiency Goodput η _goodput
RPMA	0.00006 to 0.008	0.03	0.0026
Sigfox	1	0.013	0.00019
LoRa (US)	0.0049 to 0.047	0.00057	0.000013
LoRa (Europe)	0.0049 to 0.0015	0.00017	0.000019

And that’s not all! η_goodput is a per Hertz of bandwidth number, it still needs to be multiplied by the amount of spectrum available! Since RPMA is at 2.4 GHz where 80 MHz is available, we can use significantly more spectrum that the fractured, regional subGHz bands that the Sigfox technology and LoRa use.

Technology	Spectral Efficiency Goodput η _goodput	Available Spectrum (MHz) BW	Total Goodput (bps) η _goodputxBW
RPMA	0.0026	16	41,000
Sigfox	0.00019	0.2	37
Lora (US)	0.000013	12.8	170
Lora (Europe)	0.000019	1.6	30

As you can see, RPMA has a massive advantage in the resultant goodput. That means a piece of RPMA network infrastructure (typically on a tower or rooftop) will be capable of servicing orders of magnitude more devices and devices with far greater data needs. We’ll crunch some numbers on this in Blog 5 – The Economics of Receiver Sensitivity and Spectral Efficiency.

At any point, if you’re interested in a more in-depth treatment of these topics, please download, How RPMA Works: The Making of RPMA.

July 22, 2016

Chirp Spread Spectrum: The Jell-O of Non-Coherent M-ary Modulation

We have asserted textbook style categories for Low-Power, Wide-Area (LPWA) connectivity Blog 1: Categories of LPWA Modulation Schemes. Then we went back to basics to get some framework and vocabulary in place to discuss these various schemes in Blog 2: Back to Basics – The Shannon-Hartley Theorem, where we introduced the critical concepts of spectral efficiency (η), Eb/No, and how a point of diminishing returns can be reached in the quest to reach the Shannon Limit.

Based on this framework, what I would like to discuss here is why Chirp Spread Spectrum (CSS) can be treated more generically as Non-Coherent M-ary Modulation (NC-MM). The title of this blog likely makes you ask what could CSS possibly have to do with Jell-O? With brilliant marketing, Jell-O has eclipsed the more generic terminology of “gelatin” defined as “protein obtained by boiling skin, tendons, ligaments, and/or bones with water.” Non-Coherent M-ary Modulation (NC-MM) is analogous to “gelatin”; LoRa™(CSS) is analogous to “Jell-O”. It is actually a compliment to the marketing around LoRa (CSS). As you can imagine, if you want to study the science behind Jell-O, you would have better luck starting with “gelatin”. The same is true with understanding LoRa(CSS).

Category

Local Area Network

Wide Area Network

Ultra Narrow Band

(UNB)

Sigfox

Telensa

N-Wave

WaveIOT

Non-Coherent M-ary Modulation (NC-MM)

Bluetooth

802.11b

LoRa (CSS)

Sensus (7-FSK)

GSM/GPRS

EC-GSM

Direct Sequence Spread Spectrum (DSSS)

802.11

Zigbee

W-CDMA

RPMA

Orthogonal Frequency Division Multiplexing

(OFDM)

802.11a/g

LTE

WiMax

NB-IOT

But first, we must justify why we’re lumping CSS into the NC-MM category. There are two parts to the argument:

It’s the same math. The mathematical operations for demodulation are identical to other well-known NC-MM waveforms – particularly M-ary FSK.
If it looks like a duck, swims like a duck… CSS has the same performance as other NC-MM waveforms (again particularly M-ary FSK), and that is further proof that it’s best to treat CSS more generically as NC-MM.

I will back up the two arguments below.

It’s the Same Math

Please take a look at the figure below – it is the demodulation engine of 2-ary FSK. If you added M-2 more elements, this could be turned into a demodulation engine for M-ary FSK. And, most critically, if the set of waveforms in the figure Φ are sinusoids, it is an FSK demodulator, and if Φ is an orthogonal set of chirp waveforms, it’s CSS. As long as the set of waveforms Φ are an orthogonal set, the performance of this demodulator will be 100% identical regardless of the particular orthogonal set Φ.

blog3 math

If it Looks Like a Duck, Swims Like a Duck…

Beware the straw man argument! Taking a look at the document entitled “LoRa Modulation Basics”, they do a comparison of the LoRa modulation (also known as CSS) against FSK and make the claim of an impressive sensitivity improvement of 15 dB for the SF = 12 LoRa mode over FSK (There is an obvious typo in the equivalent bit rate column in SFs 6 through 9 where the data rate is expressed in b/s as opposed to the stated kb/s.)

First, we need to talk about the relationship between SF and spectral efficiency. For un-coded Non-Coherent M-ary Modulation (NC-MM) the expression for η = (SF)2^-SF (in other words a non-coherent symbol yields SF bits of information and that non-coherent symbol is of duration 2^SF.) If channel coding as added into the mix (like in the LoRa waveform), this expression becomes η = R(SF)2^-SFwhere R is the code rate (typically rate ½).

blog 3 table On face value, that makes the claim that the math is the same rather dubious. However, as discussed in Blog 2: Back to Basics – The Shannon-Hartley Theorem, there are several reasons why similar modulations may achieve different sensitivity.

Increased data rate means reduced receive sensitivity even with the same spectral efficiency (η).
Increased spectral efficiency means reduced Eb/No which reduces sensitivity.
Almost all modern communication theory approaches include channel coding. An apples-to-apples comparison would be consistent relative to whether channel coding was implemented. A very popular and effective channel coding is convolutional encoding coupled with a Viterbi decoder (invented and named after the same individual that is on our advisory board).

As it turns out, relatively higher data rate, spectrally efficient, un-coded 2-ary FSK is fairly easy to beat up on in terms of sensitivity. A more accurate comparison point would be the coded 7-ary FSK that a company called Sensus uses. The Sensus flavor of FSK is more of an apples-to-apples comparison because 7-ary FSK has a lower spectrally efficiency (η) which allows for better Eb/No.

Mode	Coding	Equivalent bit rate (kb/s)	Eb/No	η	Δ (dB)
7-ary FSK	Rate ½ (likely)	8	9	0.2	0.0
2-ary FSK	No	1.2	17.2	1	-8.2
LoRa SF = 12	Rate ½	0.293	8.3	0.0015	0.6
LoRa SF = 11	Rate ½	0.537	8.2	0.0027	0.8
LoRa SF = 10	Rate ½	0.975	8.1	0.0049	0.9
LoRa SF= 9	Rate ½	1.757	8.6	0.0088	0.4
LoRa SF = 8	Rate ½	3.125	9.1	0.016	-0.1
LoRa SF = 7	Rate ½	5.468	9.6	0.027	-0.7
LoRa SF = 6	Rate ½	9.375	12.3	0.047	-3.3

Thus, from a fundamental technology viewpoint, the well-marketed LoRa (CSS) is really no different than the Sensus 7-FSK approach that has been around in the utility space for years. The Sensus technology is known to be very low capacity, typically appropriate for meter billing reads only, and requiring licensed spectrum to operate robustly. The Sensus technology is certainly not a WAN as we define it.

In the table above, notice many of the LoRa modes have η << 0.01. Is that bad in terms of capacity? Not necessarily, is the general answer. What’s not explored in the Shannon-Hartley theorem is the ability to send simultaneous links over the same bandwidth. That is potentially a bit of a “loophole” in the theory. Does LoRa leverage this “loophole”? What about RPMA? I’ll explore that in more detail in Blog 4: “Spreading” – A Shannon-Hartley Loophole?.

And finally, I’ll get back into the real world and test whether this discussion is even useful in Blog 5: The Economics of Receiver Sensitivity and Spectral Efficiency to attempt to quantify the commercial value of that which is discussed on these blogs.

At any point, if you’re interested in a more in-depth treatment of these topics, please download, How RPMA Works: The Making of RPMA.

July 21, 2016

Back to Basics: The Shannon-Hartley Theorem

The Shannon-Hartley Theorem represents a brilliant breakthrough in the way communication theory was viewed in the 1940s and describes the maximum amount of error-free digital data that can be transmitted over a communications channel with a specified bandwidth in the presence of noise.

As you can see, and as threatened in Blog 1: Categories of LPWA Modulation Schemes, we’re going back to basics and generating the framework and vocabulary to justify and discuss the categories of Low-Power, Wide-Area (LPWA) connectivity: Ultra-Narrow Band (UNB), Non-Coherent M-ary Modulation (NC-MM), Orthogonal Division Multiple Access (OFDM), and Direct Sequence Spread Spectrum (DSSS).

Before jumping directly into the details of the Shannon-Hartley Theorem, lets set the stage with the concepts of receiver sensitivity, Eb/No, and spectral efficiency (η).

Receiver Sensitivity and Eb/No

Receiver sensitivity is the amount of energy required for demodulation. A link with better receiver sensitivity will have longer range and more robustness, which will equate to less density of infrastructure.

Eb/No is energy per bit relative to thermal noise spectral density. To understand Eb/No, its important to understand the relationship between receiver sensitivity (related to range), data rate, and Eb/No:

Receiver sensitivity = (-174 dBm + NF) + Eb/No + 10log₁₀(Data Rate)

The number -174 dBm is based on Boltzman’s constant and represents the thermal energy at room temperature per Hz.
NF is Noise Figure, which is a figure of merit that is typically around 2 to 5 dB for a given radio front end.
There is an inverse linear relationship between receiver sensitivity and data rate. The better the Eb/No, the better the receiver sensitivity at a given data rate.

Spectral Efficiency

Spectral efficiency is the amount of data that can be transmitted in a single link per Hz of bandwidth. So why is spectral efficiency important? It’s because spectrum is typically a fixed and valuable asset. For a multiple access system to operate well, it’s often important to channelize so that multiple communications may occur simultaneously. The formula for the amount of data throughput that can be moved through a piece of spectrum is η x BW. BW is typically fixed – if it’s free spectrum, there is a very finite amount of it, if it’s licensed spectrum, then it’s extremely expensive. The role of a good modulation scheme is then to make sure that η is as high as reasonably can be designed.

The Shannon-Hartley Thoerem

The Shannon-Hartley theorem describes the theoretical best that can be done based on the amount of bandwidth efficiency: the more bandwidth used, the better the Eb/No that may be achieved for error-free demodulation. Or, equivalently stated: the more bandwidth efficient, there is a sacrifice in Eb/No. The Shannon-Hartley theorem applies only to a single radio link. It does not directly apply to a multiple access system that is important in a Wide-Area Network (WAN). Nevertheless, Shannon-Hartley is a great starting point.

Specifically, the Shannon-Hartley theorem puts a lower bound on the Eb/No for error-free demodulation given spectrum efficiency as^{^[1]}:

equation

where η is spectral efficiency measured in units of bits/Hz. This relationship is shown in the figure below. Note that what’s called the Shannon Limit (which is -1.6 dB) is asymptotically approached when η = 10^-1. As spectrum efficiency is increased, the required Eb/No increases. Stated another way, as η increases, not only is there more power required (or worse receiver sensitivity) because you’re moving more bits, but each bit actually gets more expensive in terms of power required. That’s a double whammy in terms of receiver sensitivity.

channel capacity

So, let’s be honest, the Shannon-Hartley Theorem is hardly good news for anyone. Let’s walk through what it means when you’re trying to build a system in a constrained amount of spectrum:

Say you want to cover as much distance with your link as possible. Shannon-Hartley tells you that you can reduce data rate to get better range (in theory without limit). At this limit, it costs a fixed amount of power to get a bit through – so every dB of data rate reduction buys you 1 dB of receive sensitivity. But, the more you do this, the less the capacity. You’ll be getting a trickle of data through this admirable amount of spectrum you happen to have access to.
Okay fine, you say. Then let me just get a lot of data through my spectrum (that’s a perspective that is important in the cellular industry). If it’s a fixed amount of power per bit, then I’ll crank up the power and get a factor of 10 data rate increase for every factor of 10 dB increase in transmit power. But, not so fast, says Shannon-Hartley. If you look at the curve, the amount of power per bit increases dramatically as η becomes larger than 1. Not only are you spending power getting more bits through your spectrum, each bit is getting considerably more expensive. Nevertheless, in the 3GPP standards body, the quest for ever-increasing data rates has caused pushing up this curve as much as possible. As an example, 256 Quadrature Amplitude Modulation (QAM) that has a spectral efficiency of close to 9 (η=9 which is pretty much the far right point in the figure above) is being currently standardized.
Oh, and just to add insult to injury – Shannon-Hartley is the best-case scenario. It’s what could be achieved with infinite computation power, infinite latency (because the coding blocks become infinitely large), infinite frequency accuracy, etc. Everything in the real world will be worse, perhaps even much worse.

What do smart people do when confronted with such bad news? How about look for a loophole! We definitely need such a loophole to build a high coverage, yet high capacity network (a trickle won’t do). Something called “spreading” may be that loophole. “Spreading” is maintaining the channel bandwidth (BW) regardless of data rate and allowing very low link spectral efficiencies (η <<0.1) to achieve high receiver sensitivity. On the face of Shannon-Hartley, this seems like an exceedingly bad idea. Yeah, you may get the coverage you need, but you only get a “trickle” of small amounts of data through that precious spectrum. That’s okay, a lot of loopholes seems like a bad idea at first and we will explore the implications of this approach in Blog 4: “Spreading” – A Shannon-Hartley Loophole?

But first, because there are multiple ways to “spread” a signal, we need to first understand how LoRa/Chirp Spread Spectrum (CSS) can be analyzed and we lay that framework in Blog 3: Chirp Spread Spectrum (CSS): The Jell-O of Non-Coherent M-ary Modulation.

And finally, we’ll get back into the real world and test whether this discussion is even useful in in Blog 5: The Economics of Receiver Sensitivity and Spectral Efficiency to attempt to quantify the commercial value of that which is discussed on these blogs.

At any point, if you’re interested in a more in-depth treatment of these topics, please download, How RPMA Works: The Making of RPMA.

July 20, 2016

Setting the Stage: Categories of Low-Power, Wide-Area Modulation Schemes

At first glance, the number of communication technologies being discussed for Low- Power, Wide-Area (LPWA) networks may be a bit overwhelming. What may be helpful is to look at the underlying technology from a fundamental perspective and tune out the marketing component. Note that marketing is an extremely important aspect of LPWA success, but since I am a CTO, I would like to push it to the side for this discussion.

Many of you know that Communication Theory is a very mature field going back many decades with a vast wealth of generated knowledge. Tens of thousands of books, articles, and papers have been published over this time. There are giants in the field – Claude Shannon, Harry Nyquist, Ralph Hartley, Alan Turing, and Andrew Viterbi (who has been an Ingenu strategic advisor from the beginning) whose work we can turn to for clarity. This great body of work gives us frameworks and vocabularies for comparison. It’s often a drier and less interesting world once the marketing innovation is subtracted out – but please, bear with me.

The table below shows four categorizations of the various categories of modulation schemes that are being discussed for LPWA. Bold denotes those technologies being branded as applicable to Low-Power, Wide-Area networking. Since this is a technology treatment, I am defining Local Area Network (LAN) and Wide-Network (WAN) by the underlying technology as opposed to how these approaches are being marketed. Note that the most well-known approaches in each category are the Sigfox™ technology, LoRa®, also known as Chirp Spread Spectrum (CSS), Narrow-Band IOT (NB-IOT), and Random Phase Multiple Access (RPMA®). These technologies tend to be those with the best marketing (yes, marketing is very important).

Category

Local Area Network

Wide Area Network

Ultra-Narrow Band

(UNB)

Sigfox

Telensa

N-Wave

WaveIOT

Non-Coherent M-ary Modulation (NC-MM)

Bluetooth

802.11b

LoRa (CSS)

Sensus (7-FSK)

GSM/GPRS

EC-GSM

Direct Sequence Spread Spectrum (DSSS)

802.11

Zigbee

W-CDMA

RPMA

Orthogonal Frequency Division Multiplexing

(OFDM)

802.11a/g

LTE

WiMax

NB-IOT

From a technology perspective, the definition of being appropriate as a WAN is whether the multiple-access considerations of coverage and capacity are taken into account:

Coverage. If you want to build a WAN, you would like a single piece of network infrastructure (often on a tower or rooftop) to cover as much area as possible.
Capacity. There’s not a lot of good in covering a massive area if you cannot support the data needs of all the devices in that footprint.

Giving a bit more color on the categories:

Ultra-Narrow Band (UNB). The reason many companies have elected this approach is the advantage of low barrier to entry. Companies in this category can leverage commodity radios and skip any technology development. These companies tend to argue that no new technology is required. We disagree for many reasons including the inability to make the economics of LPWA work as discussed in Blog 5: The Economics of Receiver Sensitivity and Spectral Efficiency.
Non-Coherent M-ary Modulation (NC-MM). This is a commonly used modulation in both LAN and WAN applications. Cellular 2G technology was based on GSM/GPRS which uses a modulation approach called Minimum Shift Keying (MSK) and is also being repurposed to Extended Coverage GSM (EC-GSM) which is cellular LPWA in 2G spectrum. The LoRa modulation (CSS) is a member of this category (as is justified in Blog 3: Chirp Spread Spectrum: The Jell-O of Non-Coherent M-ary Modulation. The “spreading” of CSS has no discernible advantage and indeed, as discussed in Blog 4: “Spreading” – A Shannon-Hartley Loophole?, has some significant drawbacks in terms of spectral efficiency.
Direct Sequence Spread Spectrum (DSSS). We described LoRa as “spreading” for no discernible reason. Well it turns out NC-MM does not have the monopoly on this. DSSS has a couple of technologies that also spread for no discernible reason – IEEE 802.11 (the original 1 and 2 Mbps data rates) and Zigbee (based on IEEE 802.15.4). This is just one example to show you that standards bodies are less about technology and more about politics. I will discuss this in more depth in a future blog.
Orthogonal Division Multiple Access (OFDM). This is the way you get extreme spectral efficiency. It’s great for voice high-speed data and has enabled LTE (4G) to become the dominant cellular standard. When you try to point this approach at LPWA (such as NB-IOT), significant problems emerge. I will discuss this in more depth in a future blog.

In the next blog, we visit the building blocks of communication and translate the perfectly good and intuitively understandable terms of coverage and capacity to nerd in Blog 2: Back to Basics – The Shannon-Hartley Theorem.

Coverage translates to receive sensitivity, which is a function of something called Eb/No (energy per bit relative to thermal noise spectral density).
Capacity translates into something called spectral efficiency and we need to go even one level deeper into nerd and assign it a Greek letter (of course) and that Greek letter is…. η.

Using this fundamental framework, in Blog 3: Chirp Spread Spectrum: The Jell-o of Non-Coherent M-ary Modulation, I’ll talk about a category of approaches that is very similar from a technology point of view and one very successfully marketed approach in this category called Chirp Spread Spectrum (also known as LoRa).

And while we’re on the subject of spreading, I’ll discuss how there a potential capacity loophole to take advantage of in Blog 4: “Spreading” – A Shannon-Hartley Loophole?.

And finally, let’s get back into the real world and test whether this discussion is even useful in Blog 5: The Economics of Receiver Sensitivity and Spectral Efficiency to attempt to quantify the commercial value of that which is discussed on these blogs.

At any point, if you’re interested in a more in-depth treatment of these topics, please download How RPMA Works: The Making of RPMA.

Is the cellular standard roadmap the answer to Low Power Wide Area Connectivity? Pt. 1

Part One: Introduction

The “WA” in LPWA is Wide Area… and Pizza

The Economics of Receiver Sensitivity and Spectral Efficiency

“Spreading” – A Shannon-Hartley Loophole?

Chirp Spread Spectrum: The Jell-O of Non-Coherent M-ary Modulation

Back to Basics: The Shannon-Hartley Theorem

Setting the Stage: Categories of Low-Power, Wide-Area Modulation Schemes

Ted’s Corner

Recent Posts

Archives

Categories

Tags

JOIN THE INGENU NEWSLETTER