One often finds an impression about paging, relative to other forms of wireless communication available now (say cellular in any of its forms, or WiFi), in which paging is viewed as an old and dead technology. It is “narrowband” in a world of increasingly “broadband” communication methods. A bigger, broader data pipe is almost universally seen as superior.
On this blog, I adopt the contrarian point of view. Having worked for decades both in cellular and paging, I think that I can present a more balanced point of view than others. I believe that it is extremely rare to find anyone who has worked at any real technical depth in both the cellular and paging sectors. In my personal experience, folks in cellular look at paging with very real contempt. In contrast, folks in paging are seriously jealous of the huge market that the cellular industry has developed since the early 1980s, while observing the collapse of their markets especially since 2001.
It is extremely unfortunate, but much of what has proven successful in the paging and Narrowband PCS sector will have to be relearned by the cellular industry, especially as this sector branches into more data oriented, machine to machine communication.
So, you might ask, what are the fundamental differences between paging/NPCS and cellular/WPCS? Here is a short list:
- forward channel simulcast versus point to point
- reverse channel diversity at several tiers versus point to point
- high SNR per bit versus low CIR per bit, by design
- connectionless store and forward versus connection-oriented
- non-linear modulation versus linear modulation
These 5 fundamental differences have a set of consequences for the quality of service, capacity, battery life, and costs of the two forms of wireless transmission. Of course, the quality of service, capacity, battery life, and cost results yield services that may be better delivered using one or the other form of wireless network.
At the time I’m writing this, the story of paging versus cellular is pretty much cast in stone. Rather like the story of the replacement of steam locomotives by diesel-electric locomotives just after WWII, not every rail application was better served by the new technology. Nonetheless, in the 10 or so years from 1945 to 1955 there was a dramatic displacement of steam by diesel-electric. It is difficult now to find a steam locomotive in active commercial service anywhere outside of a historical park. Perhaps the historical parks of 100 years from now will feature set scenes from the 1980s in which visitors are given working pagers as they enter the gates.
Still, there are millions of paging/NPCS units in service; and the cellular industry, in its campaign to win the hearts and minds of everyone on the planet may, just may, be missing a few lessons learned by their competition.
Let’s start by considering simulcast. Early forms of digital paging broadcast at as low as a few 100 bit/s. Such a system was standardized in the UK by its Post Office Code Standardization Advisory Group (POCSAG), which had a base rate of 512 bit/s. Current “high speed” digital paging/NPCS schemes like FLEX™ and ReFLEX™ can run as high as 6400 bit/s on the forward channel; but this is still a drop in the bucket compared to the data rates of any digital cellular system. There is an essential reason why the data rates are so low in a paging system.
The answer is that exactly the same sequence of bits is transmitted on the forward channel in a highly synchronized manner by every base station in a serving area. In the simplest form of this scheme, a large urban market like New York, Los Angeles, London, Paris, Hong Kong or what have you, might involve nearly 100 transmitters sending out exactly the same forward channel information stream. Of course, this would yield an incredibly low system capacity if measured in terms of bit/s per square kilometer. A cellular system with the same number of transmitters would have a unique forward channel information stream on each antenna. Combining this feature with the ability to reuse allocated spectrum at anywhere from every 21st to every individual antenna in the system (as opposed to a simple paging system’s inability to achieve any spectrum reuse within a market at all) might make one wonder “why bother?”
The simplest answer in terms of a user experience of the service comes in the form of the signal to noise ratio per bit. A paging transmitter might operate a 300 W ERP. If it were sending at 512 bit/s, the bit time is nearly 2 ms. This yields an energy per bit as transmitted at around 590 mJoules. In contrast, a typical GSM system would transmit at around 35 W ERP at a bit rate of 300 kbit/s. The energy per bit as transmitted is around 120 microJoules. The difference is nearly 40 dB.
Now, I have met satcomm engineers who said they’d kill for 2 dB. Of course, they don’t have Rayleigh fading to contend with; but even in the context of land mobile, 40 dB gain is, put simply, huge. Further, since each transmitter in the system is sending exactly the same information via a non-linear modulation, say Frequency Shift Keying (FSK), if for any reason a nearby transmitter’s signal fades at the mobile receiver, the FM capture effect allows the receiver to obtain the bit energy from the next lowest transmitted signal in its vicinity. This produces a natural form of forward channel diversity that requires no computational or hardware overhead in the receiver at all. Nothing more complex than a limiter detector is required.
Because of this multi-channel diversity from base to mobile, the characteristic “contrast curve” that shows bit error rates versus signal to noise ratio for a POCSAG or low-speed FLEX simulcast system is very nearly approximated by a non-fading AWGN channel. Most cellular channels, because of their design points, would benefit relatively little from diversity. Instead, some form of coding gain is almost essential, together with its complexity. Of course, by the time any manufacturer actually comes around to building a cellphone, their design choices are extremely limited. The combination of the system standard (3G GSM for example) and the chip design for the build have pretty much dictated performance. One can play games with the antenna and perhaps introduce a gain block, but each of these design features have their trade-offs, and getting a dB or two with these components can introduce other problems along the way.
In summary, the combination of low bit rates, higher power per transmitter, simulcasting, and non-linear modulations give paging/NPCS schemes a huge quality of service benefit relative to cellular schemes. Within their serving areas, paging systems are much more likely to deliver an accurate copy of an intended message than a cellular system on the first attempt. Having made that claim, there are obvious limits. One would not attempt to send a 20-page PDF attachment to a paging subscriber; and this works reasonably well in a 3G GSM service to, say, an iPhone. That reliability benefit that simulcast delivers is constrained to relatively short messages, say less than 1000 octets or so. As a consequence, a broadband service can resend failed message packets several times over within the interval of time that a simulcast service needs in order to send a message just once. The question is then, do the coding gain and ARQ methods of cellular not trump the diversity gain of simulcast?
Apparently the answer is that cellular wins, at least in the view of a large segment of the market for wireless services. Apple can sell nearly 2 million iPhone 4s in the space of a few weeks, and this is around the total number of UIS on USA Mobility (YE2009, there were around 2.2MM UIS). However, there may yet remain services for which this simplicity in the mobile device is of paramount importance.
We’ll consider this, among other matters, in up-coming posts.