19

Network Software – Microdevices and Microdevice Networks – The Software of the Very Small

This is a chapter about very small things and the software that drives them. This includes very small things that when interconnected can become very large networks. In previous chapters we have discussed the evolving role that microelectrical mechanical systems are playing in cellular handset design and beyond that the role that nanoscale manufacturing may have in transforming material properties. The combination of silicon geometry scaling and microminiaturisation techniques together are transforming phone form factor and functionality.

There are, however, other devices that are acquiring new capabilities and shrinking both in size and the amount of energy they consume. These devices use manufacturing techniques at the scale of nanometres (one billionth of a metre) to build structures at atomic or molecular level that are combined in to devices that are either measured in micrometres (one millionth of a metre), or millimetres (one thousandth of a metre).

We review three classes of ‘microdevice’, ‘Memory Spots’ ‘motes’ (being developed in the US as a result of various ‘smart dust’ projects) and the Hitachi Mu chip (an ultraminiaturised RF ID device).

Two of these devices have failed to gain market traction for reasons that we need to analyse. All three device classes are, or were, intended to store information and/or collect information and/or provide identification. All of them are, or were, communication devices. A particular interest is to consider how we communicate with these communication devices and whether the cellular phone has a valid role in this communication process.

We compare these microdevice applications with a present larger form factor application (contactless smart cards) and suggest some technical and commercial commonalities. We discuss the present status of ‘phone to device’ communication systems, particularly RF systems and protocols and question whether there are plausible cellular phone and mobile broadband business models to justify development investment in this intriguing but challenging application sector. We highlight certain factors that suggest adoption time scales may be longer than commonly supposed.

19.1 Microdevices – How Small is Small?

MEMS and microminiaturisation techniques in combination with silicon geometry scaling are allowing us to deliver storage functionality, information gathering functionality, identification and communication capability in increasingly small devices.

Table 19.1 lists three form factors ranging from a grain of rice (small) to a grain of sand (very small) to a grain of dust (very very small). A grain of dust typically has a diameter of a few tens up to 200 or 300 micrometres, hence the term ‘microdevice’.

We review products developed for all three form factors. All three product sectors present distinct communication challenges and (hence) potential communication value opportunities.

On 14 March 2003 Hitachi announced a product called a Mu chip. It was very small (or else was photographed against a very big finger). This device is illustrated in Figure 19.1.

Actually it was 0.3 millimetres square, half the size of the RF tags on the market at the time and was manufactured using a 0.18-micrometre process as opposed to the 0.35-micrometre process generally in use then. The chip operated at 2.4 GHz and stored a 128-bit ID and could be read from a foot away (30 cm). The antenna, which cannot have been very big, was attached to two electrodes at the top and bottom of the chip, which apparently was very clever.

The Mu chip was intended as an alternative to bar codes or conventional RF tags for use in supply-chain management, product traceability and security and was packaged as a complete system including readers, software and networking infrastructure. RF tags at the time cost about 40 cents each.

In practice, the device never achieved traction either in the business and asset tracking market or in the consumer space and the web site is now closed.

Table 19.1 Microdevice form factors and functionality – three examples

Small	Very small	Very very small
Grain of Rice < 4 millimetres	Grain of Sand <2 millimetres	Grain of Dust <500 micrometres
Example Product
Memory Spots1 (HP Labs)	Motes (Intel web site)2	Mu Chip3 Hitachi
Information delivery device	Information collection device	Identification device (RFID)
2 mm by 4 mm including integrated antenna 256 kilobit to 4 megabit storage	1 mm by 1 mm (mote sized) Typically with sensing capabilities – temperature, light, vibration, acceleration, air pressure	0.4 mm by 0.4 mm by 60 micrometre 0.15 mm by 0.15 mm by 7.5 micrometre
Data transfer rate up to 10 Mbits/s	Data transfer rate Device and application dependent	Data transfer rate 12.5 kbits/s
Distance – close coupled	Distance – 20–30 metres	Up to 400 mm

On 17 July 2006 HP announced ‘a grain sized chip that could be attached to almost any object, making information more ubiquitous’. The device was called the Memory Spot and is illustrated in Figure 19.2.

Here is the launch press release:

HP Unveils Revolutionary Wireless Chip that Links the Digital and Physical Worlds

PALO ALTO, Calif., Jul 17, 2006

HP today announced that its researchers have developed a miniature wireless data chip that could provide broad access to digital content in the physical world.

With no equal in terms of its combination of size, memory capacity and data access speed, the tiny chip could be stuck on or embedded in almost any object and make available information and content now found mostly on electronic devices or the Internet.

Some of the potential applications include storing medical records on a hospital patient's wristband; providing audio-visual supplements to postcards and photos; helping fight counterfeiting in the pharmaceutical industry; adding security to identity cards and passports; and supplying additional information for printed documents.

The experimental chip, developed by the “Memory Spot” research team at HP Labs, is a memory device based on CMOS (a widely used, low-power integrated circuit design) and about the size of a grain of rice or smaller (2 mm to 4 mm square), with a built-in antenna. The chips could be embedded in a sheet of paper or stuck to any surface, and could eventually be available in a booklet as self-adhesive dots.

“The Memory Spot chip frees digital content from the electronic world of the PC and the Internet and arranges it all around us in our physical world,” said Ed McDonnell, Memory Spot project manager, HP Labs.

The chip has a 10 megabits-per-second data transfer rate – 10 times faster than Bluetooth™ wireless technology and comparable to Wi-Fi speeds – effectively giving users instant retrieval of information in audio, video, photo or document form. With a storage capacity ranging from 256 kilobits to 4 megabits in working prototypes, it could store a very short video clip, several images or dozens of pages of text. Future versions could have larger capacities.

Information can be accessed by a read-write device that could be incorporated into a cell phone, PDA, camera, printer or other implement. To access information, the read-write device is positioned closely over the chip, which is then powered so that the stored data is transferred instantly to the display of the phone, camera or PDA or printed out by the printer. Users could also add information to the chip using the various devices.

“We are actively exploring a range of exciting new applications for Memory Spot chips and believe the technology could have a significant impact on our consumer businesses, from printing to imaging, as well as providing solutions in a number of vertical markets,” said Howard Taub, HP vice president and associate director, HP Labs.

The chip incorporates a built-in antenna and is completely self-contained, with no need for a battery or external electronics. It receives power through inductive coupling from a special read-write device, which can then extract content from the memory on the chip. Inductive coupling is the transfer of energy from one circuit component to another through a shared electromagnetic field. A change in current flow through one device induces current flow in the other device.

Memory Spot chips have numerous possible consumer and business-based applications.

Some examples are:

• Medical records: Embed a Memory Spot chip into a hospital patient's wrist band and full medical and drug records can be kept securely available.
• Audio photo: Attach a chip to the prints of photographs and add music, commentary or ambient sound to enhance the enjoyment of viewing photos.
• Digital postcards: Send a traditional holiday postcard to family and friends with a chip containing digital pictures of a vacation, plus sounds and even video clips.
• Document notes: A Memory Spot chip attached to a paper document can include a history of all the corrections and additions made to the text, as well as voice notes and graphical images.
• Perfect photocopies: A Memory Spot chip attached to a cover sheet eliminates the need to copy the original document. Just read the perfect digital version into the photocopier and the result will be sharp output every time, no matter how many copies are needed, and avoiding any possibility of the originals jamming in the feeder.
• Security passes: Add a chip to an identity card or security pass for the best of both worlds – a handy card with secure, relevant digital information included.
• Anticounterfeit tags: Counterfeit drugs are a significant problem globally. Memory Spot chips can contain secure information about the manufacture and quality of pharmaceuticals. When added to a drug container, this can prove their authenticity. A similar process could be used to verify high-value engineering and aviation components.

Like the Hitachi Mu chip, the Memory Spot never gained mass-market traction, but both products were precursors of things to come. Sometimes ideas just get to market too early but that does not mean that these devices were not important or no longer have relevance.

Both the Memory Spot (HP Labs) and Mu chip (Hitachi) products were passive, generating power inductively from the interrogating device.

As stated in the press release the Memory Spot was a CMOS device with an intended storage capacity of between 256 kilobits and 4 megabits and a claimed transfer rate of 10 megabits per second. The device had a built-in antenna and could be embedded on a sheet of paper or any suitably friendly surface. Suggested consumer applications included adding a memory chip to photographs or postcards or books to provide an audio (voice or music) or imaging (still and video imaging) or extended text download capability (microdot applications). As such, it could be described as an ‘information storage and delivery device’. The public announcement of the device created substantial interest.

In the middle ‘grain of sand’ category we have the various ‘mote devices’, also sometimes described as ‘smart dust devices’ with their genesis in US projects aimed at realising an autonomous sensing computing and communication system within a cubic millimetre (a ‘mote’), though present devices are significantly larger.

The idea of these devices is that they can be scattered across an area and form a self-organising network of interconnected interactive objects that are either battery powered, solar powered or vibration powered. Their purpose is primarily to capture information on the physical world, for example temperature, ambient light, vibration, acceleration and air pressure – they are information-collection devices.

However, the question we are trying to answer is whether we want or need cellular phones to be able to talk to these microdevices? If we do, what are the technology, engineering and business challenges and opportunities? Why did the MU chip and Memory Spot fail to make market headway and do they have contemporary equivalents?

19.2 Contactless Smart Cards at 13.56 MHz – A Technology, Engineering and Business Model?

A starting point is to look at a (larger) device that we already talk to, the contactless smart card, and see what lessons we might learn. Contactless smart cards are similar to the Memory Spot in that they are typically close-coupled applications in which the two devices (the reader and the smart card) are either touching or within a few millimetres of each other. This is ‘near-field communication’.

There is an international standard ISO14443 for contactless smart cards, also known as RF tags, operating at 13.56 MHz, close to the GSM clock reference at 13 MHz. The devices are ASK or BPSK modulated, either passive (load modulation) or active with a range from 0 to 200 millimetres. There are 4 types of contactless smart card (tag) categories determined by memory footprint and transfer speed detailed in Table 19.2.

Table 19.2 Contactless smart cards

In the UK, every time you use a London Train and Underground Oyster Card4 the 13.56-MHz transmitter in the oyster terminal device at the turnstile is irradiating your ‘smart’ oyster card with RF energy, your oyster card then uses this energy to transmit your ID back to the device.

In 2006, when we first researched this topic there was no compelling technical reason why this function could not be included as a mobile phone function. It could be passive or active and unidirectional or bidirectional. If bidirectional, the phone could read information stored at the turnstile (time tables, delays, special travel offers) and of course the cost of any transaction would just be added to the monthly phone bill.

Five years on, NFC-enabled handsets are available and applications are finally beginning to emerge. The lesson here is that fast innovation adoption only happens when there are tangible benefits to all parties in the value chain. In this example, there is clearly user value in terms of convenience but the mobile-phone operator has become an intermediary in a transaction that has cost and risk with income going to a third party (the train company).

In the end, if there is a genuinely compelling user experience benefit then the innovation will happen, but much more slowly than people generally appreciate. The same lag effect meant that the Memory Spot and Mu products were never a commercial success. The devices potentially created user experience value but needed to be networked to realise that value and the incentive for third parties to invest in those networks was insufficient. The device business model worked. The network business model failed.

Anyway, back to Oyster cards. In practice, we still use dedicated smart cards not cellular phones to access these systems. This is because we do not have NFC transponders as a standard item in phones. The reason we do not have NFC transponders as standard items in phones is that it adds cost and the associated value model depends on having commercial agreements in place, which require crossindustry consensus and therefore take time to negotiate.

19.3 Contactless Smart Cards and Memory Spots – Unidirectional and Bidirectional Value

The Memory Spot in some ways represented a development of the contactless smart card business model but with the focus on information transfer rather than transaction facilitation. This implied a need for higher data rates (at 2.4 GHz) than those available using NFC (at 13 MHz). Although the Memory Spot was a contactless smart memory information dispensing device, the provision of information and transactional value are closely related. Reading about things prompts us to buy things.

This comes down to the simple principle that a unidirectional exchange has a certain value. Uploading our ID to a turnstile has convenience value, reading a Memory Spot would have had interest and information value. If the exchange can be made bidirectional then the value increases.

19.4 Contactless Smart Cards, RF ID and Memory Spots

The second microdevice example, the Hitachi Mu chip, was intended to function as an RF bar code and/or RF ID device. RF bar codes offer some advantages over conventional bar codes, for example nonline-of-sight and multiple read capability and additional address bandwidth to support electronic product codes rather than standard (universal) product codes. In 2006, they were also beginning to be used in passport systems though that proved complicated as well. For consumer and retail applications the problem remained one of cost. Even a few cents matters if it's a can of beans that you are trying to sell.

Traditional printed label bar codes, however, are effectively zero cost and have benefited from years of intensive investment in optical scanning techniques.

RF ID tags also have competition from other offerings such as long-wave magnetic systems5 operating at wavelengths below 450 kHz. These system options have their own standard (IEEE P1902 Standard for Long Wavelength Wireless Network Protocol) and some application advantages, for example the ability to work underwater and/or underground.

As with RF tags, magnetic tags can be active or passive. The active devices have a claimed life of 10 years or more from a (coin-sized) lithium battery. Read rates are typically between 300 and 9600 bits/s.

RF ID tags and long-wave magnetic tags both have sufficient address bandwidth to support IPV4 or IPV6 addresses. This suggests a shift in functionality beyond present ‘visibility network’ or ‘visible asset’ applications to a broader application base in which wireless interrogation of an RFID prompts the interrogating device to access a web page.

To place this in the context of our Oyster card and Memory Spot examples, we do not necessarily need to download information from an embedded device other than an IP address that is then used to access information from a supporting web site.

In 2006 and 2007 the IEEE coordinated substantial standards work6 on these long-wavelength device systems but five years on ubiquitous applications have failed to emerge. Probably an example of a solution to a problem that didn't really exist.

19.5 Contactless Smart Cards, RF ID, Memory Spot and Mote (Smart Dust) Applications

Our third microdevice example, mote-sized smart dust devices, seem very different. These devices are intended to be deployed as self-configuring ad hoc networks that interact with each other. These devices can potentially all have IP addresses and can be a part of a sensing and surveillance network. They combine local storage capability, intelligence and communications capability.

The relevance of these networks to cellular phones may seem tenuous. However, the protocols for ad hoc networking are well established and already deployed in a number of two-way radio system solutions so the concept of cellular phones and mobile broadband devices interacting with these devices is at least plausible.

19.6 The Cellular Phone as a Bridge Between Multiple Devices and Other Network-Based Information

The logic of using a cellular phone in any or all of the above applications is that the cellular phone provides a bridge to the outside world. Advantageously, of course, this is a toll bridge with an efficient and robust revenue capture (billing) capability including pre- and postpayment collection.

The technical challenge of using a cellular phone is that we need to support additional wireless systems and protocols that add cost and complexity to an already overloaded product platform.

19.7 Multiple RF Options

Frequencies used presently include long wave (below 500 kHz), 1.95, 3.25, 4.75 and 8.2 MHz (typically used for antishop-lifting tags in retail stores), near-field communication in the 13 MHz ISM (industrial scientific medical) band, the 27-MHz ISM band, the 430–460-MHz ISM band available in Region 1 (Europe and Africa), the 902–916-MHz band available in Region 2 (North and South America), the 918–926-MHz band used for RF ID in Australia, the 2.35–2.45-GHz ISM band, a possible band at 5.4–6.8-GHz and/or the use of ultrawideband systems between 3 and 10 GHz.

UWB was potentially appealing in that there was a very scalable relationship between bit range and range (low bit rate/long range, high bit rate/short range). The frequency band (up to 10 GHz) is also useful for microdevice form factors. However, there was a massive standards battle combined with a failure to establish a consistent spectral policy interregionally, so UWB never made it to the market.

There does now seem to be a regulatory consensus that there should be three internationally agreed universal allocations for device-to-device communication, a low-frequency allocation at 125 kHz, the NFC allocation at 13.56 MHz and the 2.45 GHz allocation. These devices need to coexist with the other radio systems within the phone including wide-area cellular and local area (Bluetooth and WiFi) and receive only functions such as GPS.

19.8 Multiple Protocol Stacks

Standardised radio protocol stacks include Bluetooth7 (optimised for throughput) and ZigBee8 (optimised for low power consumption), both of which have well-developed vendor support. There are additionally operating systems optimised for low power consumption such as Tiny OS9 and a range of specialised proprietary offerings.

19.9 Adoption Time Scales – Bar Codes as an Example

Contactless smart cards and each of the microdevice options referenced in this chapter (HP Memory Spot, smart dust motes and micro-RFID) all offer or offered intriguing and potentially compelling cellular phone device and system-level integration opportunities.

They are, or were, technically feasible and supported by radio standards and protocol stacks that are relatively mature. The devices do not necessarily need to reduce in size but will increase in functionality and reduce in cost, thus broadening their application profile. However, business models may take longer to evolve than might be expected. Traditional bar codes provide an example.

The beginning of bar codes as we know them today can be traced back to a patent for ‘a Classifying Apparatus and Method’ filed in 1949 by Bernard Silver and Norman Woodland. Silver and Woodland were graduates at the Drexel Institute of Technology and were responding to a local food store's request for an automated method of reading product information.

It took twenty five years for the idea to evolve and for standards to be agreed, mostly getting agreement on the Universal Grocery Products Identification Code that evolved into the Universal Product Code. The first UPC scanner was installed in a supermarket in Ohio in 1974. The first product to have a bar code was a packet of Wrigley's gum.

Bar codes are now ubiquitous, but it has taken another 25 years for this to happen. Bar codes have taken over 50 years to become universally adopted. Hitachi and HP could not afford to wait that long.

19.10 Summary

Optical bar codes provide an example of a product that is now an integral part of everyday life. The enabling idea was simple but required an enabling technology (optical laser scanning) to be available together with universal application standards in order to support ubiquitous deployment.

Today we have a new generation of MEMS-based enabling technologies that are allowing us to build supersmall devices that can collect and store information and/or perform identification or labelling tasks that help us to interact more efficiently with the physical world around us.

It is tempting to position the cellular phone as the ‘device of choice’ for communicating with this new generation of microdevices. The technical logic is that it is relatively easy to extend the present communications systems within the phone to include phone-to-device and device-to-phone communication applications. However, commercial logic suggests the business models to support these mass-market application sectors will take some years to emerge.

Consumer applications at the time of writing look as though they may take off, provided there is wide enough support from the handset vendor community. Examples on the RFID Journal web site10 at June 2011 included contactless wrist bands for the Isle of Wight Festival, a ticketless check in system for Scandinavian airlines, a mental-health facility application and a smart-city system.11

The smart-city application is based on a sensor board that measures noise pollution (real-time noise maps), dust quantities, (real-time dust maps) the structural health of buildings, bridges and road, garbage levels (only empty bins when they are full), gas monitoring, environmental monitoring (CO2 and NO2) radiation detection and smart parking.

The sensor nodes communicate with the network using ZigBee12 radio links mainly to ensure that sensor devices have a sufficiently low energy requirement to be able to be powered from solar power or a battery lasting several years or more.

And actually this is quite a nice example of radio hardware being developed with an application appropriate protocol stack with a radio range up to 12 kilometres using ZigBee, albeit at a low data rate or Bluetooth or GPRS with some extended hibernation capabilities that reduce power drain to the sort of levels needed.

Note that the ZigBee standards process is reasonably closely linked with the Home Plug Alliance (see previous chapter) and the WiFi alliance. The ZigBee physical layer consists of two different options in three license-free frequency bands with sixteen channels at 2.4 GHz (the same as Bluetooth and WiFi) with a maximum data rate of 250 kbps, ten channels at 902 to 928 MHz (The US) with a maximum data rate of 40 kbps and one channel at 868 to 870 MHz with a maximum data rate of 20 kbps (Europe, Australia and New Zealand).

Both physical layers use direct sequence spread spectrum (DSSS) modulated O-QPSK with a 32 PN-code length and an RF bandwidth of 2 MHz in the 2.4-GHz band and BPSK modulation with a 15 PN-code length and an RF bandwidth of 600 kHz in Europe and 1200 kHz in North America.

And in the consumer space we may now finally be on the cusp of seeing broader market adoption. On 16 June 2011 Sony Ericsson announced a range of Android NFC phones13 with NXP, the semiconductor business formerly owned by Philips.

So using our phones to pay for things finally looks set to happen. Using our phone to talk to other devices via Bluetooth already happens, hands-free Bluetooth headsets being the most ubiquitous example. Using our phone to interrogate and communicate with microdevice networks, for example sensor networks is also beginning to happen. We can go online to look at video feed from CCTV cameras, we can go online to turn appliances on and off at home, we can go online to turn lights on and off at home via ZigBee or WiFi-enabled wireless switches.

But just because we can do these things does not mean we will. Many of us are quite happy opening and closing our own curtains. A suitable point on which to end this chapter.

1 http://www.hp.com/hpinfo/newsroom/press/2006/060717a.html.

2 http://embedded.seattle.intel-research.net/wiki/index.php?title=Intel_Mote_2.

3 http://www.hitachi.co.jp/Prod/mu-chip/.

4 http://www.tfl.gov.uk/tickets/14836.aspx.

5 http://www.rfidjournal.com/article/articleview/2436/.

6 http://www.rfida.com/apps/passive.htm.

7 http://www.bluetooth.com/Pages/Bluetooth-Home.aspx.

8 http://www.ZigBee.org/.

9 http://www.tinyos.net/.

10 http://www.rfidjournal.com/article/articleview/8553.

11 http://www.libelium.com/.

12 http://www.ZigBee.org/.

13 http://www.nearfieldcommunicationsworld.com/2011/06/16/38069/sony-ericsson-to-launch-range-of-android-nfc-phones/.