Nordic Semiconductor technical article (Nordic Semiconductor editorial contact: Steven Keeping, e-mail: steven.keeping@nordicsemi.no, Tel: +61 (0)403 810827) TITLE: Bringing the power of wireless to medical applications STANDFIRST: A proprietary RF alternative based on a custom ASIC design adds wireless capability to medical devices with none of the IEEE standards-based design overhead that burdens Bluetooth and ZigBee. Dr. Torstein Heggebø explains TEXT: Portable instruments such as blood pressure monitors, pulse oximeters, spirometers and heart rate monitors are the tools of the doctor s trade in intensive care. And, thanks to modern electronics, this equipment has the accuracy, reliability and repeatability critical in life-and-death situations. Unfortunately, because the sensors for these instruments are typically attached to the patient by wires, they become bed-bound. Moreover, caring for the patient becomes awkward because whenever they need to be moved all the instrumentation has to be disconnected and later reconnected. Adding low power, low cost wireless technology could eliminate this tedious job. These devices could use their embedded RF link to communicate with a gateway in the same room, connected to the hospital LAN/WAN and transmitted to a centralised hub for control, monitoring or further analysis. At first glance, IEEE802.15.x standards-based RF wireless technologies such as Bluetooth, or its lower-powered cousin ZigBee, would appear the obvious choice for adding wireless to medical equipment. After all, both have been marketed aggressively courtesy of members of the Bluetooth Special Interest Group (SIG) or ZigBee Alliance. In addition, a lot of work has gone into these technologies to ensure compatibility, and there are plenty of chipsets available from major silicon vendors. However, for certain specialist applications, both Bluetooth and ZigBee have some disadvantages. Firstly, they are relatively expensive, and likely to remain so as the semiconductor companies attempt to recoup their huge investment in ensuring compatibility. Moreover, qualifying the product to the standard adds costly NRE charges. And, due to relatively inefficient protocols, power consumption is comparatively high - a major drawback in battery-powered devices. Other non-standard wireless alternatives deserve consideration. Indeed, these can offer superior integration (and thus lower cost), reduced power consumption and simplified architecture compared to the IEEE 802 offerings. An example is Nordic Semiconductor s nrf24xx series custom ASIC. This proprietary solution is based on a field-proven standard product widely used in products ranging from wireless gamepads and PC mice to intelligent sports equipment and automotive applications. Because they don t have to adhere to tightly defined standards, proprietary devices offer greater capacity for customising to meet the specific needs of the end application. For while Bluetooth and ZigBee protocols undoubtedly have their merits, ensuring interoperability with the standard can make optimised on-chip
systems integration (such as putting a sensor controller, ADC, microcontroller and radio on a single die) difficult. The best way to demonstrate the proprietary solution s potential advantages for a point-to-point wireless link for medical equipment is to consider an example. Here, we ll consider Nordic s nrf24e1 transceiver. The device can be readily compared with Bluetooth and ZigBee. The device is a system-on-chip comprising RF transceiver, 8051 microcontroller, 9- channel, 12-bit ADC and various standard interfaces manufactured using a 0.18 µm CMOS process. All necessary peripherals such as timers, real-time clock, watchdog, parallel IO and serial IO are available. On-chip regulators enable the platform to be powered from a single 1.9 to 3.6 V power supply. Measuring blood-oxygen levels Pulse oximeters are typical of the monitoring equipment used for an intensive care patient. These measure blood-oxygen level and pulse rate non-invasively by a clip attached to a patient s finger or earlobe that contains a photo detector and two LEDs. One LED emits red light and alternates with another emitting infrared (IR). The light passes through the body tissue to the photo detector. Some of the emitted light is absorbed by haemoglobin in the red blood cells. The absorption at different wavelengths (determined by the colour of the blood) varies with the oxygen saturation of the blood. The oxygen saturation level can then be calculated by comparing the absorption measured at the two wavelengths (by using spectrophotometry). Due to increased volume of arterial blood with each beat, it s also possible to measure a pulsatile component that accurately indicates heart rate. By sampling the photo detector data at a sufficiently high rate (e.g. 300 samples/s), it is also possible to measure variations in blood pressure during each cardiac cycle that can provide important diagnostic information of cardiovascular function (plethysmogram). The graph in figure 1 illustrates a typical pulsatile signal captured over a period of several seconds from the intensity variation detected in light shone through a finger using an oximeter. Figure 1: Signal captured from the intensity variation detected in light shone through a finger using an oximeter The red and infrared LEDs of the sensor typically draw 10 ma or less and are activated for 50 µs or less for each measurement. At this sampling rate, with 50 µs LED activation time per sample, each LED has a duty cycle of 1.5 percent. The average current consumption for the two LEDs is then: 2 10 ma 1.5 percent = 0.3 ma. The other components of the sensor consumes little power compared to the 2 10 ma of the two LEDs, and can thus be safely ignored here.
The oximeter s 300 samples/s corresponds to one measurement every 3.3 ms. To be able to update a remote display in (to the human eye) real-time and assuming a transmission latency of 30 ms, it s possible to carry 10 measurements in each transmitted package across a wireless link. Each wavelength measurement takes 8 bits of data totalling 16 bits in all for both the red and IR bands. The payload of each transmitted package with 10 samples then becomes 160 bits. Comparing 2.4 GHz protocols Bluetooth, ZigBee and the proprietary solution all offer methods of adding the wireless link to the pulse oximeter. Bluetooth is a standards-based RF technology operating in the licence-free 2.4 GHz Industrial, Scientific and Medical (ISM) band. (See box for more detail). The proprietary solution uses GFSK (Gaussian Frequency Shift Keying) modulation (similar to Bluetooth) and offers a nominal data rate of 1 Mbit/s. It has been intentionally designed with a low function overhead to maximise RF utilisation and minimise the power budget. The product introduces a hardware-based physical layer protocol processing that is transparent during normal operation. The proprietary solution uses the same channel scheme as Bluetooth. Both utilise up to 83, 1-MHz wide channels between 2.400 and 2.483 GHz. This compares with ZigBee s 16 channels (Figures 2 (a) and (b)). This offers both the Bluetooth and proprietary solution many more alternative relocation frequencies should they encounter interference from other devices operating in the crowded 2.4 GHz band. Figure 2: Zigbee s channel scheme (a) comprises 16 channels compared with Nordic s 83- channel scheme (b)
Bluetooth was primarily designed as a readily-compatible RF link operating, for example, between PDA, mobile and headset in a piconet or Personal Area Network (PAN). Its strength compatibility is also a weakness in that the protocol carries a major overhead to ensure one Bluetooth device talks to another without fuss. Consequently, a Bluetooth stack requires at least 250 kbytes of program memory (compared to the proprietary device s 4 kbytes and ZigBee s 28 kbytes see Table 1). In addition, Bluetooth constantly needs to maintain synchronisation between paired devices and therefore has to send a packet every 675 µs (equating to 1600 packets/s) to maintain the link, significantly increasing the duty cycle. If this synchronisation is not maintained re-acquisition is needed, which can take up to 3 seconds. Bluetooth uses more current than Nordic s transceiver on this synchronisation requirement alone. Even though the synchronisation packages are short, the high number of packets compared with Nordic s chip (1600 packets/s for Bluetooth versus 30 packets/s for the proprietary solution) significantly increases the duty cycle for Bluetooth s radio sub-system. In contrast to Bluetooth, ZigBee was designed with low power in mind and is endowed with a simplified protocol to reduce the packet overhead. (See Box 1.) The proprietary solution has also been specifically designed to be very economical with batteries. In both cases, the key to this low power consumption is minimising the duty cycle, so that the radio is in standby mode for as much time as possible. The sequence diagram for the proprietary solution (Figure 3) shows us the radiosystem of the device has to be active for about 200 + 48 + 208 + 40 + 200 + 56 = 752 µs for each acknowledged packet. This corresponds to an active duty cycle of about 2 percent.
Figure 3: Sequence diagram for the proprietary solution. This corresponds to an active duty cycle of about 2 percent The nrf24e1 transceiver consumes less than 20 ma when active, and 12 µa when in stand-by mode. The average current consumption of the radio and the microcontroller then becomes approximately: 20 ma 2 percent = 0.4 ma (ignoring the negligible standby current). For the ZigBee solution, the device is active for 200 + 200 + 200 + 60 + 1120 + 60 + 200 + 352 = 2392 µs per acknowledged packet. This corresponds to an active duty cycle of about 7 percent. Assuming the ZigBee chip s average current consumption is similar to the proprietary device, for this example this will be around: 20 ma 7 percent = 1.4 ma. Consequently, in the pulse oximeter the simple proprietary protocol will consume one-third the transmission power of a ZigBee solution. For the ZigBee solution, the power consumption required for the RF link dominates, while for the proprietary protocol the power consumption for the sensor and the RF link are in balance. (Comparing the 0.3 ma for the sensor with 0.4 ma for the radio.) The proprietary solution could run continuously for more than a week on a small 200 mah battery, while a ZigBee solution would run less than half that time, or require a battery twice the size. Looking beyond the standard Bluetooth and ZigBee are wonderful examples of what standards-based technologies are designed to do: ensure wide and seamless compatibility across international markets. Bluetooth enjoys a huge installed base of PDAs, headsets, mobile phones and laptop PCs. For these applications adhering to a standard does indeed eliminate much of the design challenge. You know your Bluetooth product will be able to communicate with any other Bluetooth device at the desired range and data transfer rate. The newly ratified ZigBee standard was specifically developed for low power, low cost, low data rate, reliable wireless monitoring and control applications relevant to industrial and in particular home automation applications. Nonetheless, it can be over-engineered and expensive for many low-demand applications. Designers are forced to use functionality, performance and compatibility that is simply not required. Because the oximeter s RF link doesn t have to be qualified to a standard, the time-tomarket schedule for the product is shortened. However, as with all radio devices, the oximeter will be required to conform to the appropriate communications authority s regulations such as those of Europe s ETSI or the US s FCC, this is true of any RF communication whether it is designed to a standard or not. For many applications - particularly in the medical sector - the complexity, cost and qualification challenges of Bluetooth and ZigBee outweigh their advantages. Silicon radios that don t conform to IEEE standards offer inexpensive, power-frugal and proven alternatives. BODY COPY ENDS
Further information: Torstein Heggebø has a Ph.D. in computer science from the Norwegian University of Science and Technology in Trondheim. He has 20 years IC design experience, and has been R&D manager for several Norwegian technology companies. Dr. Heggebø currently works as a senior project manager for customer projects using the Nordic Semiconductor RF-ASIC design platform. Table 1 Bluetooth packet structure: 1. Access code 68 or 72bit 2. Header 56 bit 3. Data payload 160 bit Total: 288 or 292 bit ZigBee packet structure: 1. Preamble 32 bit 2. Frame de-limiter 8 bit 3. Frame length 8 bit 4. Frame control 16 bit 5. Data sequence number 8 bit 6. Address ID 32 bit 7. Data payload 160 bit 8. Frame checksum 16 bit Total: 280 bit Proprietary packet structure: 1. Preamble 8 bit 2. Address 32 bit 3. Data payload 160 bit 4. CRC 8 bit Total: 208 bit Box 1: Bluetooth and ZigBee Technical Primer Bluetooth operates in the licence-free ISM 2.4 GHz band and is available in 3 power levels: Class 1 (100 m line-of-sight range), Class 2 (10 m), and Class 3 (2-3 m). Data is transferred between 1 master and up to 7 slaves in a WPAN or piconet at up to 723 kbit/s. GFSK (Gaussian Frequency Shift Keying) modulation is employed across 83 channels of 1 Mbit/s each. GFSK applies Gaussian filtering to the modulated baseband signal before it is applied to the carrier. This results in a dampened frequency swing between 1s and 0s and a narrower, cleaner spectrum for the transmitted signal compared with straight FSK (Frequency Shift Keying). Each piconet device is assigned a unique 48-bit identifier and the first one detected (usually within 2 seconds) becomes master, and sets 1600 frequency slots to be used each second. All other devices in the piconet lock to this sequence. Masters transmit in even slots, slaves respond in odd slots. Active slave devices in the piconet are assigned an address, and listen for slots addressed to them.
Each slave can also go into lower power sniff (periodically listen during specific sniff slots but lose synchronisation); hold (listen only to determine whether to become active); or park mode (gives up address). Although hold and park modes extend battery life, the device loses synchronisation for at least 1600 hops and has to wait for a new link to be set up. This can take several seconds so doesn t assist applications that need a continuously fast response. To combat interference, Bluetooth Version 1.2 incorporates Adaptive Frequency Hopping (AFH) that allows two communicating Bluetooth devices to constantly change their mutual communications frequency across the band to avoid clashing with other 2.4 GHz devices. Bluetooth also includes profiles that designers can select to target their development. Naturally, commercial pressures from Bluetooth SIG members mean most target media and file transfer applications on mobile phones. Consequently, development using Bluetooth profiles is not trivial particularly for simpler, low capacity applications common to industrial. ZigBee is a simpler data protocol that offers high reliability, including acknowledgement of each transmission burst and other techniques to maintain communications integrity. ZigBee doesn t require Bluetooth s synchronisation, decreasing power requirements considerably and is a third as fast with a maximum 250 kbit/s data rate. Like Bluetooth, ZigBee operates in the ISM 2.4 GHz band (16 channels at 5 MHz spacing), with additional European 868 MHz (single channel) and US 915 MHz (10 channels spaced at 2 MHz) options. ZigBee exploits DSSS (Direct Sequence Spread Spectrum) for data transmission that offers some immunity to interference, but at the cost of transmitting excessive data packets, incurring bandwidth usage and current consumption overheads. Although ZigBee attempts to address the potential weaknesses of Bluetooth in low-latency, low data rate applications, applications at the RF physical layer still have to carry the overhead of interoperability functions required by the 802.15.4 spec. According to the Bluetooth and ZigBee organisations the standards are complementary rather than competitive. ZigBee does allow for many more nodes up to 4090 compared to Bluetooth s 7 plus master. Power consumption, however, is a major differentiator. ZigBee is designed for very low duty cycle, ultra long life applications where battery life is measured in years. Continuous Bluetooth communications, however, can drain batteries in hours. ZigBee chipsets also cost a fraction of a Bluetooth solution (although there are variants of the Bluetooth protocol stack that offer less than the full range of profiles for less expense). NORDIC SEMICONDUCTOR, www.nordicsemi.no May be reproduced with permission from Nordic Semiconductor