| 2008-11-22 16:01:39 | 传感器趣谈（It’s the Sensors, Stupid） |
| Your design is only as good as its sensors. With the latest sensors incorporating smarter silicon， alternative energies， and wireless technologies， the possibilities for your next system are limitless. This month， Tom shows you some of the interesting sensors he discovered at the 2008 Sensors Expo & Conference. I hope you‘ll forgive the headline. Consider it a last hurrah as the seemingly perpetual political season finally totters to an end. Posturing and pandering the headline may be， but there's a lot of truth behind it. As I‘ve said before， the promise of all the wonderful MIPS，megahertz， and megabytes at our disposal can be fulfilled only to the degree sensors make the real world accessible. Having attended the Sensors Expo last June in Rosemont， IL， I invite you to look over my shoulder as I review what was on the show floor and wade through the press releases I received. I think you'll agree there‘s progress on all fronts with sensors that are more intelligent， green， and wireless than ever. To every engineer across this great profession， I say to you： the state of the sensors union is strong. （Cue obligatory applause from both sides of the aisle.） DigiLog？ It‘s true computers are digital and sensors are analog， and never the twain shall meet. They may never get married， but that doesn't mean they can‘t live together. What the silicon wizards create let no engineer cast asunder！ Indeed， more than anything， it‘s using the combination of digital intelligence and analog sensing to create ever "smarter" sensors that defines the march of progress. I remember in the old days when the show floor at Sensors was dominated by big， shiny metal things that were expensive to buy and use. Sure the old-school sensors are still around， but they're increasingly outnumbered by sensors that look， and think， like chips. Appearances can be deceiving and the analog or digital gender of a device may not be apparent at first glance. For example， consider the TLE4997 Hall effect sensor from Infineon Technologies. A Hall effect sensor can detect changes in the local magnetic field， making it ideal for noncontact position and rotation sensing. Notably， Hall sensors are a robust solution in harsh and dirty environments（e.g.， wheel speed sensor） that would quickly render prissier parts（e.g.， optical encoders） useless. Of course， a magnetic field is quite an analog phenomenon. And the TLE4997 has just three pins： Power，Ground， and an analog output corresponding to the strength of the magnetic field. It‘s pretty much analog from every angle. Just hang the TLE4997 on the ADC that comes with the typical MCU these days and have at it. Not so fast， grasshopper. A look under the covers reveals the chip，with a complete DSP processing subsystem，is really a hybrid of digital and analog （see Figure 1）。 
Clearly， introducing superfluous analog conversion stages into a design is a bad idea. Each extra conversion both wastes power and degrades performance（speed， accuracy）。 Nevertheless， it‘s the only option if you need a sensor with both digital smarts and analog I/O. For a fully digital solution， Infineon also offers the TLE4998， which is exactly the same chip except with a digital PWM output. While I'm on a roll， let‘s talk about the good， bad， and ugly of the PWM concept （i.e.， a digital way to represent an analog value）。 The good is that a PWM can kind of easily go both ways. You can hang an RC filter on it to get an analog voltage. Or， you can feed the PWM directly to a digital input（e.g.， a timer） to measure the duty cycle. Either option is presumed better than the alternative， namely adding an ADC （or DAC）。 But that may not be true if the MCU has otherwise unused converters built-in. The bad and ugly are that as a jack of all interfaces （i.e.， analog or digital），PWM is also master of none. To generate a clean analog （i.e.， DC） voltage from a digital PWM output requires a lot of filtering. That means the PWM has to run at a high frequency all the time， which consumes power （as does the RC filter itself） and generates noise. The digital option has the otherside- of-the-coin problem， requiring a timer to run in order to measure the duty cycle. Indeed， said timer may need to run at a high frequency in order to meet both measurement speed and resolution requirements. Consider the timer specifications required to accurately measure the duty cycle of PWM-encoded audio with， for example， 8-kHz sampling at 8-bit resolution. Cranking through the numbers， the timer would need to run at more than 2 MHz， and that‘s for mere voice-grade audio. Deep in the fine print of the TLE4998 press release it states that in addition to PWM， it supports something called SENT. Wonder what the heck that is？ Could it be a solution to A/D bipolar disorder？ Don‘t think so. It turns out SENT，which stands for Single Edge Nibble Transmission， is a new SAE standard for single wire data transfer （SAE J2716）。[1] As you can see in Figure 2， a J2716 packet comprises a fixed-length start frame followed by the data （and CRC） nibbles encoded as variablelength（i.e.， 16 different lengths） pulses. 
Figure 2-It seems there's no end to clever ways of sending multiple bits of data across a single wire. The SENT standard (single edge nibble transfer, SAE J2716) encodes each nibble as a variable-length pulse.
Frankly， SENT seems like the worst of both worlds. I see no obvious way to coerce a meaningful analog voltage from the SENT waveform. The showstopper is that the nibble pulse length doesn't vary with the bit significance. For instance， consider sending a 12-bit value. If you just hang a resistor and capacitor on SENT as you would a PWM， the voltage you get will be the same whether the value is 0xF00h，0x0F0h， or 0x00Fh. Oops， don‘t forget the CRC tacked onto the end， which just makes the situation worse. From the digital perspective， measuring the duty cycle of a PWM may be tedious and annoying， but it's surely a heck of a lot easier than deciphering SENT. SENT just coughs up more edges for you to deal with， yet transfers no additional information. Worse，the total time to transfer data with SENT varies depending on the value of that data. That leaves you （i.e.， application software） to deal with the timing uncertainty and having to presume the worst-case latency. （A SENT transfer can take anywhere from 456 to 816 μs.） Meanwhile， SENT offers no significant savings on the timer front either because you still need microsecond resolution just to accommodate 1-kHz sampling of the 24 bits （16-bit magnetic field strength plus 8-bit temperature）the TLE4998 has for you. Gee， I guess PWM isn‘t that bad after all. ANALOG ENOUGH FOR YA？ Let‘s start with the premise that it's best to digitize data at the source （i.e.，sensor） near the beginning of the signal chain. That‘s because the further an analog signal is shipped （traces，wires， connectors） before it's digitized，the more susceptible it is to interference. By contrast， it‘s easy to preserve the fidelity of digital data as it is passed along because ones and zeros can easily tolerate noise levels that would otherwise overwhelm a feeble analog signal. Beyond the better noise margin， digital signals are also amenable to error-correction techniques（e.g.， ECC）， making signal integrity virtually bulletproof. So it seems pretty obvious that the sensors themselves should be smart enough to include the A/D and a digital interface，as many do these days. It sounds good， except there's just one problem. Exactly what specs should that ADC have？ The answer to that question is different for each application， right？ Certainly， that‘s historically been the case， which explains why there are a zillion different A/D chips to choose from. Such musings were furthered when I came across the new Analog Devices AD7190 （see Figure 3）。 The headline that grabs your attention is 24-bit resolution with accuracy measured in nanovolts. Of course， these are marketing specifications and your mileage will vary. And it's not as if the typical application needs to measure a millionth of a volt anyway， although some do （e.g.， precision scale， medical）。 
Figure 3-Today, the ultra-high 24-bit resolution of the Analog Devices AD7190 positions it for high-end designs like ultra-precise weigh scales. Tomorrow could find chips like it serving the vast majority of ADC applications.
Multi-hop （e.g.， mesh， tree） networks typically rely on routing tables to define the path a packet takes from source to destination. One problem is that dynamic changes in the network topology （i.e.， the addition， removal，or movement of a node） can cause the entire network to hiccup while the new routes are discovered and the routing tables across the network are updated. Particularly in a situation where lot‘s of nodes are coming and going （i.e.， mobility）， it's not hard to imagine the network struggling to keep up. Virtual Extension claims their proprietary ISM band （868 and 915 MHz）radio and novel "flooding mesh" protocol is the answer. Having just glanced at a press release and their web page I have as many questions as answers. As best I can tell， the scheme is one where the entire network is synchronized and every node acts as a repeater for everything it receives. Thus， there is no need for routing or （re）configuration and nodes are free to join， leave， or move around within the network without causing network delays or downtime. It seems to me that there would be a price to pay， namely power consumption， for the redundant activity， but the press release does say a node will run for five years on a "C" cell. Anyway， it is a reminder that the "（re）configuration"issue is a factor to consider when designing a wireless sensor network for a particular application. GreenPeak Technologies is another new outfit showing off some innovation. On a schematic， their GP500C looks like many other IEEE 802.15.4 radios， a complete antennae-to-onesand- zeros solution with a SPI port for connecting to an MCU. But under the hood， you can see they‘ve tweaked the usual partitioning by moving the MAC and power management functions， typically handled by the MCU， into the radio （see Figure 5）。 According to GreenPeak， this enables a significant reduction in overall power consumption，especially when used with a power-optimized stack. 
It was seven years ago that I first looked into Bluetooth （"Bluetruth，Houston， We Have a Problem ...，"Circuit Cellar 134， 2001）。 Way back when， I praised its long-term prospects even as I cautioned that it would take time to finesse issues like interoperability，power consumption， and pricey silicon. Well， it has been a long time and a lot has changed. The Bluetooth powers that be have embarked on a mission to extend the standard‘s reach beyond audio with power-efficient control-oriented protocol upgrades. Meanwhile， what was rocket science silicon （e.g.， 32-bit MCU） back then is now easy to do. And with an installed base said to be 2 billion units， the fact that virtually all cell phones （and now a lot of cars as well）come with Bluetooth built-in can‘t be ignored. AIRcable specializes in using Bluetooth for wireless sensor networking. They've got a line-up of goodies， including long-range modules（up to 30 km！）， USB and RS-232 bridges， and even a tiny Bluetooth module with digital and analog I/O that‘s wirelessly programmable in BASIC （see Photo 1）。 
GOOD VIBRATIONS As oil prices go up， so does interest in alternative energy sources such as solar， wind， geothermal， and more. On a smaller scale， embedded designers also have alternative energy opportunities to consider， such as harvesting energy from vibration. Give the prize for cutest product name to AdaptivEnergy who came up with the "Joule-Thief" moniker for their energy-harvesting development kit. Inside the small module is a piezoelectric beam （i.e.， lever and mass） that converts vibratory motion to electrical energy （see Photo 2a）。Beyond the harvester itself， the module also integrates storage （e.g.，capacitor or battery） and low-power collection electronics to maximize efficiency （i.e.， amount of energy harvested over and above the amount consumed by the harvester）。 It‘s a small， light-weight， reliable， and maintenance-free solution that generates enough juice to easily power a wireless sensor. 点击查看Photo 2 Advanced Cerametrics is another company， pardon the pun， shaking things up （see Photo 2b）。 Their piezoelectric fiber composites provide a backbone for rolling your own vibration energy harvester design. Putting the pieces together， KCF Technologies offers a power harvesting sensor demonstration pack （see Photo 2c）。 The package includes three self-powered wireless sensors each with an accelerometer， temperature sensor， voltage sensor， and connection for an external pressure sensor. The sensors talk to a PC via a USB radio dongle and network monitoring software. One thing to keep in mind with vibration harvesters is the importance of matching the resonant frequency of the generator to the vibration. Sensors are typically available optimized for 60 Hz or multiples there of， which probably work well with motors and such. By contrast，squeezing the energy from random vibration is more of a trick. It can be done， but it will require more thought and analysis related to the application specifics （e.g.， storage capacity required based on the dutycycle requirements of the load）。 Not that batteries are going to disappear. Indeed， down the road maybe chips will come with a sticker that says "Batteries Are Included" （see Photo 3）。 
SHAPES OF THINGS After all these years， I‘m still amazed and thankful that silicon continues to inspire. Every time I think the business is getting a little stale， all it takes is something like a trip to the Sensors Expo to get the juices flowing again. Put it all together-smarter silicon，alternative energy， and the freedom of wireless-and the future seems even more exciting. Let's retrace our steps back to the Infineon Technologies booth to see an example of today‘s technology of tomorrow. Admittedly， their SP35 tire pressure sensor targets a niche market，albeit a giant one. Nevertheless， the SP35 serves as a template for tomorrow's mass-market chips. Behind its mild-mannered justanother- chip facade the SP35 packs a pressure sensor， radial accelerometer，temperature sensor， battery level sensor，and a 315-/434-MHz radio， all under the supervision of an 8051 MCU with 6 KB of flash memory （see Figure 6）。 Sensing， processing， storage，communication all in one device. That‘s the way the wind is blowing. 
We get the chips we deserve， and the ballot is full of candidates that can satisfy all the applications all of the time. So remember to vote early and vote often for parts that can truly make a difference. Tom Cantrell has been working on chip, board, and systems design and marketing for several years. You may reach him by e-mail at tom.cantrell@ circuitcellar.com. REFERENCE [1] SAE International, "SENT-Single Edge Nibble Transmission for Automotive Applications," J2716,2008, www.sae.org/technical/standards/J2716_200802. SOURCES Joule-Thief Energy-harvesting development kit AdaptivEnergy www.rlpenergy.com Piezoelectric fiber transducers Advanced Cerametrics, Inc. www.advancedcerametrics.com Bluetooth wireless adapters AIRcable www.aircable.net AD7190 and AD7626 ADCs Analog Devices, Inc. www.analog.com AT86RF212 Transceiver Atmel Corp. www.atmel.com Thin film battery Cymbet Corp. www.cymbet.com GP500C Transceiver GreenPeak Technologies www.greenpeak.com SP35 Tire pressure sensor and TLE4997/4998 Hall sensors Infineon Technologies www.infineon.com Power-harvesting sensor demo pack KCF Technologies www.kcftech.com ZigBit 900 ZigBee PRO Radio module MeshNetics www.meshnetics.com MiWi P2P Protocol stack and MRF24J40MA transceiver module Microchip Technology, Inc. www.microchip.com Wireless mesh-RS sensor network Virtual Extension www.virtual-extension.com |
评论 (5) | 阅读(937)
|
以下网友评论只代表其个人观点，不代表本网站的观点或立场
数据加载中......