New Chip Promises Longer Battery Life

Roland Piquepaille writes "It always happens when you need it the most: the battery of your cellphone just died. But now, researchers of the University of Rochester have developed a wireless chip that needs ten times less power than current designs. The new chip relies on a technology named injection locked frequency divider (ILFD) which dramatically reduces the time needed to check for transmission frequencies which are performed several billion times per second by your current phone. The new chip uses five transistors and can perform divisions by 3 instead of only 2 by previous circuits, allowing a perfect communication between two phones communicating at 2.0001 and 2.0002 gigahertz respectively."

Comments

Not A Big Deal

[Bruce Perens]

The PLL component this is supposed to replace is a small-signal component. It is not a major user of the power budget of a cell phone. The big power users are the transmitter and the microprocessor. The PLL is not heat-sinked and does not run warm. If it's not hot, it's not a power hog.

Bruce

Re:Not A Big Deal

[geoskd]

The PLL component this is supposed to replace is a small-signal component. It is not a major user of the power budget of a cell phone. The big power users are the transmitter and the microprocessor. The PLL is not heat-sinked and does not run warm. If it's not hot, it's not a power hog.

The Problem is not that the PLL uses lots of energy, the problem is that digital circuitry, which the PLL feeds, uses power that is proportional to the frequency at which the PLL drives it. If you have a digital circuit at 2 GHz, it will use one tenth of the power of a circuit which runs at 20 GHz. This is important because traditional digital circuits which communicate with each other on specific frequencies, do so by running a clock speed of at least 10 times the communication frequency, and then using a microporcessor to count up clock pulses in order to exactly equal the right frequency. If you are running at 10x the communication frequency, then you need to count ten clock pulses for each communication signal cycle. If you need greater accuracy, then you need more clock pulses per communication cycle to get that accuracy. Thus, your digital circuits are in effect running at much higher clock frequencies than are necesary to actually achieve the communication. This is why your little 2 watt tx/rx chip actually consumes closer to 20 watts when it is communicatng actively.

What these researchers have done is found a way to adjust the frequency of the digital circuitry to exactly match the communication frequency, so instead of counting pulses, we can safely assume that 1 digital signal cycle = 1 communication cycle. This is just as good as clock pulse counting when it comes to processing digital communication signals, but up until now there was no way to adjust the source frequency with any real accuracy, so you had to run the source frequency very fast and count up pulses to get accuracy. Now, we no longer have to count, we just use one pulse / cycle, and were all set.

To explain in a slightly different way, we'll use the analogy of trying to accurately count a mountain of pennies. The easiest way to do so, is to weigh the whole pile, and then divde by the average weight of a single penny, and you get the total number of pennies. The question is how you get the "average weight" of a single penny. If you weigh just one penny, and use that as the average, then you have some total inaccuracy X. If you instead weigh 10 pennies and divde the weight by 10, the inaccuracy is much less: roughly X/10. This is how the old method of PLL circuit design worked. The greater the frequency, the more pennies you used to find the average weight, and so the greater the accuracy you could get in finding out the total number of pennies in the whole pile, or the exact frequency.

The new method described in the Article is roughly analagous to modifying all of your pennies to ensure that the variation in the weights of the pennies is much lower, so you can rely on just one penny to provide you with the precision needed to determine the total number in the pile.

I hope this cleared up some of the confusion.

-=Geoskd

Re:Not A Big Deal

[Dis*abstraction]

Actually, I'm still a little confused. Could you try an analogy using cars instead? Thanks.

Re:Not A Big Deal

[chriso11]

No, the digital circuitry does not run at the PLL frequency in a cell phone. The stable reference frequency from the crystal is upconverted to what is called the LO - this LO is mixed in with received signal from the antennea to downconvert the signal to a lower frequency. No digital processing occurs at 1.8GHz/1.9GHz on a cell phone - it is all much lower in frequency. That also goes for Bluetooth and WiFi.

The article is really short on details. The real power hog in a cell phone is the transmitter - it will draw 3Amps of current - while the rest of the receiver and up-conversion components are maybe 10% of that. And transmitters are already quite efficient - generally, ~50% of the input DC power winds up going out as RF power.

The lower power version of the PLL will be useful, since it needs to run constantly, even while not actively in a call.

Re:Not A Big Deal

[alx5000]

To explain in a slightly different way, we'll use the analogy of trying to accurately count a mountain of cars. The easiest way to do so, is to weigh the whole pile, and then divde by the average weight of a single car, and you get the total number of cars. The question is how you get the "average weight" of a single car. If you weigh just one car, and use that as the average, then you have some total inaccuracy X. If you instead weigh 10 cars and divde the weight by 10, the inaccuracy is much less: roughly X/10. This is how the old method of PLL circuit design worked. The greater the frequency, the more cars you used to find the average weight, and so the greater the accuracy you could get in finding out the total number of cars in the whole pile, or the exact frequency. The new method described in the Article is roughly analagous to modifying all of your cars to ensure that the variation in the weights of the cars is much lower, so you can rely on just one car to provide you with the precision needed to determine the total number in the pile.

Re:Not A Big Deal

[Jeff DeMaagd]

Another issue with your claims is that the power needed to operate a CMOS digital circuit goes up not linearly but by the square. A circuit that operates at 20GHz would consume about 100x the power as the same circuit that operates at 2GHz. I'm not aware of any commercial digital IC that can operate at 20GHz anyway.

Conversation

[LiquidCoooled]

Dude: Hui Wu invented this new chip that saves loads of power.

Bloke: Who?

Dude: Yes

Bloke: so who invented this chip.

Dude: Hui did.

Bloke: Thats what I'm asking you.

Dude: Yer I know, Hui did.

Bloke: Quit it and tell me who invented the chip.

Dude: Im not joking, Hui did.

Re:Thats interesting and all

[Bruce Perens]

The various segments of the readout are wired in a sort of square matrix to save wires at the chip outputs and driver transistors inside the chip or on the circuit board. They can't all be on at once because of the wires they share. They have to be driven in sequence. So, you see a sort of strobe light effect where each different part of the number is flashed at a different point while your head and eyes vibrate in a sort of arc.

Bruce

Re:Not(?) A Big Deal

[green1]

the problem is, even in "standby" the phone does a lot of transmitting, and that transmitting is still a power hog.

I'm not quite as negative as the grandparent poster, in that I'm happy if any component uses less power (every bit helps) but in reality, it's the transmitter that uses the lions share of the juice, not the reciever (even in standby).

Re:"ten times less power"?

[Guuge]

Imagine all the power the old chip doesn't use. Multiply this number by ten. This is the amount of power the new chip doesn't use. So you end up not using ten times as much power as you used to not use.

Two/Three

[samkass]

The new chip uses five transistors and can perform divisions by 3 instead of only 2 by previous circuits

Bender: "Ahhh, what an awful dream. Ones and zeroes everywhere... and I thought I saw a two!"
Fry: "It was just a dream, Bender. There's no such thing as two."

Re:Why are we still using batteries?

[Com2Kid]

Out of curiousity, why have we not yet figured out how to wirelessly power devices?

Short answer: We already have, it is just so inefficient that nobody uses it. (in fact it was invented over 100 years ago!)

Long answer: Electromagnetic waves radiate outwards. Either you have a simple non-directional antenna that radiates in all directions at the same time (in a sphere basically) and you lose power REALLY fast, or you have a directional antenna that radiates power in a cone at a target destination.

The omni-directional radiators suck so much that they are absolutely useless. Inverse square means 1/(x^2). Basically (and this is crappy math but gets the point across) if you have 10 watts at 1 feet, you would have 10*(1/(2^2)) = 2.5 watts at 2 feet. At 3 feet you would have 10*(1/9) = 1.11 watts. Please ignore that you would use meters instead of feet and that all my units are all messed up in various other ways as well. The point is that your power drops off REALLY fast.

So what about those directional antennas?

Well, you have to find some way to really accurately track someone's cell phone position, and have a world-wide array of directional antennas so that you can beam power to them no matter where they are at.

Oh and remember to keep those power levels low, else you will fry anything that gets in the way.

People worry about cell phones causing cancer as it is, directional power beamed at your head WOULD cause some serious issues!

Wireless power is possible, just not feasible!

Re:Thats interesting and all

[ltbarcly]

The red cones in your eye react more slowly than the other color receptors. Therefore when you are looking at the red led, which is a pure red light surrounded by darkness (not common in nature btw) what you are actually seeing is the vibration due to the crunch momentarily after they happen. Your brain adjusts so that you do not notice the vibration in your vision, except it cannot take into account the slight delay in the red cones. Therefore the red light appears to jump around.

Re: usb to 9v battery charger

[cheekyboy]

Make a 9volt USB battery charger

http://www.hackaday.com/entry/1234000520028239/

Or a WIND UP charger

http://www.edirectory.co.uk/pf/pages/moreinfoa.asp ?pe=CBHJGEGQ_+mobile+phone+wind+up+charger&cid=880

or a WIND TURBINE PHONE CHARGER

http://www.bytesurgery.com/gearedup/2006/02/a-wind -powered-phone-charger.html

What a crock

[amjohns]

This is mostly BS. First off, the PLL is a small fraction of the power consumed by a modern phone, even though it is running all the time. Far more power is consumed in the rest of the receiver chain, from the LNA (low nose amplifier) and the digital demodulator. And no, this does not do a thing to minimize the demod, as it is running all the time too, to detect an incoming call notification.

Second, the statement that a "phase-locked loop multiplies the pulse from a highly-stable reference clock, such as a quartz crystal oscillator, up to the desired frequency" is 100% false. The function of a PLL is to lock (in phase...) a divided down version of a totaly independent RF oscillator, called a VCO, to a divided down version of the reference clock. The distinction may appear subtle, but it's enormous. Multipliers are large, power consuming IC's, while dividers are fairly small and efficient. There are NO multipliers in a PLL, period. Also, PLL's can already do split division, it's called a fractional-N PLL.

Mobile, battery powered electronics will never achieve decent battery life beyond a few GHz. There are several effects coming into play, from cosmic noise to H2O and O2 molecular resonances to increased multipath effects, and most importantly path loss. RF power spreads in a spherical wavefront, so there is a 1/R^2 power falloff. BUT, you need to recognize that this is in terms of wavelength (lambda), which is mathematically equal to C/f (speed of light / frequency). The net result is that doubling the frequency on a radio link incurs a 4-fold power fallof for a fixed distance.

So if I want to go from say just under 2GHz w/ a current GSM system to say 8GHz, then I need an effective 16 times the power output from my transmitter. I say effective, because you can use antenna gain, but not in the mobile handset (it needs to be omnnidirectional), and base stations directionality is very limited, since they need to support many users on the same antenna, and can't steer the beam to all of them simultaneously. You wouldn't be allowed ot put out that much powr form a safety perspective, never mind the power consumption and heat requirements in the power-amplifier. Handsets are at 600 milli-watts now, we're not going to put out >10 watts!

Re:What a crock

[thestuckmud]

RF power spreads in a spherical wavefront, so there is a 1/R^2 power falloff. BUT, you need to recognize that this is in terms of wavelength (lambda), which is mathematically equal to C/f (speed of light / frequency). The net result is that doubling the frequency on a radio link incurs a 4-fold power fallof for a fixed distance.

Sorry, but this last point is wrong. The inverse square law for power is, indeed, in terms of power, not wavelength. Actual radiated power depends on the power input to the final stage of the transmitter times the efficiency of that stage, the transmission line, and antenna. It does not drop simply because of an increase in frequency.

Wavelength and frequency are related to a photon's energy, by the equation e = h*f (= h*c/lambda), but this is not relevant here.

Your physics inspector (and amateur extra, AB0VV)

Re:Why are we still using batteries?

[darthwader]

A while ago, Mythbusters did a "free energy" show. They collected a bunch of plans from "the Internet", built the devices, and tested them.

One of the devices that surprised me was a 50' long aerial, attached to some simple circuitry. The aerial absorbed RF energy, and the electronics converted it into a somewhat useful DC power supply. I think it was producing somewhere around 1 volt, no idea how much current, indoors. IIRC, they said it was "almost as good as a AA battery".

So, not only is is possible in theory, it's possible in practice. But it's still wildly impractical.

I think it's episode 24 (http://dsc.discovery.com/fansites/mythbusters/epi sode/episode_06.html).

Re:get your pronounciation right

[phlipped]

1: "Hoo" invented this new chip
2: No "Hway"!
1: YES way!
2: That's what I said
1: What?
2: The name of the guy is pronounced "Hway", not "Hoo"
1: Oh. I thought it was "Hoo"
2: No it's "Hway"
1: I see
2: Yes. Well ... Um ... I have to go ... over ... there now
1: Ok ...

This is all incorrect. PR & media idiocy as us

[3flp]

I don't post here very often, but this time I couldn't handle this. (Maybe I should drink less coffee). There was probably some paper at that uni, talking about an incremental improvement in frequency divider design. Ok, cool ... we may or may not see in in a PLL chip in a few years. But the news release (TFA) and RP's writeup are rubbish. Actually, after a bit of Googling, it's all over the net. Next thing I expect, my PHB will ask me to change my totaly unrelated design to use ILFD. My signature notwithstanding, I'll try to pick out some of the c***p, and put some actual information in. BTW, I design 3G mobile terminal circuitry full time. And yes, I am an arrogant SOB. That doesn't make me wrong.

"...But now, researchers of the University of Rochester have developed a wireless chip that needs ten times less power [GC] than current designs."

So far so good.

The new chip relies on a technology named injection locked frequency divider (ILFD) which dramatically reduces the time needed to check for transmission frequencies which are performed several billion times per second by your current phone.

This statement is wrong 2 times. First of all, the time needed to check for transmission frequencies depends on PLL settling time. Nothing to do with divider technology. Even broader scope, it is a rare occurence in 3G that the phone needs to change RF frequency. It's WCDMA, so all cells from a given operator transmit on the same channel. Secondly, tthe checking for transmission does NOT occur "several billion times per second". The RF carrier frequency is several billion cycles per second (ie several GHz). But the carrier frequency is changed on every 10ms roughly, even when it needs to happen. That's 100 times per second. GSM is different, as it does frequency hopping normally, but that doesn't change the point: nothing to do with divider technology.

The new chip uses five transistors and can perform divisions by 3 instead of only 2 by previous circuits

OK, agreed. Anyway, who gives a f**k. A modern PLL chip has a programmable divider, settable from 3 to several thousand. Yes, 3, because it is different technology.

..., allowing a perfect communication between two phones communicating at 2.0001 and 2.0002 gigahertz respectively.

That's not how mobile phones work. Mobiles establish connection with the cell (base station), then remain frequency locked to it, to compensate for temperature dependant frequency variation of their reference reference crystal oscillators - and Doppler shift, if they are moving. A "perfect" communication hardly ever depends on this. And frequency locking does not happen via changing PLL settings in this case anyway - too coarse steps, so other techniques are used.

Anyway, as other people posted already, the frequency synthesizer is not significant contributor to mobile terminal power consumption. Even old PLL chips only use a few milliamps

The ILFD technology seems to be good for building efficient frequency dividers at higher microwave frequencies. That will probably not affect current mobile phones anyway, because all the current systems work around 1-2GHz. Higher up, it's difficult to achieve coverage. Again, other people already pointed this out.

If you want real news in this area, go to sites like this, or this. Slashdot's editorial quality has degraded in the last few years so much that I am thinking about deleting it from my bookmarks.

[/rant]

Perpetuating the propagation loss myth

[dtmos]

I'm considering the devotion of the rest of my professional career to the eradication of the "propagation loss increases with frequency" myth.

Repeat after me:

Propagation loss does not increase with frequency!
Propagation loss does not increase with frequency!
Propagation loss does not increase with frequency!

Think about it: If the propagation loss of an electromagnetic wave increased in proportion to its frequency, there would be so much so much attenuation at the THz frequency of light that we'd never see sunlight--or stars. Propagation loss is independent of frequency, except for scattering due to molecular and atomic resonances that are insignificant at the frequencies we're discussing. (There are also changes in scattering behavior that become relevant in indoor applications, like propagation around corners.)

What is dependent on frequency, however, is the performance of the antennas we use to transmit and receive electromagnetic waves. Antennas can be characterized by a parameter called effective area. Returning to the sunlight example, recognize that the output power of a solar panel is proportional to its physical area; the larger this area, the greater the fraction of the incident power transmitted by the sun is received by the solar panel and converted to available output power. Receiving antennas, and antennas in general (even wire antennas), have an effective area; it's the area required to produce the measured output power, based on the density of transmitted power (watts/unit area) at the location of the receiving antenna.

Antennas can also be characterized by their gain, a function of their directivity and efficiency.

Interestingly, based on these two parameters any given antenna can be placed into one of two categories: There are constant-area antennas, the effective area of which is constant with frequency, and constant-gain antennas, the gain of which is constant with frequency. Constant-area antennas have gain that increases with frequency; constant-gain antennas have effective area that decreases with frequency.

The source of the myth is that most portable consumer wireless products use constant-gain antennas, usually some variant of a dipole. While the gain of a resonant dipole is constant with frequency, as the frequency goes up its physical length, and therefore its effective area, goes down. 2.4 GHz dipoles are physically smaller than 900 MHz dipoles. They therefore have less effective area, and recover less power from the incident wave. It seems like the path loss at 2.4 GHz is greater, but it's really just a result of the antenna choice in the product design. If consumer products used constant-area antennas, like a parabolic dish of fixed physical dimensions, exactly the opposite result would be found: Since constant-area antennas have gain that increases with frequency, the recovered power at 2.4 GHz would be greater than that at 900 MHz, and we could start a myth that propagation loss decreases with frequency.

Interestingly enough, if the transmitter has a constant-gain antenna and the receiver has a constant-area antenna (or vice-versa), the recovered power at the receiving antenna terminals would be independent of frequency (i.e., constant), and we could avoid the generation of propagation loss myths entirely.