WO2010043992A1

WO2010043992A1 - Crystalless transceiver

Info

Publication number: WO2010043992A1
Application number: PCT/IB2009/054296
Authority: WO
Inventors: Michiel Van Elzakker
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2008-10-13
Filing date: 2009-10-01
Publication date: 2010-04-22

Abstract

In wireless transceivers the crystal oscillators for frequency reference are often replaced oscillators that do not rely on any crystal. By combining the accuracy of these kinds of crystalless oscillators over the network into a higher shared accuracy, the required individual accuracy is reduced. This severely reduces the penalty in power and consequently, form factor and cost.

Description

CRYSTALLESS TRANSCEIVER

TECHNICAL FIELD

The present invention relates to transceivers in wireless networks. More specifically, the transceivers in accordance with the present invention do not contain any crystal for providing reference frequency to an oscillator. The invention also relates to a method of adjusting reference frequency of a transceiver and to a corresponding computer program product comprising instructions for implementing the steps of said method.

BACKGROUND OF THE INVENTION

It is generally the aim to utilise in wireless networks transceivers that have a small form factor and are inexpensive. Closely related to low cost and small form factor is low power, because in a wireless application a larger power source directly increases cost and form factor. Wireless transceivers rely on accurate frequency and timing information in order to achieve several necessary and desirable properties.

A wireless transceiver has to obey strict regulations concerning the frequency band or bands in which it is allowed to transmit. An example is the CEPT/ERC Rec. 70-03 868-870 MHz industrial, scientific and medical (ISM) band. The complete band is only 2 MHz wide and is divided into sub-bands which can be as narrow as 50 kHz. A narrow band like this can be attractive for a low power wireless sensor network because transceivers can be implemented with a low (peak) power for both transmit and receive. For such a sub-band the absolute accuracy of the transmitted frequency needs to be better than 50 parts per million (ppm).

Absolute frequency accuracy is required for regulations. Apart from that wireless communication benefits from relative accuracy in timing and frequency as well. Protocol timing needs sufficiently accurate relative timing. This timing ensures that the receiver is ready to receive in time for the transmission. A sufficiently accurate relative frequency ensures that the receiver is active in the same frequency (sub-) band as the transmitter. A high absolute accuracy leads to a high relative accuracy. A high relative accuracy does not automatically lead to a high absolute accuracy. For instance when every clock is approximately 20% too fast the relative accuracy can still easily be within 1%. Margin in timing and frequency is added by making the receiver and/or the transmitter active for a longer time and in a wider frequency band. A little margin is inevitable but especially in a low power application it is desirable to have a small margin. Often the required absolute frequency accuracy is significantly higher than the required relative timing accuracy. Transceivers use accurate oscillators to provide time and frequency references. Oscillator accuracy is limited by phase noise, manufacturing variations, stability over temperature and stability over time, i.e. aging. When accuracy is limited by phase noise it can generally be improved by using more power. As a result an accurate oscillator generally needs more power than a less accurate one. For protocol timing a transceiver in a wireless network needs a timing reference. However, it only needs the more accurate frequency reference when it is actually transmitting or receiving. In a wireless sensor network this is usually happens during a small fraction of the time. Consequently wireless sensor transceivers are often designed with one low power oscillator as timing reference and a second more accurate oscillator as frequency reference. This frequency reference oscillator is only turned on for the small amount of time when it is required. This method is commonly referred to as duty-cycling.

For various practical reasons the centre frequency of the modulated signal is often not the frequency of the reference oscillator. The frequency is scaled instead. The same is true for the timing reference oscillator. It is common practice to use quartz crystals in these reference oscillators. An example of a low power low cost transceiver is given in a publication by Wong, et al., "A IV Micropower System-on- Chip for Vital-Sign Monitoring in Wireless Body Sensor Networks," ISSCC Dig. Tech. Papers, pp. 138-139, Feb. 2008. In this example the frequency reference oscillator uses a 16 MHz crystal and the timing reference oscillator a 32 kHz one.

Quartz crystal oscillators can provide references with an excellent accuracy. The manufacturing process is very mature and includes a calibration step that can achieve for example 10 ppm accuracy. Especially high frequency crystals have a high stability over temperature. Stability over time is sufficient for the total lifetime of a transceiver. Finally these properties can be combined with low phase noise at low power consumption because of the mechanical resonance with a very high quality factor.

The main disadvantages of a crystal are the relatively bulky form factor and the relatively high cost. The high cost is partially caused by the component itself and partially by the additional manufacturing effort for integration. Thus, the crystal is becoming a bottleneck in the further reduction of form factor and price of wireless sensor nodes.

One technology trend replaces crystals by micro-electrical-mechanical structure (MEMS) resonators, which can be cheaper and less bulky, but also have a lower intrinsic accuracy. For application in a transceiver especially the high manufacturing variations and the small temperature stability need to be compensated. A possible compensation method locks a complementary metal-oxide-semiconductor (CMOS) oscillator through a fractional N phase locked loop (PLL) to the MEMS resonator, compensating offset and temperature variations through the control signal for the PLL. This is done at the expense of increased power consumption.

Another potential replacement for a crystal reference oscillator is a free running CMOS oscillator. Like a MEMS oscillator, extra power is required for a given phase noise compared to a crystal oscillator. The intrinsic accuracy is even lower than that of a MEMS resonator. The component price is potentially low and can very well be dominated by costs associated with calibration over a temperature range. A main advantage is the possibility to directly compensate systematic errors right in the oscillator. Because of measures to prevent aging and because of phase noise, high accuracy requires high power consumption.

Relative accuracy is commonly obtained through synchronisation. There are several known methods to achieve synchronisation, for instance the timing synchronisation function within the IEEE 802.11 standard. These synchronisation methods rely on some form of phase lock between continuously running timing reference oscillators. The result is relative timing accuracy, not absolute frequency accuracy. It is widely recognised that the crystal is becoming a bottleneck in the further reduction of form factor and cost of wireless transceivers. It is becoming apparent that drop-in replacements for a crystal, like a MEMS resonator and a free running CMOS oscillator, will need a lot of power to achieve the required very high absolute accuracy. This results in a larger power source which directly increases cost and form factor and therefore makes them hardly suitable for application in wireless (sensor) nodes. The present invention overcomes the problem of obtaining very high individual frequency accuracy.

SUMMARY OF THE INVENTION According to a first aspect of the invention, there is provided a method of adjusting reference frequency of a wireless transceiver, the method comprising:

- receiving at least one shared reference frequency together with at least one of the following: magnitude of error of the shared frequency and origin of the magnitude of error; - allocating a weighting factor for the at least one shared frequency; and comparing the received at least one shared frequency with frequency provided by a frequency source of the transceiver for adjusting the frequency provided by the frequency source by taking into account the weighting factor. Thus, the present invention is based on sharing reference frequencies and possibly magnitude and origin of associated inaccuracies between transceivers. This allows the accuracy of many oscillators to be combined into a shared higher accuracy. The invention results in lower power consumption, form factor and cost.

According to a second aspect of the invention, there is provided a computer program product comprising instructions for implementing the method according to the first aspect of the invention when loaded and run on computer means of a transceiver.

According to a third aspect of the invention, there is provided a transceiver capable of adjusting frequency provided by its frequency source, the transceiver comprising: a demodulator for demodulating a signal comprising at least one shared reference frequency, the signal further comprising at least one of the following: magnitude of error of the shared frequency and origin of the magnitude of error; - a controller for allocating a weighting factor for the at least one shared frequency; and a processor for comparing the received at least one shared frequency with frequency provided by the frequency source of the transceiver for adjusting the frequency provided by the frequency source by taking into account the weighting factor.

Other aspects of the invention are recited in the dependent claims attached hereto.

BRIEF DESCRIPTION OF THE DRAWINGS

Other features and advantages of the invention will become apparent from the following description of non- limiting exemplary embodiments, with reference to the appended drawings, in which:

- Figure 1 is a schematic showing five nodes of an exemplary sensor network;

- Figure 2 is a simplified block diagram illustrating a state of the art transceiver;

- Figure 3 is a simplified block diagram illustrating a transceiver in accordance with an embodiment of the present invention;

- Figure 4 is a simplified block diagram illustrating a reference frequency oscillator in accordance with the present invention;

- Figure 5 shows two controllers as separate physical units;

- Figure 6 shows one integrated controller being part of the reference frequency oscillator;

- Figure 7 is a simplified block diagram illustrating the controller in accordance with the present invention;

- Figure 8 shows another configuration of an exemplary sensor network;

- Figure 9 shows another configuration of an exemplary sensor network; and - Figure 10 is a flow chart illustrating a method of adjusting reference frequency in accordance with an embodiment of the present invention. DETAILED DESCRIPTION OF AN EMBODIMENT OF THE INVENTION

The teachings of the present invention are described in the context of a wireless sensor network, although the invention is not limited to this environment. Figure 1 shows a simplified schematic diagram of this kind of sensor wireless network. There are shown five nodes: Nl, N2, N3, N4 and N5 that are arranged to communicate with each other over wireless links. The present invention is based on an idea that the transceivers in the network not only share their reference frequencies to other nodes in the network, but also magnitude and/or origin of associated inaccuracies is/are shared. Figure 2 illustrates in simplified block diagram form an exemplary conventional transceiver 200 that has a crystal 201 for providing reference frequency to a frequency reference oscillator 203. There is also shown a timing reference oscillator 205 for time keeping. The timing reference oscillator 205 provides timing reference to a processor 207 which is further connected to a modulator/demodulator unit 209 that obtains frequency reference from the frequency reference oscillator 203. The modulator/demodulator 209 is further connected to an antenna 211 for sending/receiving radio signals.

In accordance with an embodiment of the present invention, the crystal 201 is no longer needed in the transceiver 300 as is illustrated in Figure 3 representing in block diagram form an embodiment of the present invention. The changes that are made to the transceiver 300 to implement this invention are shown in Figure 3. The crystal 201 is removed, the operations of the frequency reference oscillator 203 and the modulator/demodulator 209 are modified as will be described later in more detail and a (digital) controller 313 is added. The demodulator 209 is arranged to sample the modulation frequency of the transmitter and to participate in the calculation of the frequency difference between the reference frequency provided by the frequency reference oscillator 203 and the reference frequencies received from other transceivers. The calculated frequency differences are then communicated to the controller 313 which further provides frequency setting to the oscillator 203, i.e. the output frequency of the frequency reference oscillator 203 is set (or changed) according to the input provided by the controller 313. Information of the frequency inaccuracies are communicated between the controller 313 and the processor 207. Although the accuracy requirements are significantly relaxed, the frequency reference oscillator 203 needed for this invention is similar to a drop-in crystal replacement as described earlier. Such a frequency reference oscillator 203 is shown in Figure 4 in more detail. A drop-in replacement contains some form of an oscillator controller 411 that can (partially) be integrated with the controller 313 of the present invention, but can also be left separate. Frequency setting and/or calibration signal(s) are input externally to the oscillator controller 411. In the figure there are also shown a disturbance measurements unit 401 for measuring disturbances and a controllable oscillator core 405. Different disturbances are due for instance to temperature fluctuations, supply voltage fluctuations and so forth. Supply voltage fluctuations are possible fluctuations in the supply voltage biasing the frequency reference oscillator 203, e.g. a supply voltage supplied by a battery is not constant in time, but will decrease as the battery drains. The oscillator controller 411 receives disturbance information from the disturbance measurements unit 401 and the oscillator controller 411 then feeds a correction signal to the controllable oscillator core 405 which then outputs a periodic signal with an appropriately changed frequency.

Figure 5 shows the solution with two separate controllers, whereas Figure 6 shows the solution where only one integrated controller 611 is used. The amount of integration is a power and cost trade-off and a design effort trade-off.

Specifications of the inaccuracies of the reference oscillator 203 are input to the controller 313 shown in Figure 3. At the start, i.e. just after the radio has been fabricated, the frequency accuracy of the frequency reference oscillator is known (i.e. calibrated and measured) and this initial magnitude of error is input to the controller 313, e.g. manually. In this case, information from the frequency reference oscillator 203 is used in the controller 313. After that, in actual operation, this magnitude of error is updated each time a weighted frequency sample is taken as will be described later on. In these cases, information from the controller 313 is used in the frequency reference oscillator 203. It is important that the combined accuracy of all oscillators in the network is sufficient to obey the regulations and to enable successful communication. It is not necessary for all nodes to have a similar accuracy. For instance, for a network with a low total number of transceivers the accuracy requirements can be met by including a few (more expensive) nodes with a crystal. These special nodes do not necessarily need to have a controllable frequency. Overall this can be a cheaper solution than having a similar accuracy for all nodes. In an embodiment of the present invention a receiver needs to sample, i.e. measure and store, the modulation frequency of a transmitter it is directly communicating with. A measurement requires a comparison between two quantities (e.g. frequencies) of the same sort. The most accurate frequency measurement arises from comparing the received reference frequency with (a scaled version of) the local frequency reference oscillator. The received frequency Received cannot be measured directly, but the difference Received - fold can be measured. As the present frequency fold is known, the received frequency Received can be determined. The updated frequency is calculated using f_new = wl -fu + w2 f_iecei_Ved with weighting factors w\ and w2 (wl + w2 = 1). The sampling action can be done in two steps. The first step is to generate a signal that is proportional to the difference between the frequency used to modulate in the transmitter and the frequency used to demodulate in the receiver.

The demodulation part 209 of the transceiver 300 needs to distinguish the modulated signal from the transmitter from interference and noise. The frequency is an important criterion to make this distinction. In order to better make this distinction the frequency difference is often tracked in existing demodulators. Examples are differentiate and multiply (DAM) and/or cross, differentiate and multiply (CDM) demodulators for frequency shift keying (FSK) demodulation and a demodulator that uses a fast Fourier transformation (FFT) of the IF signal. In accordance with the present invention the frequency difference is not just tracked internally but it is also made available. This extra output can be analogue or digital. The measurement of the frequency difference is one source of inaccuracy that can be taken into account for this invention. The frequency difference signal can have a different accuracy than the reference itself. The magnitude of the inaccuracy of the measurement depends on implementation, interferers and phase noise.

Measuring the frequency difference is the first step in obtaining a frequency sample. The second step is to obtain a representation of the measured frequency that can be stored. In this description the second step is performed in the controller 313, although it can also be performed in the demodulator 209 of the transceiver 300. The controller 313 for a single transceiver 300 is responsible for:

• Obtaining frequency samples from frequency difference measurements obtained by the demodulator 209;

• Managing frequency samples;

• Determining and setting appropriate weighted averages; and

• Helping the controllers 313 of the other transceivers to manage their frequency samples. This is a repeating process. The controller 313 can be completely local, but some parts can also be in a different location. It is for instance possible to integrate parts of the controllers 313 of various nodes into a central node.

A basic version of the controller 313 is shown in Figure 7. There are shown two major units, namely sample management unit 701 and weighted average determination unit 703. The frequency difference information is obtained from the demodulator 209 and fed into the controller 313. The frequency difference should be quantified in a practical unit, for example in a number of bits (in a digital implementation) or in a voltage or current (in an analogue implementation). The frequency setting (output of the controller) should be quantified in a similar unit. If the quantification of the frequency difference and the quantification of the frequency setting do not make use of the same conversion factor (e.g., "a" for frequency difference and "b" for frequency setting), this difference can be compensated for by scaling the frequency difference (e.g. by b/a) so that the scaled version can be compared to the frequency setting. In Figure 7 this is done by a divider denoted by a forward slash in that figure. Instead of the division, a multiplication could also be used. For instance in an exemplary solution the controller 313 drives a low- frequency oscillator of 10 MHz which is used as the input of a phase-locked loop generating the output reference frequency of 870 MHz. In this case the frequency setting refers to 10 MHz and the frequency difference refers to 870 MHz, and the latter needs to be divided by 87 before it can be compared to the former. However, the scaling is optional and in some solutions it is not needed. Managing frequency samples includes: • Storing samples;

• Deciding which samples to discard or to combine through a weighted average with other samples, in order to keep the collection of samples manageable, sources of inaccuracy can also be combined; • Gathering information on the network about external sources of inaccuracy associated with the samples;

• Deciding when to obtain a new sample and administering the inaccuracy of the sample action. It is possible not to sample during an established link. Also it may be necessary to periodically establish a new radio link just to ensure accuracy; and

• Periodically updating the inaccuracy of every sample to reflect continuous internal sources of inaccuracy (for example aging, improperly compensated temperature variations and low frequency phase noise).

To determine an appropriate weighted average the controller 313 takes into account which inaccuracies are correlated and to what extent. Appropriately weighting samples with uncorrelated inaccuracies gives a lower combined inaccuracy. When the weighted average has been calculated it can be used as input for the reference oscillator 203.

Information needs to be shared with other controllers 313 about the sources and magnitudes of inaccuracy in the current weighted average as being used by the reference oscillator 203. The internal inaccuracies are administered by the controller 313 itself. Information about every external inaccuracy originates from another controller 313 for which it is internal. Sharing this information between the controllers 313 allows all controllers 313 to appropriately manage their samples. In order to avoid excessive communication and calculation, the number of updates should not be too high. One method to reduce the number of updates is quantization in for example the internal aging. This may result in the aging always being overestimated a bit. In other words, the error in accuracy due to aging is not updated continuously, but instead it is updated discontinuously, e.g. each hour (updating period). At the start of a period, the error in accuracy is updated due to the expected aging in the coming period, which will result in an overestimating of the aging when a weighted frequency average is taken during this period. Some features of the controller 313 are next illustrated with an example. Figure 8 shows a wireless (sensor) network and some information about past connections. The resulting sample collection of node Nl is shown in Table 1. The first frequency sample is the result of (factory) calibration at tθ, i.e. at time instance 0. Since that moment it has been aging, the accompanying magnitude of inaccuracy has been calculated and updated by the controller 313. As shown in the tables, each frequency sample further consists of frequency subsamples with individual sources of inaccuracies and magnitudes of inaccuracies. In the tables the notation "sigma" signifies the standard deviation of Gaussian distribution. Regarding the samples from N2, the sample action corresponds to the inaccuracies due to reference frequencies communicated with Nl and regarding the samples of N3, the sample action corresponds to the inaccuracies due to reference frequencies communicated with Nl, N4 and N5.

Table 1: Simplified example sample collection of Nl

The second frequency sample is obtained from node N2 at tl. N2 has aged little at tl and was only using its own calibration as a frequency sample. The third sample illustrates samples and inaccuracies that have been grouped. The low combined sample error is the result of repetitive sampling.

In this example the magnitude is administered in sigma ppm. Thus, the combined magnitude of inaccuracy emanating from a specific node N can be calculated by using the following formula:

^MN =*"VΣ^² _' 0⁾

where x denotes sigma confidence level, y denotes the magnitude of inaccuracy for a source of inaccuracy k for a certain node N.

For the own calibration at Nl and by using one sigma accuracy, the result by using formula (1) leads to:

On the other hand, if a confidence level of ten sigma is chosen, the maximum frequency error based on Formula (1) is

10 • _ΛM² +5² +5²+2² - 84[ppm]

The present maximum frequency error of the sample obtained from N2 at tl (time instant 1) is

10 • _ΛM² +5² +10²+5² +2² - \30[ppm] ,

and for the combined samples from N3 at around t2

10 • _Λ/2.5² +3²+3²+1² +2² - 54[ppm] . For ease of calculation it is assumed that the various samples are weighted by using the following coefficients: 0.2, 0.1 and 0.7. Instead of using coefficients, it is also possible to divide the actual weighting equation by the sum of the weighting factors. Table 1 illustrates how the magnitudes of inaccuracies in the weighted average are calculated. The sources and magnitudes in Table 2 need to be shared over the network.

Table 2: Calculation of the magnitude of the inaccuracies in the weighted average of Nl

The maximum frequency error of this weighted average is smaller than that of either of the original samples. When calculating this error, the subsamples of the same type, such as 'TSf l 's aging from tl to t2" and "Nl 's aging since t2" should be combined before inserting them into Equation 1. The error is thus calculated in the following way:

10 ^• Vθ.8² + 1² + 0.4² + 0.5² + 1² + 1.5² + 1.75² + 2.1² + 2.1² + 0.7² + 2² - 4l[ppm]

In the example above, for simplicity, it was as if the reference frequency was updated once using three samples with weighting factors 0.2, 0.1 and 0.7. However, in most applications the frequency would be updated in two stages. An updated frequency /i of node Nl is first determined at tl while communicating with node N2. The updated frequency is calculated as (^₁ ⁺ refers to just after the update at tl, whereas t^ refers to just before the update at tl):

f_ι{t = t; ) = j f_ι{t = t;)+j f₂{t = ς),

where f∑ is the received frequency of node N2 (the weighting factors are derived from the 0.2 and 0.1 factors divided by 0.2+0.1). The frequency /i of node Nl is updated again at t2 while communicating with node N3. The updated frequency is calculated as:

f_γ (t = *₂ ⁺ ) = 0.3 - f_γ(t = f- )+ 0.7 ^■ /₃ fr = t- \

where fi, is the received shared frequency of node N3. It is to be noted that there is a difference between f_γ [t = ^₁ ⁺ ) and f_γ (t = t₂ ) due to aging. The magnitude of error of

/1 would also be updated twice (at tl and t2), but the net result would be as given in

Table 2 and subsequent equations.

It is also possible to invoke temperature compensation. The oscillator controller 411 for temperature compensation in the reference oscillator 203 can be integrated into the controller 313 to obtain an integrated controller 611 as shown in

Figure 6. In a straightforward implementation every frequency sample is accompanied by a simultaneous sample of the oscillator's temperature. Another exemplary wireless (sensor) network is shown in Figure 9 and the resulting sample collection of N6 is shown in Table 3.

Table 3: Simplified example sample collection of N6

The (factory) calibration has been performed over temperature by offering an accurate external frequency reference while sweeping the temperature. N6's controller 313 has sampled this reference and has simultaneously sampled the temperature it observed. Because the temperature is observed locally the sweep does not need to be accurate and can therefore be performed cheaply.

At a temperature of 10 degrees Celsius the sample obtained from N7 at tl is a lot more accurate than both calibration samples. The following allocation of weighting factors is close to optimal: 0 for the first calibration at tO, 0 for the second own calibration at tO and 1 for the sample from N7 at tl .

At 50 degrees the sample obtained from N7 at tl is not accurate, because it has been obtained at 10 degrees. The first own calibration sample is not accurate because it has been obtained at 10 degrees and has aged a lot. Also the second own calibration sample is not accurate because it has aged a lot. However, because of the correlation in the aging, the difference between both calibration samples is an accurate measurement of what the temperature difference between 10 and 50 degrees Celsius does to the frequency.

In this example weighting factors of 1, -1 and 1 are used, which is close to optimal. This means that the weighted average is the sample obtained from N7 at tl plus the difference of both calibration samples. The weighting of the inaccuracies is shown in Table 4. Table 4: Calculation of the magnitude of the inaccuracies in the weighted average of N6

When calculating the maximum frequency error, it can be realised that there are two subsamples of the same type, namely N6's aging since tO. As one of the values is negative and the other positive, they cancel each other out. Thus, now Formula (1) gives the maximum frequency error based on a confidence level of ten sigma:

10 • _A/2² + (- 2)² + 3² + l² + 2² = 47[ppm]

In this example full correlation is assumed between both instances of N6's aging since tO. For a practical oscillator this correlation is high instead of full.

A result of the temperature compensation is that temperature stability is also shared over the network. For instance an individual node can start with an accurate frequency, but lose some accuracy as a result of a rising temperature. The network still has a good representation of this frequency and the node can sample it back.

This is advantageous when not all nodes experience the same temperature changes at the same time (sun / shade). It is also advantageous if some nodes have a better temperature compensation, or just because the temperature compensation error between different nodes is partially uncorrelated.

In accordance with one embodiment of the present invention a new link is established based on sufficient frequency accuracy of both transmitter and receiver.

Initial sufficient accuracy is ensured by for instance factory calibration and subsequent sufficient accuracy is ensured by sharing the reference frequency as explained above. When the receiver (or transmitter) does not have a sufficient initial accuracy to establish a normal link, another method needs to be used to establish a link.

In a multicast link various receivers can listen to the same transmitter. They can all sample the transmitter's frequency. It is also possible for a receiver to listen to a link that has no relevant data for the node, just to sample the frequency.

When a sufficient absolute accuracy for transmitting cannot be guaranteed, the transceiver can switch to a receive-only mode. This may result in new frequency samples that will allow transmitting within regulations again. This situation may arise when the network has been aging for a long time. This mechanism can also be used for a transceiver that has not been calibrated sufficiently possibly at certain temperature. The network will be the calibration source and from that moment on the transceiver can contribute to a slower aging process of the whole network.

The controller 313 may require the establishment of a specific radio link just to ensure accuracy. This link can then also be used to communicate data, other than reference frequencies or accuracy information. This data needs to be routed to the targeted nodes by a routing algorithm, which can be extended to take this into account.

The flow chart of Figure 10 illustrates the method as described above. In step 1001 the transceiver 300 receives a shared frequency, together with at least inaccuracy information and/or the origin of the inaccuracies, from another terminal operating in the network. In step 1003 the transceiver 300 measures a frequency difference between the received shared frequency and frequency provided by the oscillator 203. In step 1006 the weighted averages are determined using the received origin and/or magnitude of error of the received reference frequency and in step 1007 the weighted averages are allocated to the frequency references. In step 1009 the frequency provided by the oscillator is adjusted based on the received shared frequency and taking into account the weighting factor. The origin and/or magnitude of the frequency inaccuracy of the newly updated frequency is calculated and stored. In step 1011 the transceiver transmits the frequency provided by the oscillator 203 to other transceivers in the network. It is to be noted that step 1011 can also be performed at other positions in the process. The "magnitude of error of a shared frequency" refers to (in the example above) a sigma and not to an absolute frequency error. The magnitude of error of each reference frequency in the network is known (in the same way as shown in Table 1), and when one reference frequency

is updated using another shared frequency heaved (and its magnitude of error), the magnitude of error of the newly updated frequency f_new is calculated from the magnitude of error of f_ou and the magnitude of error of Received as done in Table 2 and corresponding equations.

In the embodiment described above, the frequency of the transmitter is sampled at the receiver. This gives the same information as if the transmitter would sample the frequency of the receiver. Consequently the result of the sampling operation can be communicated and used to adjust the transmitter reference oscillator as well. In addition, in some transceiver architectures it may indeed be possible to actually sample the frequency of the receiver at the transmitter.

The above-described transceiver configuration has primarily been invented with narrow-band transceivers and absolute frequency accuracy in mind. It can provide relative frequency accuracy as well. It can also provide absolute or relative timing accuracy. Therefore, it can also be applied in applications that require timing accuracy like pulse-based ultra-wideband (UWB) and protocol timing. When a node contains both a frequency reference oscillator and a timing reference oscillator their accuracies can be shared as well.

The teachings of the present invention can be applied basically in any of the numerous wireless (sensor) network transceivers where it is desirable to reduce power, form factor and/or cost. There are for instance many wireless sensor network applications within Consumer Lifestyle Solutions where the teachings of the invention can be applied. There are also activities within research exploring patient tracking and monitoring, both within hospitals and at patients' houses. These activities are relevant for the healthcare business sector and can use the teachings of the present invention.

The invention also relates to a computer program product that is able to implement any of the method steps as described above when loaded and run on computer means of the transceiver. The computer program may be stored/distributed on a suitable medium supplied together with or as a part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems.

The invention also relates to an integrated circuit that is arranged to perform any of the method steps in accordance with the embodiments of the invention. While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive, the invention being not restricted to the disclosed embodiment. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims.

In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that different features are recited in mutually different dependent claims does not indicate that a combination of these features cannot be advantageously used. Any reference signs in the claims should not be construed as limiting the scope of the invention.

Claims

C L A I M S

1. A method of adjusting reference frequency of a wireless transceiver (300), the method comprising:

- receiving (1001) at least one shared reference frequency together with at least one of the following: magnitude of error of the shared frequency and origin of the magnitude of error; allocating (1007) a weighting factor for the at least one shared frequency; and comparing (1009) the received at least one shared frequency with frequency provided by a frequency source of the transceiver (300) for adjusting the frequency provided by the frequency source by taking into account the weighting factor.

2. The method according to claim 1, further comprising determining (1006) the weighting factor using the at least one of the following: received magnitude of error of the shared frequency and received origin of the magnitude of error.

3. The method according to any one of the preceding claims, wherein several reference frequencies are received and weighting factors are allocated so that a sum of the weighting factors allocated to different transceivers sharing their reference frequencies equal 1 or dividing a weighting equation by the sum of the weighting factors.

4. The method according to claim 3, wherein the shared reference frequencies having higher magnitude of error are allocated lower weighting factors than shared frequencies having lower magnitude of error.

5. The method according to any one of the preceding claims, further comprising determining (1003) frequency difference between the at least one received shared reference frequency and the frequency provided by the frequency source of the transceiver (300).

6. The method according to any one of the preceding claims, further comprising changing individual weighting factors over time.

7. The method according to any one of the preceding claims, wherein the shared reference frequencies are obtained from other nodes in the network and comprise frequency subsamples with individual magnitude of errors.

8. The method according to claim 7, wherein the frequency subsamples' origin of the magnitude of error is due to aging of the reference frequency during certain period of time.

9. The method according to claim 7, wherein the frequency subsamples' origin of the magnitude of error is due to changing frequency information with other nodes in the network.

10. The method according to any of the preceding claims, wherein the at least one shared reference frequency is further received together with temperature indicating the temperature where the shared reference frequency was taken and using the correlation of shared reference frequencies taken at different temperatures to adjust the frequency provided by the frequency source.

11. A computer program product comprising instructions for implementing the steps of a method according to any one of claims 1 through 10 when loaded and run on computer means of a transceiver.

12. A transceiver (300) capable of adjusting frequency provided by its frequency source, the transceiver (300) comprising: a demodulator (209) for demodulating a signal comprising at least one shared reference frequency, the signal further comprising at least one of the following: magnitude of error of the shared frequency and origin of the magnitude of error; a controller (313; 611) for allocating a weighting factor for the at least one shared frequency; and a processor (207) for comparing the received at least one shared frequency with frequency provided by the frequency source of the transceiver (300) for adjusting the frequency provided by the frequency source by taking into account the weighting factor.

13. The transceiver of claim 11, wherein the demodulator (209) is further arranged to calculate frequency difference between the at least one shared frequency and the frequency provided by the frequency source.