The invention relates to a method for digital quartz temperature and drift compensation for a sleep timer of a NB-IoT device.

The invention also relates to an apparatus for digital quartz temperature and drift compensation for a sleep timer of a NB-IoT device.

NarrowBand Internet-of-things (NB-IoT) devices are rather new. These devices are typically connected to an IoT network (Internet of Things) and will be used/produced cheap and in high quantities.

Many IoT devices cannot be accessed physically once they were put into operation. These are e.g. sensors in street grounds, animals or at other not easily accessible places. Protecting and saving battery power of these devices is important because the battery cannot be recharged or

replaced. Furthermore, these devices wake-up only for a short time for sending or providing data to the network, but most of the time these devices are in a sleep mode. For some applications it is important that the NB-IoT devices provide and send their data at specified points in time and hence it is important to guarantee that these devices wake up at these specific and specified times. Therefore, NarrowBand Internet-of-Things (NB-IoT) devices comprise a sleep timer with a quartz crystal in order to wake up after long deep sleep periods and to send the required data. For IoT applications normally X-cut crystals are used for a 32 kHz sleep timer of a NB-IoT device.

It is known that X-cut crystals typically have a parabolic dependency on temperature, where the offset shows a

characteristic like y- = c _T (T— T ₀ ) ² + d ₀ ® df = c _T AT ² + d ₀ (eq.

1), where c _T is the temperature coefficient of the quartz crystal in the range of for example -0.025ppm ...-0.05ppm/K ² and the so-called "turnover temperature" To is in the range of for example 20...30°C. The do term describes a pseudo-static frequency offset (initial offset plus potential drift

contribution) , which is in the order of ±10% of the

temperature offset. The Temperature T is considered to be in the range of for example -55 ...+125°C (extended industrial range) .

Figure 1 shows the influence of the temperature coefficient c _T on the quartz frequency offset d _f . Figure 2 shows the influence of the turnover temperature To on the quartz frequency offset, whereas the turnover temperature is understood as the temperature at which the tangent to the parabola is parallel to the x-axis. And figure 3 shows the influence of static offset do on the quartz frequency offset.

Until now, the compensation of quartz temperature and drift compensation in temperature measurement setups was performed analogue, either with a quartz crystal with a controlled heating or as a "pull-circuit" for the quartz frequency.

For NarrowBand Internet-of-Things (NB-IoT) devices it is absolutely disadvantageous to use quartz crystals with a power consuming heating in order to compensate drifts due to the temperature dependence of the quartz crystal of the sleep timer.

It is therefore the objective of the invention to provide a method and an apparatus which allow exploiting the known characteristic of typically used quartz oscillators in such NB-IoT devices in order to compensate temperature and drift of the quartz crystal frequency due to temperature and to make the device cheaper, less power consuming and more flexible. Generally, it is desirable to find a way for an effective quartz crystal temperature and drift compensation in NB-IoT devices.

The objective of the invention will be solved by a method for digital temperature and drift compensation of a quartz crystal from its nominal frequency for a sleep timer used for a NarrowBand Internet of Things (NB-IoT) device, the method comprising the following steps:

- determining a temperature dependence of the quartz crystal frequency offset against an external reference, resulting in quartz crystal parameters,

- storing the quartz crystal parameters for further

processing,

- acquiring a temperature measured by a temperature sensor,

- calculating a deviation of the quartz crystal frequency offset due to its temperature dependence and the measured temperature, and

- generating compensation pulses for the sleep timer

according to the deviation of the quartz crystal frequency offset to adjust a counter value of the sleep timer of the NB-IoT device.

Before other preferred and advantageous embodiments of the inventive method are described, the theory of operation of such quartz crystals used for the sleep timer of a NB-IoT device and its temperature dependence are considered.

It is known that static offset as well as dynamic offset contributors have to be considered to be able to compensate their influences in NB-IoT devices, which use quartz

crystals in a sleep timer for counting and determine its active and idle periods.

Static offset contributors are considered first. In general, the measured temperature T, as well as the quartz crystal parameters c _T and do are inaccurate to some extent. The propagation of uncertainty caused by their individual uncertainties to the overall uncertainty d _f can be estimated using the variance formula:

(eq. 2)

Looking at each contributor separately allows the following simplifications :

It is possible to determine the maximum uncertainty budget for each parameter (temperature T, temperature coefficient c _T , and pseudo-static frequency offset do) of the quartz crystal. According to a global error range, for example 0.5ppm, that will be defined first, the individual

contribution from each parameter can be calculated and considered in the hardware design of the NB-IoT device.

In order to assess the overall quartz temperature and drift compensation, dynamic offset contributors have to be

considered as well.

The frequency offset d _f calculated according to equation (eq. 1) describes the momentary deviation at a given point in time. When integrated over a certain interval, the number of missing/excessive clock pulses during that interval can be obtained by: (eq. 6) .

So far, all calculations regarding uncertainties and

resolutions were done under the assumption of a static, yet uncertainly measured temperature and static quartz crystal parameters. In reality, though, temperature is a function of time, most simply modeled as linear dependency: T = T (t) = g _T · t + T _s (eq. 7) . Inserting equation (eq. 7) into equation (eq. 1) then yields:

Tc2+d0 (eq . 8 ) .

So, while the temperature is changing and sampled only at certain intervals At, a frequency offset is accumulated.

This is depicted in figure 4, which uses equation (eq. 8) as temporal temperature dependence. Figure 4 shows the

dependence of the frequency offset d _f (t) (dashed graph) from the temperature (dotted graph) , which is assumed to change linearly with time within the interval to ... t ₀ +At. The solid line shows the quartz frequency offset as a function of time. The accumulated error corresponds to the shaded region in figure 4 and is the bigger, the greater the sampling interval and the steeper the temperature gradient at that point of the curve. Frequency offset during intervals of constant temperature (t < to or t > t ₀ +At) can be fully compensated, so these constant contributions do not add to the sampling error e. Hence: e = P _reai ^~ P _appro x ⁼ —

(eq. 9) .

Resolving the integral will then lead to:

After various transformation this eventually yields: ^e = c _T (g _T t ₀ At ² + g¾At ³ + g _T At ² T _c (eq. 11).

The error calculated according to equation (eq. 11)

basically describes the offset made when compensation is only done with the values calculated at exactly the sample times without any interpolation. In other words, the

integral P _re ai to the offset-over-time curve is approximated by a sequence of rectangles. However, a better approximation is achieved by a piecewise trapezoid shape which yields:

gT 2 tO D t2+12cfT 2D t3 +2 GίOD tT + ?TD ί2Ta+D tT 2+dOD t (eq . 12).

Replacing now the term for P _apP rox in equation (eq. 9) with the new approximation from equation (eq. 12) then yields: e _a = c _T grAt ³ (eq. 13).

This remaining (in-compensable) error would be accrued in each sampling interval as long as the temperature is

changing. For a worst-case estimation a linear temperature change within the full operational range AT _max is considered. The duration of this full-range change depends on the

AT

thermal gradient g _T : D=— ^{º i} (eq. 14) and the number of

sampling intervals therefor is: n _s = (eg. 15) .

Hence, the total accrued error for this full-range

temperature sweep is the product of the number of steps and the in-compensable error per step as given in equation (eq.

For a constant temperature, the permissible offset of a defined global error range, for example 0.5ppm, from the static contributors would accumulate a total offset of e _t ot,stat = o .5*D (eq. 17) .

Now assuming that the total temporal offset according to equation (eq. 16) shall not exceed the static contribution following the relation can be derived: -Dc _T g _j At <0.5 D (eq.

18) which then translates (eq. 19) .

So, the sampling interval At should be chosen inverse proportionally to the assumed or measured temperature gradient g _T .

Depending on all the offset contributions described above, the real frequency may be either too slow or too fast with respect to the nominal frequency. Since the real frequency drives the sleep timer counter, this means that the counter value accumulated over a given interval is off by a number of ticks, and this value is either too low or too high. As described above it is possible to determine the frequency offset d(T) as a function of mainly the measured temperature T. With this knowledge it becomes possible to compensate the counter bias of the sleep timer. Basically: C _T = f _T At =

In order to compensate the bias, it is necessary to

add/subtract C _0ffS ticks during the interval At. This is achieved by providing a clock d _c derived from the offset clock f _T like this: (eq. 21) so that C _comv

^p = d _c At (eq. 22) and hence

Equation (eq. 21) may seem tautological, but it reflects the fact that although the frequency offset d _f is known the only clock that can be used for counting is the temperature- influenced quartz clock of the sleep timer. Therefore, a fractional divider according to figure 5 is used and allows iic

generating exactly this: f _d = / _{dec+inc (} eq. 24).

Combining equation (eq. 21) and equation (eq. 24) then

. . r/ ₀ for d _f ³ 0

yields : inc = \d _f \ (eq . 25 ) and dec = \ _r , _ _n (eq. 26).

^{1 /1} (/o + 2 df for df < 0

After the derivation of the theory underlying the inventive method it is preferred in one embodiment that the

temperature is acquired at fixed temperature sampling

intervals. This is the simplest form of the inventive method and best suited for environments with mostly stable

temperatures, because only a static frequency offset has to be compensated.

In another preferred embodiment of the inventive method, the temperature is acquired at adaptive temperature sampling intervals. This means that the temperature sampling interval is dynamically re-selected at each new sampling point to minimize the in-compensable error caused mostly by

temperature fluctuations during the sampling intervals. In practical scenarios the parameters used to determine the next sampling interval (temperature differences, offsets) have less resolution than otherwise used for these types of data. This is the case to save power and area in

implementation without sacrificing the ability to control the interval selection.

In a further embodiment of the inventive method, the

temperature sampling interval is selected based on up to four temperature ranges. This will be done by specifying three limits given as absolute differences from the turnover temperature To of the used quartz crystal of the sleep timer. The effective number of ranges can be decreased by

specifying identical values for the limits, e.g. when setting two limits to the same value the number of ranges is decreased to three instead of four. The ranges defined by the temperature difference limits should be chosen according to the derivative and/or the result of equation (eq. 1) . The range around the turnover temperature To can be made wider because both the absolute result (and hence the offset to be compensated) is low as well as the susceptibility to

temperature changes, so the overall in-compensable error is limited. More outer ranges should therefore be smaller. In another embodiment of the inventive method, the

temperature sampling interval is selected based on a

temperature difference between a previous and a current sampling point. This will be used when operating in an environment with dynamic temperature changes or high

temperature gradients and has the advantage of a good balance between frequency of measurement which also consumes some power and accuracy of compensation.

In another embodiment of the inventive method, the

temperature sampling interval is selected based on a

predicted offset for compensation at a time of the previous temperature measurement and a calculated offset of the previous interval determined at a current measurement, multiplied with a length of the previous sampling interval. This leads in a residual offset which needs to be backward compensated at the beginning of the next sampling interval. This will be used when again there is a high variance in temperature fluctuation and has the advantage of minimizing the amount of post-compensation at each new sampling time.

In a further embodiment of the inventive method, the

temperature sampling interval is selected based on

temperature difference between a previous and a current sampling point and a predicted offset for compensation at a time of the previous temperature measurement and a

calculated offset of the previous interval determined at a current measurement, multiplied with a length of the

previous sampling interval. This is a combination of the embodiments of the inventive method claimed according to claims 5 and 6. This will be used when there is a more unknown thermal environment and has the advantage if

providing good compensation under arbitrary conditions. And in another further embodiment of the inventive method, the external reference is a radio cell the NB-IoT is communicating with. This has a great advantage, because the clock of the radio cell is very accurate, and hence it is preferred to use the clock of the radio cell to adjust the clock of the sleep timer of the NB-IoT device during its active periods in order to compensate quartz temperature and drifts during the much longer sleep periods.

The advantage of the inventive method can be seen therein, that the method provides a possibility to compensate drift characteristic of the quartz crystal due to its temperature dependence. As in sleeping periods of the NB-IoT device only the pulses of the quartz crystal are available, only the quartz frequency can be used as reference. By compensating the quartz frequency according to the known behavior any drift and deviation can be compensated and corrected. In the active phase the NB-IoT can get a reference time indication from the connected radio cell. It is possible to determine the quartz crystal frequency against the frequency of the radio cell in order to get the true difference value. With this knowledge the curvature and displacement of the quartz crystal frequency can be adjusted.

The object of the invention will also be solved by an apparatus for digital quartz temperature and drift

compensation of a sleep timer of a NarrowBand Internet of Things (NB-IoT) device comprising a temperature sensor connected to a temperature acquisition block, an offset calculation block connected to an internal storage block and the temperature acquisition block for calculating and determining an offset of a quartz frequency of the quartz crystal for the sleep timer due to its temperature dependence based on measured temperature values by the temperature sensor, whereas the sleep timer (11) is

connected to the apparatus for digital quartz temperature and drift compensation, the apparatus further comprising an adjustment generator providing tuning pulses for the sleep timer of the NB-IoT device in a sleep phase according to the measured and determined temperature dependence of the frequency offset of the sleep timer, whereas the adjustment generator is connected to the offset calculation block and a controlling finite state machine (FSM) , and the FSM controls the temperature acquisition block, the internal storage block and the adjustment generator.

The quartz temperature and drift compensation (QTDC) is responsible to equalize deviations of for example a 32 kHz crystal of a sleep timer from its nominal frequency, based on acquired temperature measurements and information on the frequency offset. The adjustment generator then generates compensation pulses for the sleep timer, whereas the pulses are used there to adjust the counter value of the sleep timer .

The finite state machine is used to control the compensation process. It determines the sampling interval based on register settings, acquires the temperature and controls the generation of compensation impulses.

In an embodiment of the inventive apparatus for digital quartz temperature and drift compensation for a sleep timer of a NarrowBand Internet of Things (NB-IoT) device, the adjustment generator is a fractional divider. A fractional divider also called a clock divider or scaler or prescaler, is a circuit that takes an input signal of a frequency, fi _n , and generates an output signal of a frequency: f _out = fi _n * m/n, where m and n are integers. The fractional divider provides the tuning pulses for the sleep timer of the NB-IoT device in a sleep phase according to the measured and determined temperature dependence of the frequency offset of the sleep timer. In comparison to already known NB-IoT devices with no possibility of quartz temperature and drift compensation this has the advantage that the device can stay longer in a low-power deep sleep mode, since the wake-up time can be set more closely to the actual time when the device needs to be operative. It is no longer necessary to add a big safety margin to the wake-up time to compensate for frequency uncertainty.

In another embodiment of the inventive apparatus, the temperature sensor comprises a digital interface. This has the advantage of using off-the-shelf sensors that are widely available with e.g. I2C or SPI interfaces.

The invention will be explained in more detail using

exemplary embodiments.

The appended drawings show

Fig. 1 The influence of the temperature coefficient c _T on the quartz frequency offset;

Fig. 2 The influence of the turnover temperature To on the quartz frequency offset;

Fig. 3 The influence of static offset do on the quartz frequency offset;

Fig. 4 The influence of temperature sample interval At on quartz frequency offset; Fig. 5 A Fractional divider;

Fig. 6 Workflow of the inventive method;

Fig. 7 Inventive apparatus for digital quartz temperature and drift compensation of a sleep timer of a NB- IoT device.

Figure 6 shows the workflow of the inventive method for digital temperature and drift compensation of a quartz crystal from its nominal frequency used for a sleep timer of a NarrowBand Internet of Things (NB-IoT) device. In a first step it will be decided if the quartz crystal parameters of the quartz crystal used for the sleep timer are known. If these parameters are not known, the parameters have to be determined by measuring separately. But this is not

advantageous for mass production devices. The parameters can also be determined against an external reference, like a connected radio cell as mentioned earlier.

If the quartz crystal parameters are known or have been determined and stored for further processing, a temperature measured by a temperature sensor is acquired. With the acquired temperature a deviation of the quartz crystal frequency offset due to its temperature dependence and the measured temperature is determined and compensation pulses for the quartz crystal of the sleep timer according to the deviation of the quartz crystal frequency offset are

calculated in order to adjust a counter value of the sleep timer of the NB-IoT device. Therefore, a fractional divider is used in order to generate pluses with a frequency that is a fraction of the reference clock, hence the clock of the quartz crystal for the sleep timer or the time indication of a radio cell to which the NB-IoT device is connected to. Figure 7 shows an exemplary setup of the inventive apparatus for digital quartz temperature and drift compensation of a sleep timer 11 of a NB-IoT device 1. The apparatus comprises a temperature sensor 2 connected to a temperature

acquisition block 3, an offset calculation block 4 connected to an internal storage block 5 and the temperature

acquisition block 3 for calculating and determining an offset of a quartz frequency of the sleep timer 11 due to its temperature dependence based on measured temperature values by the temperature sensor 2, an fractional divider 6 providing tuning pulses for the sleep timer 11 of the NB-IoT device 1 in a sleep phase according to the measured and determined temperature dependence of the frequency offset of the sleep timer 11. The fractional divider 6 is connected to the offset calculation block 4 and a controlling finite state machine (FSM) 7, and the FSM 7 controls the

temperature acquisition block 3, the internal storage block 5 and the adjustment generator 6.

Method and apparatus for digital quartz temperature and drift compensation for a sleep timer of a NB-IoT device

List of Reference Signs

1 Digital quartz temperature and drift compensati of a NB-IoT device

2 temperature sensor

3 temperature acquisition block

4 offset calculation block

5 internal storage block

6 adjustment generator, e.g. a fractional divider

7 finite state machine (FSM) 7

11 sleep timer