GB2367981A

GB2367981A - Transmission power control in a frequency hopping RF system

Info

Publication number: GB2367981A
Application number: GB0024680A
Authority: GB
Inventors: David Smith
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2000-10-09
Filing date: 2000-10-09
Publication date: 2002-04-17
Also published as: WO2002032006A2; GB0024680D0; WO2002032006A3; AU2002223587A1

Abstract

The signal levels in a hopping channel of a frequency hopped system are measured over a predetermined time period and a pattern in the measured signal level is identified. Future signal levels are predicted on the basis of the identified pattern and the RF output power level per carrier is controlled on the basis of the predicted signal levels.. Signal levels are monitored burst by burst to identify pattern which can be extrapolated to predict the future signal levels.

Description

RF CHANNEL POWER CONTROL FIELD OF THE INVENTION The present invention relates to RF channel power control in communications systems.

BACKGROUND OF THE INVENTION Reducing the transmit power of a radio (RF) channel results in a corresponding reduction in the interference that is presented to other channels.

Dynamic power control is a well known and widely used technique. It is also standard practice for radio channels to frequency hop. This decorrelates the channels against fading, and in some case interference.

For example, see"Digital Communication"by John G.

Proakis, Third Edition, 1995 pages 729-743.

Conventional power control solutions apply a single power value to the radio channel as a whole, this being used for all frequencies within the hopping sequence.

However, multi-path (Rayleigh) fading resulting from the phase sum over paths of all rays taken by the radio signal is dependent on the wavelength of the carrier, and thus the path-loss differs from one frequency to another. Indeed this is one reason why frequency hopping is used as a defense against fading (see"the GSM system for mobile communications", Mouly & Pautet, section 4.2. 2.2). this can affect the strength of C and I independently, and thus the C/I ratio can differ for each frequency used in the hopping sequence.

In power control as commonly practised (for example according to the GSM standard), the required

power value is applied for a period of time which is many times the length of the frame, and indeed the hopping sequence. (In GSM a power control cycle [SACCH period] is"480ms, whereas the frame rate is 4. 615mS. ie a typical hopping sequence will rotate through each carrier order of tens of times per power control order), i. e. each SACCH period reports on 480/4. 615 = approx. 104 frames. If the Hopping sequence is a cyclic pattern comprising 4 carrier frequencies, each frequency will be used 26 times per SACCH period.

A transmit power level set to give the required nominal carrier/interference ratio C/I on the most disturbed burst (s) on the channels results in unnecessary high C for all other bursts on the chanel.

In an interference limited network a higher value of C, by definition, represents a higher value of I presented to other channels. If the carrier C value is unnecessarily high, the interference I it presents to other channels is correspondingly unnecessarily high.

As noted above, a principle feature of hopping is to decorrelate interference. This means that the interference level found on the channel varies burst by burst.

However, traditional power control works by reacting to (ie following) the measured interference.

However since the SACCH period reporting frequency is many times longer than the hopping frequency, power control according to the standard is totally unable to

track the C/I variations which occur on hopping channels frame by frame as a result of the C/I variations due to the different frequencies as explained above. Rather, power control as practiced today (and standardised) simply tracks the mean characteristics of the channel line 15.

SUMMARY OF THE PRESENT INVENTION According to the present invention, there is provided a method of controlling output power of an RF channel in a frequency hopping RF system having plurality of carriers, the method comprising: measuring signal levels in the channel over a predetermined time period; identifying a pattern in the measured signal level; predicting future signal levels on the basis of the identified pattern; and controlling the RF output power level on the basis of the predicted signal levels.

It is emphasised that the term"comprises"or "comprising"is used in this specification to specify the presence of stated features, integers, steps or components, but does not preclude the addition of one or more further features, integers, steps or components, or groups thereof.

DESCRIPTION OF THE PREFERRED EMBODIMENT The preferred embodiment of the invention can be described in the context of a cyclic hopping channel of a cellular radio system constructed according to the GSM standard. It will be readily apparent, however, that the techniques described are also applicable to other systems. In an embodiment, the channel is analysed in order to identify a pattern in burst by burst received signal. This pattern can then be extrapolated to predict the approximate transmit power for future bursts to come, per hopping channel. In essence, the preferred embodiment of the present invention analyses the characteristics of the channel

burst by burst, rather than looking at some aggregate characteristics reported per SACCH period or similar cycle that is substantially slower than the hopping mechanism, said analysis identifying any recurring pattern in the signal characteristics that is due to the effect of the hopping sequence of the channel on the channel's C, and the interaction between the hopping channel and (independently) hopping interference sources. Further, the embodiment then extrapolates this pattern to predict the likely characteristics of subsequent bursts, and derives from this power control values than can be optimised to the propagation characteristics of each frequency in the hopping sequence individually, rather than having to respond simply to a mean or worst case value.

Since the characteristics of the channel are the sum of the characteristics of each burst on the channel, and since the characteristics of the network are the sum of the characteristics of the channels comprising the network, improving the resolution of the power control mechanism in this manner advantageously improves the (spectral efficiency) performance of channels individually and indeed the entire network.

Burst by burst signal level on a channel is monitored for long enough to identify a recurring pattern using, for example, a"bounded comb"or Kalman filter. A simple comb implementation that can be described in a straightforward manner is to look at the signal strength on each n'th burst, where n is the number of carriers in the hopping sequence, and

separately look at the strength of each (n+l)'th etc i. e. if the hopping sequence is ABC DAB C DAB C D A B....

To look only at the A's. Separately at all the B's, separately at all the C's... Such a solution is

well motivated in theory since the signal strength is defined primarily by the path loss affecting the frequency used for each burst, and while the path loss will be affected by fading, this will typically vary slowly relative to the frame rate (but similar to or faster than the SACCH period).

A simple implementation might assume that the next burst to use A will have very similar signal strength to the last burst on A, and this will be found to be true within a reasonable confidence and accuracy tolerance. An enhanced embodiment might filter a rolling set of the last j bursts per frequency, improving confidence, however j should not be too large otherwise transient response to fading dips will be compromised. Another enhanced implementation might predict that A (next) will be of signal strength A (current)-A (last)), that is to say assuming that any recent trend in the rate of change in the signal strength will continue, and again such a strategy will (statistically) be correct more often that it is wrong, and will be a quantitatively"less wrong"more often than"more wrong", i. e. represents a net improvement.

In a simple implementation, such a comb filter can be"seeded"since the length of the hopping sequence is known.

However in more complex implementations, which also try to react to systematic variations in the interference level, the modulo of the signal pattern may not be known (since interference often comes from several surrounding cells, which may have different hopping sequences). Nevertheless, the pattern can be found by (as one simple implementation) an"exhaustive search", i. e. trial and error use of combs of all lengths from 2 to x, with some metric assigned to evaluate the correlation of the bursts dropping through

the comb. The comb length that gives the best correlation is then used, and the others abandoned.

More algorithmic solutions include the use of hashing algorithms and variations of the SW implementations of Viterbi and Fast Fourier Transform algorithms etc ( There is a strong similarity in particular with th FFT problem, where the existence of a regularly recurring pattern is deduced from a long sequence of discrete samples).

A preferred embodiment is to monitor the channel continuously, a repeating pattern being identified, thereafter this pattern is tracked.

Since the GSM air interface is synchronised to a very stable clock, the pattern of the interplay between interfering hopping sequences is extremely stable, even though the signal levels vary over time due to fading etc..

The level of signal on a given frequency will vary due to fading, but typically this will occur at a rate orders of magnitude slower than the hopping sequence repetition rate, and is thus easy to track.

The simplest case is where hopping sequences in interfering cells have equal length (equal number of ARFCN in the HFS), such that the pattern will be the same length in frames. Note that the interfering cells do not have to be synchronised-the alignment does not matter, as it will be stable.

If the sequences are of different lengths, the interfering cycles will"walk"across each other and are only guaranteed to recur every x frames, where x is the lowest common multiple of the respective lengths of the HFS's concerned. For example, a cyclic hopping sequence of 3 ARFCN's walking across a sequence with 4 ARFCN's will repeat regularly an interference pattern which is 12 frames long.

It might be reasonably expected that even a simple and naive algorithm to lock to this in < < 200mS. (a "na ve"algorithm being an algorithm that does not have any hints to guide it). A seeded algorithm (given a priori knowledge, for example the length of the

sequences,"x"which is known in the BSC) might typically lock in 100mS, in theory after just"x" frames.

Once lock is acquired, the sequence is tracked and held, adapting dynamically to changes due to, for example, fading variations in the interfering propagation multipath and DTX (discontinuous transmission) on interfering channels.

A frame can be destroyed if more than a few bits are lost. This means that the FER (frame error rate) is primarily determined by the interference level on the most disturbed bits in the burst, and the most disturbed burst in the hopping sequence. Note that in a non synchronised network the interference level is not necessarily constant throughout a burst.

The result of the above is that a"mean"BER (bit error rate) reported for the channel overall is misleading in the case that the channel is hopping orthogonally to its interferers.

Rather than detecting the mean signal level and adapting power to give a margin over that mean, it is proposed, and advantageous, to report the PEAK signal value per burst. Working in this way the likelihood of a given BER per burst will be substantially reduced.

A preferred embodiment weights the"most important"bits in the burst, due to the disproportionate effect on speech quality if these are degraded. The"most important"bits in the burst are defined in the GSM standards.

The two techniques described above are complementary and used in combination will minimise BER at the same time as minimising interference presented.

By adding further modulos in the interferences sequencing, the system can lock to and intelligently power control against decorrelated channels eg HR sharing the same hopping sequences. For example. If the algorithm can predict when bursts on the two orthogonal hopping sequences will not collide, interference can be predicted to be low and low power can be used, whereas high power can be used when they will collide.

Claims

CLAIMS : 1. A method of controlling output power of an RF channel in a frequency hopping RF system having plurality of carriers, the method comprising: measuring signal levels in the channel over a predetermined time period; identifying a pattern in the measured signal level; predicting future signal levels on the basis of the identified pattern; and controlling the RF output power level per carrier on the basis of the predicted signal levels.
2. A method as claimed in claim 1, wherein the RF channel comprises a series of data bursts, and the measurement of signal levels in the channel occurs for each burst.
3. A method as claimed in claim 1 or 2, wherein future signal levels are predicted by extrapolating results for measured signal levels.
4. A method as claimed in claim 1, wherein the RF channel concerned has a predetermined hopping sequence, and the pattern in the measured signal level is identified on the basis of the predetermined hopping sequence.
5. A method as claimed in claim 1, wherein the measured signal levels are integrated over predetermined regular intervals.
6. A method as claimed in claim 1, wherein the signal level for the channel is measured at regular intervals for each frequency of the hopping sequence of the channel.

,
7. A method as claimed in claim 1, wherein the signal level is measured for a predetermined number of bursts on the channel.
8. A method as claimed in claim 1, wherein

trends in each of the carriers characteristics are identified and extrapolated for predicting future signal levels.
9. A method as claimed in claim 1, wherein peak signal levels are measured.
10. A method as claimed in claim 1, wherein average signal levels are measured.
11. Apparatus for controlling output power of an RF channel in a frequency hopping RF system having plurality of carriers, the apparatus comprising: measurement means for measuring signal levels in the channel over a predetermined time period; identification means for identifying a pattern in the measured signal level; prediction means for predicting future signal levels on the basis of the identified pattern; and control means for controlling the RF output power level per carrier on the basis of the predicted signal levels.
12. Apparatus as claimed in claim 11, wherein the RF channel comprises a series of data bursts, and the measurement means is operable to measure signal levels in the channel for each such burst.
13. Apparatus as claimed in claim 11 or 12, wherein the predication means is operable to predict future signal levels by extrapolating results for measured signal levels.
14. Apparatus as claimed in claim 11, wherein the RF channel concerned has a predetermined hopping sequence, and identification means is operable to identify the pattern in the measured signal level on the basis of the predetermined hopping sequence.
15. Apparatus as claimed in claim 11, wherein the measurement means are operable to integrate the measured signal levels over predetermined regular

intervals.
16. Apparatus as claimed in claim 11, wherein the measurement means is operable to measure the signal level for the channel at regular intervals for each frequency of the hopping sequence of the channel.
17. Apparatus as claimed in claim 11, wherein the measurement means is operable to measure the signal level for a predetermined number of bursts on the channel.
18. Apparatus as claimed in claim 11, wherein the identification means is operable to identify trends in each of the carriers characteristics and the prediction means is operable to extrapolate identified trends for predicting future signal levels.
19. Apparatus as claimed in claim 11, wherein the measurement means is operable to measure peak signal levels.
20. Apparatus as claimed in claim 11, wherein the measurement means is operable to measure average signal levels.