, Title of the Invention
2 Burst-by-burst Adaptive Single-carrier Data Transmission
3 1 Background of the Invention
4 The invention relates to data transmission, more specifically to transmission in packets or bursts.
5 In contrast to the burst-by-burst reconfigurable wideband multimedia transceivers described in this doc- e ument, the term statically reconfigurable found in this context in the literature refers to multimedia
7 transceivers that cannot be near-instantaneously reconfigured. More explicitly, the previously proposed β statically reconfigurable video transceivers were reconfigured on a long-term basis under the base sta-
9 tion's control, invoking for example in the central cell region - where benign channel conditions prevail ιo - a less robust, but high-throughput modulation mode, such as 4 bit/symbol Quadrature Amplitude Mod- ιι ulation (16QAM), which was capable of transmitting a quadruple number of bits and hence ensured a i2 better video quality. By contrast, a robust, but low-throughput modulation mode, such as 1 bit/symbol is Binary Phase Shift Keying (BPSK) can be employed near the edge of the propagation cell, where hostile propagation conditions prevail. This prevented a premature hand-over at the cost of a reduced video is quality. is The philosophy of the fixed, but programable-rate proprietary video codecs and statically reconfigurable
17 multi-mode video transceivers presented by Streit et al in References [l]-[4] was that irrespective of is the video motion activity experienced, the specially designed video codecs generated a constant number
19 of bits per video frame. For example, for videophony over the second-generation Global System of
20 Mobile Communications known as the GSM system at 13 kbps and assuming a video scanning rate of
21 10 frames/s, 1300 bits per video frame have to be generated. Specifically, two families of video codecs
22 were designed, one refraining from using error-sensitive run-length coding techniques and exhibiting the
23 highest possible error resilience and another, aiming for the highest possible compression ratio. This
2 fixed-rate approach had the advantage of requiring no adaptive feedback controlled bitrate fluctuation
25 smoothing buffering and hence exhibited no objectionable video latency or delay. Furthermore, these
26 video codecs were amenable to video telephony over fixed-rate second-generation mobile radio systems,
27 such as the GSM.
2β The fixed bitrate of the above proprietary video codecs is in contrast to existing standard video codecs,
29 such as the Motion Pictures Expert Group codecs known as MPEG1 and MPEG2 or the ITU's H.263
30 codec, where the time-variant video motion activity and the variable-length coding techniques employed
1 result in a time-variant bitrate fluctuation and a near-constant perceptual video quality This time vaπant 2 bitrate fluctuation can be mitigated by employing adaptive feed-back controlled buffering which poten- 3 tially increases the latency or delay of the codec and hence it is often objectionable tor example in mter- 4 active videophony The schemes presented by Streit et al in References [l]-[4] result in slightly \aπable 5 video quality at a constant bitrate, while refraining from employing buffering, which again would result 6 in latency in interactive videophony A range of techniques, which can be invoked, in order to render the 7 family of variable-length coded, highly bandwidth-efficient, but potentially error-sensitive class of stan- 8 dard video codecs, such as the H 263 arrangement, amenable to error-resilient
interactive 9 wireless multimode videophony was summarised in [5] The adaptive video rate control and packetisa- 0 tion algorithm of [5] generates the required number of bits for the burst-by-burst adap e transceiver, 1 depending the on the capacity of the current packet, as determined by the current modem mode Fur- 2 ther error-resilient H 263-based schemes were contrived for example by Farber, Steinbach and Girod 3 at Erlangen University [6], while Sadka, Eryurtlu and Kondoz [7] from Surrey University proposed a 4 range of improvements to the H 263 scheme Following the above portrayal of the pπor art in both video 5 compression and statically reconfigurable narroband modulation, let us now consider the philosophy of 6 wideband burst-by-burst adaptive quadrature amplitude modulation (AQAM) in more depth 7 In burst-by-burst adaptive modulation a higher-order modulation scheme is invoked, when the channel 8 is favourable, in order to increase the system's bits per symbol capacity and conversely, a more robust
49 lower order modulation scheme is employed, when the channel exhibits infeπor channel quality, in order so to improve the mean Bit Error Ratio (BER) performance A practical scenaπo, where adaptive modula- 5i tion can be applied is, when a reliable, low-delay feedback path is created between the transmitter and
52 receiver, for example by supeπmposing the estimated channel quality perceived by the receiver on the
53 reverse-direction messages of a duplex interactive channel The transmitter then adjusts its modem mode
54 according to this perceived channel quality
55 Recent developments in adaptive modulation over a narrow-band channel environment have been pise oneered by Webb and Steele [9], where the modulation adaptation was utilized in a Digital European 57 Cordless Telephone - like (DECT) system The concept of vaπable rate adaptive modulation was also 5β advanced by Sampei et al [12, 17], showing promising advantages, when compared to fixed modula-
59 tion in terms of spectral efficiency, BER performance and robustness against channel delay spread In
60 another paper, the numeπcal upper bound performance of adaptive modulation in a slow Rayleigh flat- 6i fading channel was evaluated by Torrance et α/[10] and subsequently, the optimization of the switching
62 threshold levels using Powell minimization was used in order to achieve a targeted performance [ 1 1 , 18]
63 In addition, adaptive modulation was also studied in conjunction with channel coding and power control
6 techniques by Matsuoka et al [12] as well as Goldsmith et al [ 13]-[15] es In the narrow-band channel environment, the quality of the channel was determined by the short term
66 Signal to Noise Ratio (SNR) of the received burst, which was then used as a cπteπon in order to choose
67 the appropπate modulation mode for the transmitter, based on a list of switching threshold levels, /„ [9, ββ 10] However, in a wideband environment, this cπtenon is not an accurate measure for judging the quality
69 of the channel, where the existence of multi-path components produces not only power attenuation of the
70 transmission burst, but also intersymbol interference Subsequently, a new cπteπon has to be defined to 7i estimate the wideband channel quality in order to choose the appropπate modulation scheme
72 2 Summary of the Invention
73 Particular and preferred aspects of the invention are set out in the accompanying independent and depen-
74 dent claims Features of the dependent claims may be combined with those of the independent claims as
75 appropπate and in combinations other than those explicitly set out in the claims
76 The performance benefits of burst-by-burst adaptive modulation are descπbed, employing a higher-order modulation scheme, when the channel is favourable, in order to increase the system's bits per symbol
78 capacity and conversely, invoking a more robust, lower order modulation scheme, when the channel
79 exhibits infeπor channel quality It is shown that due to the descπbed adaptive modem mode switch-
80 mg regime a seamless multimedia source-signal representation quality - such as video or audio quality - et versus channel quality relationship can be established, resulting in a near-unimpaired multimedia source- 82 signal quality πght across the operating channel Signal-to-Noise Ratio (SNR) range The main advan- β3 tage of the descπbed technique is that irrespective of the prevailing channel conditions, the transceiver 84 achieves always the best possible source-signal representation quality - such as video or audio quality - by es automatically adjusting the achievable bitrate and the associated multimedia source-signal representation βe quality in order to match the channel quality expenenced This can achieved on a near-instantaneous or 87 burst-by-burst adaptive basis under given propagation conditions in order to cater for the effects of path- 8β loss, fast-fading, slow-fading, dispersion, co-channel interference, etc Furthermore, when a mobile is 89 roaming in a hostile out-doors - or even hilly terrain - propagation environment, typically low-order, go low-rate modem modes are invoked, while in benign indoor environments predominantly the high-rate, 91 high source-signal representation quality modes are employed
92 3 Brief Description of the Drawings
93 For a better understanding of the invention and to show how the same may be carried into effect reference
94 is now made by way of example to the accompanying drawings, in which:
95 List of Figures
96 1 Signalling scenarios in adaptive modems between a Mobile Station (MS) and a Base
97 Station (BS)
9β 2 Reconfigurable transceiver schematic
99 3 Normalized channel impulse response for the COST 207 four-path Typical Urban channel.
100 4 Transmission burst structure of the FMA1 non-spread data burst mode of the FRAMES ιoι proposal
102 5 Adaptive burst-by-burst modem in operation for an average channel SNR of 20dB, where
103 the modulation mode switching is based upon the SNR estimate at the output of the
10 equaliser, using the channel parameters defined in Table 1 ιo5 6 PDF of the adaptive modem being in a particular modulation mode versus channel SNR.
106 7 Transmission FER (or packet loss ratio) versus Channel SNR comparison of the four
10 fixed modulation modes (BPSK, 4QAM, 16QAM, 64QAM) and that of the adaptive toe burst-by-burst modem (AQAM). AQAM is shown with a realistic one TDMA frame ιo9 delay between channel estimation and mode switching, and also with a zero delay version no for indicating the upper bound performance. The channel parameters were defined in in Table 1 ιi2 8 Transmission FER (or packet loss ratio) versus Channel SNR comparison of the four
,13 fixed modulation modes (BPSK, 4QAM, 16QAM, 64QAM) with 5% FER switching and
114 adaptive burst-by-burst modem (AQAM). AQAM is shown with a realistic one TDMA us frame delay between channel estimation and mode switching, and a zero delay version
He is included as an upper bound. The channel parameters were defined in Table 1 ιi 9 Video bitrate versus channel SNR comparison of the four fixed modulation modes (BPSK, us 4QAM, 16QAM, 64QAM) and adaptive burst-by-burst modem (AQAM). AQAM is i,9 shown with a realistic one TDMA frame delay between channel estimation and mode
,20 switching, and also as a zero delay version for indicating the upper bound. The channel
,21 parameters were defined in Table 1
Decoded video quality (PSNR) versus channel SNR compaπson of the four fixed modulation modes (BPSK, 4QAM, 16QAM, 64QAM) with 5% transmission FER switching and that of the adaptive burst-by-burst modem (AQAM) AQAM is shown with a realistic one TDMA frame delay between channel estimation and mode switching, and a zero delay version tor indicating the upper bound The channel parameters were defined in Table 1 Decoded video quality (PSNR) versus channel SNR for the adaptive burst-by-burst modem (AQAM) AQAM is shown with a realistic one TDMA frame delay between channel estimation and mode switching, and a zero delay version indicating the upper bound Results are shown for three video sequences using the channel parameters that were defined in Table 1 Decoded video quality (PSNR) versus transmission FER (or packet loss ratio) comparison of the four fixed modulation modes (BPSK, 4QAM, 16QAM, 64QAM) and that of the adaptive burst-by-burst modem (AQAM) AQAM is shown with a realistic one TDMA frame delay between channel estimation and mode switching, and a zero delay version indicating the upper bound The channel parameters were defined in Table 1 Transmission FER (or packet loss ratio) versus Channel SNR compaπson of the fixed BPSK modulation mode and the adaptive burst-by-burst modem (AQAM) for the three sets of switching thresholds descπbed m Table 4 AQAM is shown with a realistic one TDMA frame delay between channel estimation and mode switching The channel parameters were defined in Table 1 Video bitrate versus channel SNR compaπson for the adaptive burst-by-burst modem (AQAM) with a realistic one TDMA frame delay between channel estimation and mode switching for the three sets of switching thresholds as descπbed in Table 4 The channel parameters were defined in Table 1
,47 4 Detailed Description
us 4.1 General Introduction to Adaptive Modem Mode Signalling Scenarios
,49 AQAM transmission parameter adaptation is an action of the transmitter in response to time-varying i5o channel conditions It is only suitable tor duplex communication between two stations, since the trans- i5i mission parameter adaptation relies on some form of channel estimation and signalling In order to
152 efficiently react to the changes in channel quality, the following steps have to be taken
153 • Channel quality estimation In order to appropπately select the transmission parameters to be
154 employed for the next transmission, a reliable prediction of the channel quality duπng the next ,55 active transmit timeslot is necessary
,56 • Choice of the appropriate parameters for the next transmission Based on the prediction of the
157 expected channel conditions duπng the next timeslot, the transmitter has to select the appropπate
,58 modulation schemes for the subcarπers
,59 • Signalling or blind detection of the employed parameters The receiver has to be informed, as
,60 to which set of demodulator parameters to employ for the received packet This information can
,6t either be conveyed within the packet, at the cost of loss of useful data bandwidth, or the receiver
,62 can attempt to estimate the parameters employed at the transmitter by means of blind detection
,63 mechanisms
16 Depending on the channel characteπstics, these operations can be performed at either of the duplex
165 stations, as shown in Figure 1 If the channel is reciprocal, then the channel quality estimation tor each
166 link can be extracted from the reverse link, and we refer to this regime as open-loop adaptation In this
16 case, the transmitter needs to communicate the transmission parameter set to the receiver (Figure 1(a)),
168 or the receiver can attempt blind detection of the transmission parameters employed (Figure 1 (c))
,69 If the channel is not reciprocal, then the channel quality estimation has to be performed at the receiver
, 0 of the link In this case, the channel quality measure or the set of requested transmission parameters is
,71 communicated to the transmitter in the reverse link (Figure 1(b)) This mode is referred to as closed-loop
,72 adaptation
, 3 4.2 A Specific Embodiment of a Video Transceiver
, The schematic of the whole system is depicted in Figure 2 In the descπbed system the wideband channel-
175 induced degradation is combated not only by the employment of adaptive modulation but also by equal-
176 ization, where the equalization process will eliminate most of the intersymbol interference based on a
,77 Channel Impulse Response (CIR) estimate deπved using the channel sounding midamble and conse-
, 8 quently, the signal to noise and residual interference ratio at the output of the equalizer is calculated
,79 We note, however that the above adaptive methodology can also be extended to employing burst-by-
,80 burst adaptive channel coding associated with different-strength error correction codecs in the different is, transceiver modes of operation
,82 4.3 Channel quality metrics
183 The most reliable channel quality estimate is the bit error rate (BER), since it reflects the channel quality,
184 irrespective of the source or the nature of the quality degradation
185 Firstly, the BER can be estimated with a certain granulaπty or accuracy, provided that the system entails
186 a channel decoder or - synonymously - Forward Error Correction (FEC) decoder employing algebraic
187 decoding las Secondly, if the system contains a soft-in-soft-out (SISO) channel decoder, the BER can be estimated leg with the aid of the Logaπthmic Likelihood Ratio (LLR), evaluated either at the input or the output of the
TO channel decoder A particularly attractive way of invoking LLRs is employing powerful turbo codecs, i9i which provide a reliable indication of the confidence associated with a particular bit decision in the
192 context of LLRs The LLR is defined as the ratio of the probabilities of a specific bit being binary zero
193 or one Again, this measure can be evaluated at both the input and the output of the turbo channel codecs ,94 and both of them can be used for channel quality estimation.
,95 Thirdly, in the event that no channel encoder / decoder (codec) is used in the system, the channel quality
,96 expressed in terms of the BER can be estimated with the aid of the mean-squared error (MSE) at the
197 output of the channel equaliser or the closely related metric, the Pseudo-Signal-to-noise-ratio (Pseudo-
1 8 SNR) The MSE or pseudo-SNR at the output of the channel equaliser have the important advantage
199 that they are capable of quantifying the seventy of the inter-symbol-interference (ISI) and/or Co-channel
200 Interference expenenced, in other words quantifying the Signal to Interference plus Noise Ratio (SINR)
2oι 4.3.1 Pseudo-SNR Embodiment
202 A specific embodiment based on the above-mentioned pseudo-SNR is now descπbed in more depth
203 Employing the pseudo-SNR has the advantage that it is generally applicable regardless of whether or
204 not a channel codec is present
We found that the residual channel-induced lnter-symbol-interference (ISI) at the output of the decision feedback equaliser (DFE) is near-Gaussian distπbuted and that if there are no decision feedback enors,
the pseudo-SNR at the output of the DFE, γ(// can be calculated as [8]
Wanted Signal Power
Residual ISI Power + Effective Noise Power E !^Σ,
=0 Cmhm =-(N
f- \)
E l/ -
■ N
v» y- '„
V,--
~o 0' I C"-'" I!"
( 1 )
where C„, and lι„, denotes the DFE's feed-forward coefficients and the channel impulse response, respectively The transmitted signal and the noise spectral density is represented by S/, and N„ Lastly, the number of DFE feed-forward coefficients is denoted by N By utilizing the pseudo-SNR at the output of the equalizer, we are ensunng that the system performance is optimised by employing equalization and adaptive quadrature amplitude modulation (AQAM) in a wideband environment according to the following switching regime
Modulation Mode = (2)
205 where fn,n = 1 . 3 are the pseudo-SNR thresholds levels, which are set according to the system's integπty
206 requirements
207 In contrast to the narrowband, statically reconfigured multimode systems of [l]-[4] constituting the state - 2oβ of-the-art, the present embodiment invokes wideband, near-instantaneously reconfigured or burst-by- 209 burst adaptive channel-equalised modulation, in order to achieve the best possible multimedia source- 2io signal representation quality - for example video quality - when transmitting over arbitranly time- vaπant 2iι channels, exhibiting short-term and/or long-term channel quality vaπations These vaπations can be
212 due to the effects of path-loss, fast-fading, slow-fading, dispersion, co-channel interference, etc Further-
213 more, when the mobile is roaming in a hostile out-doors - or even hilly terrain - propagation environment,
2.4 typically low-order, low-rate modem modes are invoked, while in benign indoor environments predomi-
2.5 nantly the high-rate, high video quality modes are employed
2.6 It is an important element of the system that when the binary BCH channel codes or FEC codes protect-
217 ing the video stream are overwhelmed by the plethora of transmission errors, the embodiment refrains
218 from decoding the video packet in order to prevent error propagation through the reconstructed frame 2i9 buffer [5] Instead, these corrupted packets are dropped and the reconstructed frame buffer will not be 220 updated, until the next packet replenishing the specific video frame area arnves The associated video
Table 1 Modulation and channel parameters
221 performance degradation is fairly minor for packet dropping or frame error rates (FER) below about 5%
222 These packet dropping events are signalled to the remote decoder by supenmposing a strongly protected
223 one-bit packet acknowledgement flag on the reverse-direction packet, as outlined in [5] In the embod-
224 iment we also invoked the adaptive rate control and packetisation algoπthm of [5], supporting constant
225 Baud-rate operation
226 As a specific example of the burst-by-burst adaptive nultimedia system we used 176x144 pixel so-called
227 QCIF-resolution, 30 frames/s video sequences encoded at bitrates resulting in high perceptual video
228 quality, in order to demonstrate the performance advantages of the video transceiver Table 1 shows the
229 modulation- and channel parameters employed, noting again that the associated pnnciples are applicable
230 in the context of a whole range of other system parameters The COST207 four-path typical urban (TU)
231 channel model was used in quantifying the associated system performance and its impulse response
232 is portrayed in Figure 3 As an example, we used the Pan-European FRAMES proposal as the basis
233 for our wideband transmission system, the frame structure of which is shown in Figure 4 Employing
234 the FRAMES Mode Al (FMA1) so-called non-spread data burst mode required a system bandwidth of
235 3 9MHz, when assuming a modulation excess bandwidth of 50% A range of other system parameters
236 are shown in Table 2
237 The specific example of the video transceiver - which is used to demonstrate the advantages of the system
238 concept - is based on the H 263 video codec The video coded bitstream was protected by binary Bose-
Table 2 Geneπc system features of the reconfigurable multi-mode video transceiver, using the non- spread data burst mode of the FRAMES proposal shown in Figure 4
Chaudhuπ-Hocquenghem (BCH) coding combined with an intelligent burst-by-burst adaptive wideband multi-mode Quadrature Amplitude Modulation (QAM) modem, which can be configured either under network control or under transceiver control to operate as a 1, 2, 4 and 6 bits/symbol scheme, while maintaining a constant signalling rate This allowed us to support an increased throughput expressed in terms of the average number of bits per symbol, when the instantaneous channel quahtv was high leading ultimately to an increased video quality in a constant bandwidth
The transmitted bitrate for all four modes of operation is shown in Table 3 The unprotected bitrate before approximately half-rate BCH coding is also shown in Table 3 The actual useful bitrate available for video is slightly less, than the unprotected bitrate due to the required strongly protected packet acknowledgement information and packetisation information The effective video bitrate is also shown in Table 3
250 4.4 Burst-by-Burst Adaptive Videophone Performance
25i The descπbed burst-by-burst adaptive modem maximizes the system capacity available bv using the
252 most appropriate modulation mode for the current instantaneous channel conditions We found that the
253 pseudo-SNR at the output of the channel equaliser was an adequate channel quality measure in our burst-
Table 3: Operational-mode specific transceiver parameters
254 by-burst adaptive wide-band modem. Figure 5 demonstrates how the burst-by-burst adaptive modem
255 changes its modulation modes every transmission burst, ie every 4.615 ms, based on the fluctuating
256 pseudo-SNR. The right-hand-side vertical axis indicates the associated number of bits per symbol.
257 By changing to more robust modulation schemes automatically, when the channel quality is reduced
258 allows the packet loss ratio, or synonymously, the FER, to be reduced, which results in increased per-
259 ceived video quality. In order to judge the benefits of burst-by-burst adaptive modulation we considered
260 two scenarios. In the first scheme the adaptive modem always chose the perfectly estimated AQAM
26, modulation mode, in order to provide a maximum upper bound performance. In the second scenario
262 the modulation mode was based upon the perfectly estimated AQAM modulation mode for the previous
263 burst, which corresponded to a delay of one Time Division Multiple Access (TDMA) frame duration
264 of 4.615ms. This second scenario represents a practical burst-by-burst adaptive modem, where the one-
265 frame channel quality estimation latency is due to superimposing the receiver's perceived channel quality 26β on a reverse-direction packet, for informing the transmitter concerning the best mode to be used.
267 The probability of the adaptive modem using each modulation mode for a particular average channel
26β SNR is portrayed in Figure 6 in terms of the associated modem mode probability density functions
269 (PDFs). It can be seen at high channel SNRs that the modem mainly uses the 64QAM modulation mode,
270 while at low channel SNRs the BPSK mode is the most prevalent one.
27, Figure 7 shows the transmission FER (or packet loss ratio) versus channel SNR for the 1 , 2, 4 and 6 272 bit/symbol fixed modulation schemes, as well as for the ideal and for the one-frame delayed realistic
273 scenaπos using the burst-by-burst adaptive QAM (AQAM) modem In the ideal - le zero-delav - Λ.QAM
274 scenario, where the modulation mode estimation is assumed to be available instantaneously the trans-
275 mission FER is zero at high channel SNRs even though 64QAM is used predominantly, while at low
276 SNRs it exhibits a similar FER behaviour to fixed BPSK modulation, since this is the most often used
2 7 mode More explicitly, at high SNRs the adaptive modem chooses the most suitable AQAM mode and
278 hence no packets are lost However, at low SNRs the adaptive modem opts for using BPSK, even when
279 the channel is so hostile that the packets are lost Hence the BPSK and ideal - le zero-delay - AQAM
280 results are very similar at low channel SNRs However, when the modulation mode estimation is delayed
281 by one TDMA frame - representing a realistic, practical AQAM modem - then the transmission FER is
282 no longer zero at high channel SNRs, since the delay results in a non-optimum modulation mode selec-
283 tion, which can result in the corresponding video packet being lost At high channel SNRs the FER of
284 the realistic, one-frame delay AQAM modem is similar to that of the fixed 64QAM modem mode By
285 contrast, at low channel SNRs its FER performance is similar to that of the fixed BPSK modem mode
286 However, at medium channel SNRs the transmission FER is almost constant at about 3% for the realistic
287 AQAM modem This is more clearly demonstrated on a logaπthmic scale in Figure 8
288 Explicitly, the ideal and realistic AQAM modems are compared to a fixed modulation based, statically
289 re-configured multi-mode system with switching at 5% transmission FER in Figure 8 The statically
290 reconfigured modem was invoked here as a benchmarker, in order to indicate, how a system would 29, perform, which cannot act on the basis of the near-instantaneous ly varying channel quality As it can
292 be infirred from Figure 8, such a statically reconfigured transceiver switches its mode of operation from
293 a lower-order modem mode, such as for example BPSK to a higher-order mode, such as 4QAM, when
294 the channel quality has improved sufficiently for the 4QAM mode's FER to become lower than 5 %
295 upon reconfigunng the transceiver in this 4QAM mode Again - as seen in Figure 7 earlier on a non- 296 loganthmic scale - Figure 8 clearly shows that the ideal AQAM modem has a similar FER performance
297 to the fixed rate BPSK modem Additionally, it indicates that the realistic AQAM modem has a similar
298 FER performance to the BPSK modem at low SNRs, yielding a near-constant 3% FER at medium SNRs
299 and a FER similar to that of the fixed 64QAM modem at high channel SNRs
300 A compaπson of the effective video bitrate versus channel SNR is shown in Figure 9 for the four fixed so, modulation schemes, and the ideal and realistic AQAM modems The effective video bitrate is the
302 average bitrate provided by all the successfully transmitted video packets It should be noted that the
303 realistic AQAM modem has a slightly lower throughput, since sometimes the incorrect modulation mode
304 is chosen, which may result in packet loss At very low channel SNRs the throughput bitrate converges
305 to that of the fixed BPSK mode, since the AQAM modem is almost always in the BPSK mode at these
306 SNRs, as demonstrated in Figure 6.
307 Having shown the effect of the burst-by-burst adaptive modem on the transmission FER and effective
308 bitrate, let us now demonstrate these effects on the decoded video quality, measured in terms of the Peak
309 Signal-to-Noise Ratio (PSNR). Figure 10 shows the decoded video quality in terms of PSNR versus 3io channel SNR for both the ideal and realistic adaptive modem, and for the four modes of the statically 311 configured multi-mode modem. It can be seen that - as expected - the ideal adaptive modem, which 3i2 always selects the perfect modulation modes, has a better or similar video quality for the whole range 3i3 of channel SNRs. For the statically configured multi-mode scheme the video quality degrades, when 3i the system switches from a higher-order to a lower-order modulation mode. The ideal adaptive modem 3i5 however smoothes out the sudden loss of video quality, as the channel degrades. The non-ideal adaptive
316 modem has a slightly lower video quality performance, than the ideal adaptive modem, especially at
317 medium SNRs, since it sometimes selects the incorrect modulation mode due to the estimation delay.
318 This can inflict video packet loss and/or a reduction of the effective video bitrate, which in turn reduces
319 the video quality.
320 The difference between the ideal burst-by-adaptive modem, using ideal channel estimation and the non- 321 ideal, realistic burst-by-burst adaptive modem, employing a non-ideal delayed channel estimation can be
322 seen more clearly in Figure 11 for a range of video sequences. Observe that at high and low channel
323 SNRs the video quality performance is similar for the ideal and non-ideal adaptive modems. This is
324 because the channel estimation delay has little effect, since at low or high channel SNRs the adaptive
325 modem is in either BPSK or 64QAM mode for the majority of the time. More explicitly, the channel
326 quality of a transmission frame is almost always the same as that of the next, and hence the delay has
327 little effect at low and high SNRs.
32β The video quality versus channel quality trade-offs can be more explicitly observed in Figure 12. This
329 figure portrays the decoded video quality in PSNR versus the packet loss ratio or transmission FER.
330 The ideal and practical adaptive modem performance is plotted against that of the four fixed modulation
331 schemes in the figure. It can been seen that the adaptive modems' video quality degrades from that
332 achieved by the error-free 64QAM modem towards the BPSK modem performance as the packet loss
333 ratio increases. The practical adaptive modems' near constant FER performance of 3% at medium SNRs
334 can be clearly seen in the figure, which is associated with the reduced PSNRs of the various modem
335 modes, while having only minor channel enor-induced impairments.
Table 4: SINR estimate at output of the equaliser required for each modulation mode in Burst-by-Burst Adaptive modem, ie. switching thresholds
336 4.5 Switching Thresholds
337 The burst-by-burst adaptive modem changes its modulation modes based on the fluctuating channel con-
338 ditions expressed in terms of the SNR at the equaliser's output. The set of switching thresholds used in
339 all the previous graphs is the "Standard" set shown in Table 4, which was determined on the basis of the
340 required channel SINR for maintaining the specific target video FER.
34i In order to investigate the effect of different sets of switching thresholds, we defined two new sets
342 of thresholds, a more conservative set, and a more aggressive set, employing less robust, but more
343 bandwidth-efficient modem modes at lower SNRs. The more conservative switching thresholds reduced
344 the transmission FER at the expense of a lower effective video bitrate. The more aggressive set of thresh-
345 olds increased the effective video bitrate at the expense of a higher transmission FER.
346 The transmission FER performance of the realistic burst-by-burst adaptive modem, which has a one
347 TDMA frame delay between channel quality estimation and mode switching is shown in Figure 13 for
348 the three sets of switching thresholds of Table 4. It can be seen that the more conservative switching
349 thresholds reduce the transmission FER from about 3% to about 1% for medium channel SNRs. The
350 more aggressive switching thresholds increase the transmission FER from about 3% to 4-5%. However,
351 since FERs below 5% are not objectionable in video quality terms, this FER increase is an acceptable
352 compromise for a higher effective video bitrate. The effective video bitrate for the realistic adaptive
353 modem with the three sets of switching thresholds is shown in Figure 14. The more conservative set
354 of switching thresholds reduces the effective video bitrate but also reduces the transmission FER. The
355 aggressive switching thresholds, increase the effective video bitrate, but also increase the transmission
356 FER. Therefore the optimal switching thresholds should be set such that the transmission FER is deemed
357 acceptable is the range of channel SNRs considered.
358 5 Summary
359 The above-descπbed burst-by-burst adaptive multimedia transceiver concept exhibits substantial advan-
360 tages in compaπson to conventional fixed-mode or statically reconfigurable transceivers, which was sub- 36i stantiated in the context of a specific embodiment of the advocated system concept, namely with the aid
362 of a burst-by-burst adaptive video transceiver
363 Specifically, the main advantage of the descπbed burst-by-burst adaptive transceiver technique is that lr-
364 respective of the prevailing channel conditions, the transceiver achieves always the best possible source-
365 signal representation quality - such as video, speech or audio quality - by automatically adjusting the
366 achievable bitrate and the associated multimedia source-signal representation quality m order to match
367 the channel quality expenenced This is achieved on a near-instantaneous or burst-by-burst adaptive 36β basis under given propagation conditions in order to cater for the effects of path-loss, fast-fading, slow-
369 fading, dispersion, co-channel interference, etc Furthermore, when the mobile is roaming in a hostile
370 out-doors - or even hilly tenain - propagation environment, typically low-order, low-rate modem modes 37i are invoked, while in benign indoor environments predominantly the high-rate, high source-signal repre-
372 sentation quality modes are employed
3 3 The descnbed system embodiment has the following features
374 1 A reliable instantaneous channel quality metπc is employed, in order to appropπately configure
375 the AQAM modem for maintaining the required target BER and the associated source signal rep- 37β resentation quality The range of potential channel quality metπcs entails the pseudo-SNR, SINR,
377 BER and its LLR-based channel estimates
378 2 The perceived channel quality determines the number of bits that can be transmitted in a given
379 transmitted packet or burst, which in turn predetermines the number of bits to be generated by the see associated multimedia source codec, such as for example the associated video, audio, speech or 3β, handwπting codec Hence the multimedia source codec has to be capable of adjusting the number 3β2 of bits generated under the instruction of the burst-by-burst adaptive transceiver
383 3 The transmitter mode requested by the receiver, in order to achieve the target performance has to
384 be signalled by the receiver to the remote transmitter Another scenaπo is, where the uplink and
385 downlink channel quality is sufficiently similar for allowing the receiver to judge, what transmis- 3β6 sion mode the associated transmitter should use, in order for its transmitted signal to maintain the 387 required transmission integrity Lastly, the mode of operation used by the transmitter can also be see detected using blind detection techniques, for example in conjunction with the associated channel
389 decoder.
390 In the studied example of the system embodiment we have characterised a wideband burst-by-burst adap-
391 tive multimedia transceiver, which employed the pseudo-SNR at the output of the channel equaliser as
392 the quality measure for controlling the AQAM modem modes. Whilst in reference [16] the through-
393 put upper-bound of such an AQAM modem was analysed, in this document a practical multimedia
394 transceiver concept was described and the achievable performance gains due to employing the described
395 wideband burts-by-burst adaptive modem were quantified. An adaptive packetiser was used in conjunc-
396 tion with the adaptive modem, which continually adjusted the video codec's target bitrate, in order to
397 exploit the instantaneous bitrate provided by the adaptive modem.
398 In the example the delay between the channel estimation and modulation mode switching was shown
399 to have a considerable effect on the performance achieved by the adaptive modem. This performance
400 penalty can be mitigated by reducing the modem mode switching latency, for example by employing 40, adjacent slots for the uplink and downlink of of a TDD system. However, at lower vehicular speeds 02 the effects of AQAM mode switching latency are less crucial and the practical adaptive modem can
403 achieve a performance that is close to that of the ideal adaptive modem exhibiting no switching latency,
404 that we used as an upper-bound benchmarker. We have also demonstrated, how the transmission FER
405 performance is affected by changing the switching thresholds. Therefore the system can be tuned to the
406 required FER performance using appropriate switching thresholds.
407 It will be appreciated that although a particular embodiment of the invention has been described, many 4oβ modifications / additions and / or substitutions may be made within the spirit and scope of the present 409 invention.
4,0 References
4iι [1] J. Streit, L. Hanzo: Vector-quantised low-rate cordless videophone systems, IEEE Tr. on Vehicular
4i2 Technology, Vol. 42, No.2. May 1997, pp 340-357
4i3 [2] L. Hanzo, J. Streit: Adaptive low-rate wireless videophone systems, IEEE Tr. on Circuits and
4. Systems for Video Technology, Aug. 1995, Vol. 5, No. 4, 305-319, ISSN 1051-8215
4.5 [3] J. Streit, L. Hanzo: Quadtree-based reconfigurable cordless videophone systems, IEEE Tr. on Cir-
4.6 cuits and Systems for Video Technology, Apr. 1996, Vol. 6, No. 2, pp 225-237
4.7 [4] J. Streit, L. Hanzo: Comparative study of programmable-rate videophone codecs for existing and
4.8 future multimode wireless systems, European Transcations on Telecomms., Vol.8, Issue No 6, 1997,
419 pp 551-572
420 [5] P. Cherriman, L. Hanzo: Programable H.263-based wireless video transceivers for interference-
421 limited environments, IEEE Tr. on CSVT, June 1997, Vol. 8, No.3, pp 275-286
422 [6] N. Faerber, E. Steinbach, B. Girod: "Robust H.263 Compatible Transmission for Mobile Video
423 Server Access", Proc. First International Workshop on Wireless Image/Video Communications,
424 Loughborough University, UK, 4-5 September 1996, pp 8-13
425 [7] A.H. Sadka, F. Eryurtlu, A.M. Kondoz: Improved performance H.263 under erroneous transmission 42β conditions, Electr. Letters, 16th of Jan., 1997, Vol. 33, No. 2, pp 122-124
427 [8] J.C.S. Cheung, R. Steele: Soft-decision feedback equalizer for continuous-phase modulated sig-
42β nals in wide-band mobile radio channels, IEEE Tr. on Communications, 1994, Vol.42, No.2-4 Pt3,
429 pp.1628-1638
430 [9] W.T. Webb and R. Steele, "Variable Rate QAM for mobile radio," IEEE Transactions on Commu-
431 nications, vol. 43, pp. 2223 - 2230, July 1995.
432 [10] J.M. Torrance and L. Hanzo, "On the Upper bound performance of adaptive QAM in a slow
433 Rayleigh fading," IEE Electronics Letters, pp. 169 - 171, April 1996.
434 [1 1] J.M. Torrance and L. Hanzo, "Optimization of switching levels for adaptive modulation in a slow
435 Rayleigh fading channel," IEE Electronic Letters, pp. 1 167 - 1 169, June 1996.
436 [ 12] H Matsuako, S Sampei, N Monnaga and Y Kamio, "Adaptive Modulation systems with vaπable
437 coding rate concatenated code for high quality multi-media communication systems," in Proceed- 43B ings of IEEE Vehicular Technology Conference, pp 487 - 491 , 1996
439 [13] A J Goldsmith and S G Chua, "Vanable Rate Vanable Power MQAM for Fading Channels," -MO IEEE Transactions on Communications, vol 45, pp 1218 - 1230, October 1997
[ 14] A Goldsmith, "The capacity of downlink fading channels with vaπable rate and power." IEEE Ti
442 on Veh Techn., vol 46, pp 569-580, Aug 1997
443 [15] A Goldsmith and S Chua, "Adaptive coded modulation for fading channels," IEEE Tt on Commu-
444 nications, vol 46, pp. 595-602, May 1998
445 [ 16] CH Wong, L. Hanzo- Channel Capacity Upper-bound of a Wideband Burst-by-burst Adaptive
446 Modem, Proc. of VTC'99, Houston, USA, May, 1999, pp 1851 -1855
447 [17] S Otsuki, S. Sampei and N. Monnaga, "Square-QAM adaptive modulation/TDMA/TDD systems 4β using modulation estimation level with Walsh function," IEE Electronics Letters, vol 31. pp 169 — 49 171, February 1995.
450 [18] J.M. Torrance and L. Hanzo, "Demodulation level selection in adaptive modulation," IEE Electron- 45, ics Letters, pp. 1751 - 1752, September 1996.