With the advent of data applications, interference
is a challenge for wireless carriers. Consequently, mitigating interference to maximize spectral efficiency and improve network throughput is on the minds of operators and handset makers. This
article describes an interference cancellation technology comprising
an ASIC/core hardware and DSP-based software, which
cancels interference from all traffic channels, and from all interfering sources for 2.5G, 3G and 4G networks.
Canceling intracell interference induced by channel effects and intercell interference introduced by channel effects, as well as serving and non-serving base stations is a technological challenge. And, it is particularly difficult when the code space is heavily loaded, as in dense voice networks such as fully loaded CDMA2000 1xEV-DO data networks or mixed-voice and HSDPA networks.
While there may be several receiver architectures in the market to address the interference challenge, in reality there are four general classes of baseband receivers. And, in terms of increasing complexity, these include RAKEs, equalizers, linear minimum mean square error (MMSE) receivers, and multi-user detectors (MUDs). Each of these receivers is an optimum solution for a different combination of communication protocol, channel condition, and code-space loading. Consequently, when a RAKE is optimum, there is no performance gap to be filled by a more advanced receiver. However, when an equalizer is optimum, the only performance gap to be filled is between a RAKE and an equalizer. Likewise, when a linear MMSE receiver is optimum, there are performance gaps to be filled between a RAKE, an equalizer, and a linear MMSE receiver.
Thus, to fill available performance gaps between a RAKE, an equalizer, a linear MMSE receiver, at low complexity, TensorComm has developed a novel interference cancellation technology (ICT), comprising an ASIC/core hardware and DSP-based software, that cancels interference from all traffic channels, and from all interfering sources for 2.5G, 3G and 4G networks. A protean signal-processing operator, ICT exploits all time-varying source characteristics that code-space profiles and channel conditions leave open for exploitation, in a hardware-efficient architecture.
Interference sets the limit on performance in code-based systems.By canceling unwanted interference, ICT reduces power requirements and increases spectrum efficiency so that cells can maintain size and network capacity. The technology requires no modification to existing or evolving air interface standards. Because it is a receive-only technology, it requires no base station modification to effectively reduce interference on the forward link. It also can be integrated into CDMA and WCDMA chipsets today. The same technology may be applied to the reverse link and integrated into base stations. The important effect is to maintain cell size at high traffic density and to reduce the frequency of dropped calls in hand-off.
Besides canceling intersymbol and interchannel interference from pilot, paging, synchronization, traffic and high-rate data channels, ICT is specifically designed to cancel direct and multipath interference from adjacent base station sectors and interference from multipath within the serving sector. Also, this technology is complementary to, and can co-exist with, other technologies, such as transmit/receive diversity technologies, to re-capture power and/or spectral efficiency. Additionally, it is independent of carrier frequency bands, and will apply at 450 MHz, 800 MHz, 1900 MHz and elsewhere.
Theoretical foundations of ICT
Interference is the limiting factor in the performance of CDMA and WCDMA wireless networks. Field conditions such as fading and multipath defeat all attempts to maintain orthogonality between traffic and control channels in multi-access voice and data systems based on CDMA and WCDMA standards. The lack of orthogonality leads to interference and a consequent reduction in signal-to-interference and noise ratio (SINR). Thus, CDMA and WCDMA are interference-limited rather than noise-limited. As a consequence:
After multipath resolution with a RAKE receiver, every resolved baseband path contains interchannel interference (ICI) and intersymbol interference (ISI) from every other path.
This produces a bit-error-rate (BER) higher than a target energy-to-noise density ratio (Eb/No) would predict, requiring a) increased signal strength and SINR, b) reduced traffic-loading and/or c) reduced bit-rate to maintain network quality of service (QoS).
Transmit power increases because neighboring devices ask for more power to contend with more interference.
Overall network capacity should be maximized by having each device use the minimum required transmission power so that the interference caused to other devices in the network is minimized.
There is a body of work for interference cancellation and multi-user detection suggesting that all interference effects can be managed with signal processing if the channel can be accurately estimated and the optimal signal-processing solution can be implemented at the symbol rate for voice and data. Neither of these ideals is achievable. The TensorComm approach approximates the optimum solution by factoring interference cancellation into a sequence of signal-processing steps that remove ISI and ICI.
The company has filed more than 75 patents in this area and indicates its strength is to blend advanced signal processing with existing transceiver architectures for CDMA and WCDMA modems.
The example illustrated in Figure 1 describes the application of ICT. Consider y to be the complex baseband signal arriving at a handset (after reception at the antenna and downconversion). This signal can be resolved into multiple components that represent the different paths arriving at the antenna, plus thermal noise.
For example, we could represent the complex signal in path 1 as:
where s1 and s2 are signals from two different paths from base station 1, while s3 and s4 are multipath signals from base station 2 in soft hand-off. n is the thermal noise in the received signal.
A conventional RAKE receiver assigns these paths to different fingers, which then recover the message by applying the correct aligned codes for recovery of the transmitted symbol. While the design and selection of the codes attempts to minimize the cross-correlation of the desired codes with the codes of other paths, the presence of multipath and hand-off defeats orthogonality and produces non-zero cross-correlation between signal components.
Let the codes for the channels of interest be:
x1, x2, x3 and x4 for the respective paths.
The RAKE receiver then recovers the symbols
m1, m2, m3 and m4 by computing inner products with the corresponding codes. The estimated symbol in path 1, is obtained using the inner product or correlation
where m1 is the symbol of interest in path 1, i1 is interference and n1 is noise.
Similarly, each path experiences interference from all the other paths:
The combiner combines the symbol estimates from each path to arrive at a soft decision, usually using a maximal ratio combiner:
where β1, β2, β3 and β4 are the maximal ratio coefficients associated with each path (usually pilot amplitudes).
Because of the correlations between the multipath signals si and the codesxj, the symbol estimates contain interference from all paths.
The estimated SINR, also referred to as Ec/Io, is the ratio of the signal energy (Ec) to the total noise and interference (Io) in the system. Let's call Ij the variance (or power) of interference ij, and σ2 the variance of the noise. Then ignoring correlation between interferences, a crude but descriptive estimate of SINR is:
In an ICT-enabled handset, the interference in the signal y1 is canceled, so that the application of the desired code yields a smaller interference term. Thus,
where y1ICT is the signal for path 1 with interference canceled.
where the variance (or power) of ε1i1 is much smaller than the power of i1, and α1 is approximately 1. Therefore, the maximal ratio combined symbol estimate is:
The new estimated SINR is:
The ICT gain G can be roughly estimated as:
To illustrate, if all εj2 are roughly equal to ε2 <<1, then the gain is:
This gain is realized without impacting the diversity that multipath brings or from the gain due to hand-off. Interference cancellation preserves all the advantages of the system while increasing SINR. In hand-off, the receiver resolves multipaths with higher SINRs than it would have in the absence of ICT.
Architecture and integration
As shown in Figure 2, ICT is an ASIC/core solution that integrates into the modem of the baseband chipset. When deployed in the handset to mitigate forward link interference, no other components and no re-design of the handset RF are required, resulting in no form factor changes to the handset. Furthermore, no base station modifications are required.
The sequence of signal-processing steps in ICT occurs after RF processing, delivering interference-cancelled signals to the RAKE receiver. ICT delivers a subspace version of the original path signals to each finger, relatively free of ISI and ICI, with higher SINR. This is illustrated in Figure 3.
ICT has been integrated into platforms and successfully tested in the field on commercial networks. In addition to developing the basic ICT algorithms, TensorComm has evolved a process for integrating its algorithms and intellectual property into a customer's CDMA or WCMDA modem.
Impact on performance of network
In a CDMA network, as traffic load increases, the total base station transmit power increases, because a handset requires more transmit power from the base station to maintain the same performance in dense interference. The effect on the network is that probability of coverage decreases and network performance degrades. Furthermore, when the network load exceeds 75%, degradations are more pronounced as cell boundaries collapse, creating coverage holes. The result is that customers experience a greater number of dropped and blocked calls.
The sequence of panels in Figure 4 demonstrates in a qualitative way how ICT reclaims power and spectral efficiency to maintain cell coverage and traffic density. That is, interference cancellation increases the capacity of the network for more users or for higher data rates, while expanding or maintaining network coverage. In essence, it shows that base stations no longer have to transmit so much power to each handset. And can use extra power to increase capacity, coverage, data rates and quality.
A network attempts to maintain uniform quality for users by setting performance targets and adjusting transmit power to meet these targets.
At a low level, ICT operates in tandem with fast and slow power control to provide performance improvement. ICT improves SINR on each signal, which translates to a reduction in demodulation errors and lower frame error rate (FER). Accordingly, the handset compensates for this performance improvement by requesting a lower forward-link transmit power in an attempt to keep base station transmit power at a minimum.
Additionally, the increase in SINR due to interference cancellation allows a greater number of base station sectors to remain in the active and candidate sets. This provides greater signal diversity, which is invaluable in compensating for signal fading. The technology can provide large instantaneous gains in fades that would limit handset performance. The interference cancellation gains also “soften” the impact of fades, since power control movements are minimized during fades.
From the base station perspective, each ICT-enabled handset requires reduced transmit power to maintain the same performance. As a result, the base station is able to increase the number of served users in a cell. From the network, a second-order effect is observed: for a given number of users, each base station lowers the transmit power for ICT-enabled handsets, reducing the noise on all mutually interfering sectors. This leads to further reduction in network transmit power.
In field environments, there is almost always an opportunity for interference cancellation. As a result, ICT will be operational for much of the time. However, the design incorporates a level of intelligence, so that it may turn off the cancellers when cancellation is not beneficial. This “no-harm” feature guarantees that an enabled handset always provides performance better than or equal to that of a non- ICT-enabled handset.
The company has proven its technology through comprehensive technology development, testing, and evaluation, consisting of simulations, laboratory testing and field validation. Figures 5-7 demonstrate the performance of ICT by documenting its impact on base station transmit power and file transfer speed.
Laboratory tests were conducted to evaluate the prototype under various signal conditions, such as fading, multipath and hand-off, over a range of network loads from 0% to 75%. In the laboratory, test scenarios were specified with a pilot Ec/Io for each base station that depended on proximity to the handset, from the edge of the cell, where pilots were equal, to close-in, where the difference in pilot strengths was 6 dB.
The forward-link traffic channel power gain ranged from 0 db to 4 dB. Figure 6 illustrates the forward traffic channel power reductions as a function of the cell loading, at three different pilot strengths. When pilot strength separation is medium to low, ICT provides significant gains that increase with % OCNS (orthogonal channel noise simulation).
Drive tests were conducted on a major U.S. operator's CDMA2000 commercial network over an extended period at various times during the day and night. Tests were conducted simultaneously with two side-by-side prototypes comparing the network impact between a prototype with ICT enabled and another prototype with ICT disabled. In all cases, the critical metric recorded was the base station forward-link traffic channel power.
The commercial test results of Figure 7 demonstrate a substantial reduction in forward link traffic channel power. Gains averaged 2.5 dB in a heavily loaded environment with peak gains of up to 6 dB.
Laboratory tests of data throughput revealed a doubling of data rates for those handset prototypes enabled with ICT. As illustrated in Figure 8, a file transferred to an ICT-enabled prototype took half the time of a file transfer to a prototype with ICT disabled. This was further validated by the recorded data rate, which was doubled for the ICT prototype at a reduced data retransmission rate.
Simulation evaluations of WCDMA versions (release 99 and HSDPA) have been completed and observed gains are comparable to CDMA2000 gains.
Network modeling of these test results shows that the recaptured base station power can support greater than 40% more subscribers and make all future network capacity expenditures 40% more efficient. This gain can be realized and used in multiple ways: to increase higher data rates for data applications, support high traffic densities, provide better quality of service for voice applications, and reduce base station transmit power, leading to increased network capacity. This capacity relief allows the wireless operator to delay and reduce significant capital and network operational expenditures on expensive infrastructure and spectrum. Network modeling indicates that for a 20 M subscriber system, the realized savings would be more than one billion dollars in a five-year period.
By adopting ICT, a wireless service operator can increase network capacity, accompanied by a substantial savings of capital expenditure. The ICT's patented approach delivers gains in power and spectral efficiency, thereby improving cell capacity, increased coverage, improved quality of service and increased data rates. Based on commercial network trials, this technology has proven to increase a CDMA wireless service operator's network capacity by greater than 40%, thus delaying and reducing the capital and spectrum expenditures required to support subscriber growth, increased minutes-of-use, expanded coverage, and increased data services.
About the author
John Thomas is CEO and co-founder of TensorComm. Prior to founding TensorComm, Thomas co-founded Data Fusion Corp. Before co-founding Data Fusion, he was at NASA's Jet Propulsion Laboratory. Thomas earned a Ph.D in Electrical Engineering/Signal Processing from the Univeristy of Colorado at Boulder.