Chapter 2 Multi-Band/Multi-Mode Radio Systems 5
3.2 Concurrent Receiver Architectures
In the following sections a few potential concurrent transceiver architectures are studied.
On the receiver side, desirable architectures are those with a smaller chip area and/or
5 It is already assumed that the environment in our hypothetical case does support both standards and channels at both frequency bands for our single user. Just recently, single wireless network access points that can simultaneously support multiple standards at 2.4GHz and 5.2GHz have been introduced to the market [40].
consume less battery power. It is worth mentioning that in the following description of various concurrent transceiver architectures, only signal-path designs are presented. As discussed in the previous chapter, a proper frequency planning can be applied to most of these architectures to reuse some of the resources (e.g., frequency synthesizers) in LO- generation sections.
3.2.1 Parallel Single-Band Receivers
Probably the most trivial solution for a concurrent multi-band receiver is to have a number of single-band receivers in parallel and use each for one particular frequency band. Figure 3.1 shows examples of the aforementioned implementation where each single-band receiver is using a standard heterodyne architecture in (a), and direct down-conversion architecture in (b). More generally, different architectures for each of the parallel receivers can be conceived.
Despite the fact that this scheme is general and can be expanded to any number of frequency bands, it is not the most efficient architecture. Later, we will show architectures that share the valuable resources (i.e., battery power, chip area) by introducing novel and unconventional building blocks.
Figure 3.1: Concurrent receiver implementations using multiple parallel single-band, receivers in (a) heterodyne (b) direct down-conversion architectures
3.2.2 Parallel Receivers with a Wide-Band Front-End
A receiver with a wide-band front-end consisting of an antenna and low-noise amplifier (LNA) that can collect signal power at a large range of frequencies followed by appropriate down-converters can be used to receive any number of frequency bands within that range.
In Figure 3.2, a broadband front-end is followed by separate down-conversion paths for each frequency band. This scheme might have some advantage in terms of chip area and/or power consumption over the previous method, depending on the implementations of wide- band LNA6. The direct down-conversion architecture of Figure 3.2 (b) is fairly versatile in that by controlling the local oscillator frequency, the operation bands of the concurrent receiver can be varied.
Broadband front-end nevertheless suffers from a serious shortcoming: it not only receives the frequency bands of interest, but also all the other undesirable frequencies get amplified by it. These unwanted signals might have a large signal power and can potentially
6 It is also noteworthy that normally antenna size increases with its bandwidth.
LO1 LO2
LNA BPF2 BPF3
BPF1
ω ω ω ωin1
ωω ωωin2
LO1,I
BPF1 LNA
ωω ωωin1
LO1,Q
ω ω ω ωin2
ω ω ω ωin3
(a) (b)
ωω ωωin3
limit the dynamic range of the receiver due to the nonlinearities inherent to any implementation.
Figure 3.2: Concurrent receiver implementation using a wide-band front-end in (a) heterodyne (b) direct down-conversion architectures
3.2.3 Parallel Receivers with a Multi-Band Front-End
The shortcoming of the previous architecture can be improved by replacing the wide-band front-end with one that has a multiple narrow-band response at frequency bands of interest (Figure 3.3). This multi-band front-end consists of a multi-band antenna (e.g., [28]-[30]), followed by a multi-band filter (e.g., [31]) and a concurrent multi-band LNA that provides simultaneous gain and matching with a low added noise at multiple frequency bands [26].
More discussions on the design of a concurrent multi-band front-end will be provided in section 3.3.
LO1 LO2
broadband LNA
BPF2 BPF1
broadband antenna
(a) (b)
broadband LNA broadband
antenna
LO1,I
LO1,I
Figure 3.3: Concurrent receiver implementation using a multi-band front-end in (a) heterodyne (b) direct down-conversion architectures
3.2.4 Multi-Band Sub-Sampling Receiver
According to Shannon’s original statement of sampling theorem, any signal that contains no frequencies higher than BW [Hz] can be completely determined by discrete samples of the signal spaced (1/2 BW) [sec] apart [21]. Sampling of band-limited signals has been extended to bandpass signals that have energy in the frequency interval (fL, fU) where BW=fU-fL [22]. The classical bandpass theorem for uniform sampling states that the signal can be reconstructed if the sampling rate is at least fs(min)
=2fU/n, where n is the largest integer within fU/BW. A graphical representation of bandpass sampling in the frequency domain can be seen in Figure 3.4. In the frequency domain, the effect of uniform sampling is equal to repeatedly shifting the signal by integer multiples of the sampling rate, fs. If these shifted versions of the signal can not be superimposed on one another, the original signal can be recovered easily with appropriate filtering. Using this graphical representation, the aforementioned condition for permissible sampling rates can be derived.
There are two important observations in the bandpass sampling scheme. First, note that the sampling rate can be much lower than the center frequency of the signal. Second, we not only can recover the signal in its original form by filtering at the correct center frequency,
LO1 LO2
multiband LNA
BPF2 BPF1
multiband antenna
(a) (b)
LO1,I
LO1,I
multiband LNA multiband
antenna
multiband filter
but also can filter any other replica of the signal and effectively shift the center frequency of the signal without disturbing its frequency contents. This point has instigated schemes that use the so-called sub-sampling to down-convert the RF signal to a lower frequency [23],[24]. Subsampling receivers traditionally have a good linearity performance, but have the disadvantage of noise-folding and hence a lower SNR at the output. As shown in Figure 3.4, both the signal and the wide-band background noise are shifted several times, the latter resulting in an increased total in-band background noise. Filtering the bandpass signal prior to sampling can lower the noise and increase the receiver sensitivity.
Figure 3.4: Frequency domain representation of down-conversion using bandpass subsampling
Bandpass subsampling can be further expanded to bandpass signals. Similar to the previous case, expressions for permissible sampling rates can be derived (e.g., [25]). Once again, Figure 3.5 shows the essence of this concept without any mathematical details and can be used to find the allowable rates for any set of discrete signals. In practice, the frequency bands of interest might not have the same bandwidth and most likely are not equally spaced
sampling rate
noise floor
f
f
BW
fu
fL
fs
increased noise floor Ts=1/fs
Ts
sampler low-pass filter
filter
in frequency domain. This may result in very low sampling rates which will consequently increase in-band noise due to noise-folding. Therefore, a multi-band filtering scheme is desired at the front-end to limit the effect of out-of-band noise. Additionally, using the multi-band filtering instead of a wide-band front-end is required to reduce the undesirable interferences that limit the dynamic range of the receiver as previously discussed.
Figure 3.5: Frequency domain representation of multi-band subsampling
3.2.5 Direct Digitization (Digital Radio)
Most modern wireless radio receivers down-convert the input RF signal to a lower frequency (intermediate frequency, IF, or baseband) and then convert the low-frequency analog signal to a digital one for further processing. With the rapid improvement in device technology providing faster, smaller, and cheaper active elements, digital circuits are now capable of operating at GHz range frequencies. Hence, there is a large interest in moving the analog-to-digital conversion to higher frequencies (Figure 3.6). Versatile digital circuitry can then process the high-frequency signal (e.g., filter, demodulate). Ideally, such a digital processor can be programmed to process signals at different frequency bands and standards (software defined radio), as well as multi-band signal structures. In theory, direct digitization of a radio-frequency signal is an intriguing idea, but there are extremely difficult challenges in practice with the current technologies. High resolution analog-to-
sampling rate
f
f fs
filter
digital converters (ADC) that can operate at multiple Gsamples/sec range with the tough dynamic-range requirements of radio standards are still not on the horizon for implementation. At the same time, the dynamic-range of these ADCs usually comes at the price of increased power consumption. As power consumption of digital circuitry increases in proportion to the frequency of operation, justifying the direct digitization for low-power portable radio receivers becomes more difficult. Upon availability of the desired ADC, the clear advantages of such implementations in the future are their performance scaling with technology improvements, ideally automated and faster design cycles, programmability, and extreme versatility in architectures and applications. In the following section we introduce an alternative approach that does not suffer from these deficiencies.
Figure 3.6: Generic architecture of a direct digitization radio