Receiver Architectures for Multiband/Multimode

The traditional receiver architecture of choice has been, and in many instances continues
to be, the superheterodyne, or “superhet.” The principle of the superhet, invented by
Edwin Armstrong early in the twentieth century, is to take the incoming received signal
and to convert it, together with its modulation, down to a lower frequency—the
intermediate frequency (IF), where channel selective filtering and most of the gain is
performed. This selected, gained up channel is then demodulated to recover the baseband
signal.
Because of the limited bandwidth and dynamic range performance of the superhet
stages prior to downconversion, it is necessary to limit the receiver front-end bandwidth.
Thus, the antenna performance is optimized across the band of choice, preselect
filters have a similar bandwidth of design, and the performance and matching efficiencies
of the low-noise amplifier (LNA) and mixer are similarly tailored.
For GSM GPRS, the handset preselect filters have a bandwidth of 25 MHz
(GSM800), 39 MHz (GSM900), 75 MHz (GSM1800), or 60 MHz (GSM1900). The filter is
used to limit the RF energy to only that in the bandwidth of interest in order to minimize
the risk of overloading subsequent active stages. This filter may be part of the
duplex filter. After amplification by the LNA, the signal is then mixed with the local
oscillator (LO) to produce a difference frequency—the IF. The IF will be determined by
the image frequency positioning, the selectivity capability and availability of the IF filter,
and the required LO frequency and range.
The objective of the superhet is to move the signal to a low-cost, small form factor
processing environment. The preselect filter and LNA have sufficient bandwidth to process all possible channels in the chosen band. This bandwidth is maintained to the
input to the mixer, and the output from the mixer will also be wideband. Thus, a filter
bandwidth of just one channel is needed to select, or pass, the required channel and
reject all adjacent and nearby channels after the mixer stage. This filter is placed in the
IF. In designing the superhet, the engineer has chosen the IF and either designed or
selected an IF filter from a manufacturer’s catalog.
The IF filter has traditionally had a bandwidth equal to the modulation bandwidth
(plus practical tolerance margin) of a single channel. Because the output from the
mixer is wideband to support multiple channels, it is necessary to position the wanted
signal to pass through the selective IF filter. For example, if the IF filter had a center frequency
of 150 MHz and the wanted channel was at 922 MHz, the LO would be set to
1072 MHz (1072-922 = 150 MHz) or 772 MHz (922-772 = 150 MHz) to translate the center
of the wanted channel to the center of the IF filter. The designer must ensure that the
passband of the filter can pass the modulation bandwidth without distortion. Following
the selective filtering, the signal passes to the demodulator where the carrier (IF) is
removed to leave the original baseband signal as sourced in the transmitter.
The IF filter and often the demodulator have traditionally been realized as electromechanical
components utilizing piezoelectric material—ceramic, quartz, and so on.
This approach has provided sufficient selectivity and quality of filtering for most
lower-level (constant envelope) modulations, such as FM, FSK, and GMSK. However,
with the move toward more complex modulation, such as π/4DQPSK, QPSK, and
QAM, the performance—particularly the phase accuracy of this filter technology—
produces distortion of the signal.
The second problem with this type of filter is that the parameters—center frequency,
bandwidth, response shape, group delay, and so on—are fixed. The engineer is designing
a receiver suitable for only one standard, for example, AMPS at 25 kHz bandwidth,
IS136 at 30 kHz, GSM at 200 kHz. Using this fixed IF to tune the receiver, the LO must
be stepped in channel increments to bring the desired channel into the IF.
Given the requirement for multimode phones modes with different modulation
bandwidths and types, this fixed single-mode approach cannot be used. The solution
is either to use multiple switched filters and demodulators or to adopt an alternative
flexible approach.
The multi-filter approach increases the cost and form factor for every additional
mode or standard added to the phone and does not overcome the problems of insufficient
phase/delay performance in this selective component. Amore cost-effective, flexible
approach must be adopted.
It is the adoption of increasingly capable digital processing technology at an acceptable
cost and power budget that is providing a flexible design solution. To utilize digital
processes, it is necessary to convert the signal from the analog domain to the digital
domain.
It would be ideal to convert the incoming RF to the digital domain and perform all
receive processes in programmable logic. The ultimate approach would be to convert the
whole of the cellular RF spectrum (400 MHz to 2500 MHz) in this way and to have all
standards/modes/bands available in a common hardware platform—the so-called software
radio. The capability to convert signals directly at RF—either narrowband or
wideband—to the digital domain does not yet exist. The most advanced analog-to-digital converters (ADCs) cannot yet come near to this target. To configure a practical cost
effective receiver, the ADC is positioned to sample and digitize the IF; that is, the conventional
downconverting receiver front end is retained.
The receiver design engineer must decide the IF frequency and the IF bandwidth to
be converted. In the superhet architecture, the higher the IF that can be used, the easier
the design of the receiver front end. However, the higher the frequency to be converted,
the higher the ADC power requirement.
If an IF bandwidth encompassing all channels in the band selected could be digitized,
the receiver front end could be a simple non-tuning downconverter with channel
selection being a digital baseband function. This is a viable technique for base
station receivers where power consumption is less of an issue; however, for handsets,
the ADC and DSP power required restricts the approach to digitization of a singlechannel
bandwidth.
This then returns us to single-channel passband filtering in the analog IF prior to
digitization—a less than ideal approach for minimum component multimode handsets.
However, minimum performance IF filters could be employed with phase compensation
characteristics programmed into the digital baseband filtering to achieve
overall suitability of performance.
Another possible approach is to use a single IF selective filter but with a bandwidth
suitable for the widest mode/standard to be used. For W-CDMA, this would be 5 MHz.
The 5 MHz bandwidth would then be digitized. If it was required to work in GSM mode,
the required 200 kHz bandwidth could be produced in a digital filter. This approach
needs careful evaluation. If the phone is working predominantly in GSM mode, the sampling/
digitizing process is always working at a 5 MHz bandwidth. This will consume
considerably more power than a sampling system dimensioned for 200 kHz.
So, in summary, the base station may use a wideband downconverter front end and
sampling system with baseband channel tuning, but the handset will use a tunable
front end with single-channel sampling and digital demodulation. The required number
of converter bits must also be considered.
Again, the power consumption will be a key-limiting parameter, given the issues of
input (IF) frequency and conversion bandwidth. The number of bits (resolution)
equates directly to the ADC conversion or quantization noise produced, and this must
be small compared with the carrier-to-noise ratio (CNR) of the signal to be converted.
In a GSM/GPRS receiver, 8 to 10 bits may be necessary. In a W-CDMA receiver, since
the IF CNR is considerably worse (because of the wideband noise created signal), 6 or
even 4 bits may be sufficient.
In a mobile environment, the received signal strength can vary by at least 100 dB,
and if this variability is to be digitized, an ADC of 20 bits plus would be required.
Again, at the required sample rates this is impractical—the dynamic range of the signal
applied to the ADC must be reduced. This reduction in dynamic range is achieved
by the use of a variable-gain amplifier (VGA) before the ADC. Part of the digital processing
function is to estimate the received signal strength and to use the result to
increase or decrease the gain prior to the ADC.
This process can be applied quite heavily in the handset, since it is required to
receive only one signal. However, in the base station, it is required to receive strong
and weak signals simultaneously, so dynamic range control is less applicable. In 3G
networks, aggressive power control also assists in this process. We consider further
issues of the IF sampled superhet in node B design discussions in Chapter 11.