The emergence of new components requires a hardware comparison of receiver architectures

The tsunami of data being carried over cellular networks is driven by the dramatic advances in smartphones, tablets, and other devices that use these frequency bands to access the Internet. This has led to an increase in technology requirements while forcing vendors to reduce costs. New base stations come in many forms, from traditional rack-mounted equipment to smaller units that can operate on just a few watts of power. The circuitry required to support multiple channels in a tiny base station uses a variety of integration approaches. Given recent developments, how big is the difference between superheterodyne hardware and direct conversion hardware?

Review of basic architecture

By most accounts, Edwin Armstrong invented the superheterodyne receiver architecture in 1918. In this common type of receiver, a radio frequency (RF) signal is mixed with a local oscillator (LO) signal to produce an intermediate frequency (IF) signal, which is then demodulated. The LO frequency is offset from the RF carrier frequency, which creates image signals of that signal. The IF signal is passed through a filter, while all other image signals are rejected by the filter. In modern receivers, the IF signal is converted to a digital signal using an analog-to-digital converter (ADC) and then demodulated in the digital domain (see Figure 1).

Figure 1 : Superheterodyne receiver architecture

A few years later, the direct conversion receiver was developed as an alternative to the superheterodyne receiver. However, unlike the superheterodyne receiver, in the direct conversion receiver the LO frequency is not offset from the received signal frequency but is equal to the received signal frequency. The signal mixer is replaced by two mixers, one receiving the RF signal and the LO signal and the other receiving the RF signal and the orthogonal LO signal. The result is a demodulated output that is digitized by two baseband ADC converters ( see Figure 2) . In other words, the intermediate frequency is zero. The filtering requirements are simplified because only a low-pass filter is required, rather than a band-pass filter as in the superheterodyne receiver.

Figure 2 : Direct conversion receiver architecture

The evolution of hardware

In both cases, architectures have continued to improve over the past few decades. The performance of all integrated circuit (IC) components has continued to improve while consuming less power and requiring less printed circuit board (PCB) area. The resolution and sampling rate of ADCs have also improved to allow wider bandwidth signals and higher input frequencies.

The early appeal of direct conversion receivers was the single frequency conversion to baseband. For decades, superheterodyne receivers have used multiple down-conversion stages. As mixer and filter technology improved, the stages were combined and now a typical superheterodyne receiver has only one frequency conversion stage in the analog section and only one digital down-conversion stage in the digital signal processor.

Another attractive aspect of the direct conversion architecture is low-pass filtering. The superheterodyne architecture requires a bandpass filter at the IF. In many cases, the bandpass filter is of high order or surface acoustic wave (SAW) type. SAW filters require hermetic packaging and are often quite large and expensive. Although there have been great improvements in SAW filter technology and packaging, low-pass filters are still considered more attractive.

Latest Hardware Comparison

To make a reasonable comparison of cost, power, and board space, it is necessary to summarize the components necessary for four receiver channels used in a small base station and suitable for a 20MHz signal bandwidth. Each superheterodyne receiver uses a mixer, a variable gain amplifier, a SAW filter, a second IF amplifier, and a high-speed ADC. Each direct conversion receiver uses an I/Q demodulator, two baseband amplifiers, and two high-speed ADCs. A specific board layout example is used to compare the estimated board space required for these components, and the nominal power consumption is calculated directly from the parameters in the data sheet. The direct conversion architecture is expected to perform much better in both aspects.

Superheterodyne Architecture Example

For the 4 channels of a superheterodyne receiver, two mixers are typically provided in a 5mm x 5mm QFN package, so two of these dual-channel mixer devices are needed. Since the baluns and internal matching components for the RF and LO inputs are integrated, the number of passive components is minimal and the sizes are mostly 0201 and 0402, which are ignored in the comparison because these parts are also required for the direct conversion architecture. Similarly, dual digital VGAs are available in the appropriate frequency range. Such dual VGAs also use 5mm x 5mm QFN packages, so two of these devices are also needed to achieve 4 channels. After the mixer stage, a little filtering may be required, so a few 0402-type inductors and 0201-type capacitors are appropriate. To achieve the required selectivity, the superheterodyne receiver requires a SAW bandpass filter. A separate SAW filter is required for each of the 4 channels. At RF frequencies, the SAW filters can be quite small. In the common IF range of 70MHz to 192MHz, SAW filters are found in 5mm x 7mm packages. Even if the preceding VGA output and the succeeding amplifier input are both 50Ω , the SAW filter will require several matching components. Typically, another gain stage is also required to compensate for the insertion loss of the filter. However, a new 4-channel ADC with integrated amplifiers, the Linear Technology LTM9012-AB µModule ^® ADC, is available in a system-in-package (SiP). The µModule is available in a 15mm x 11.25mm package, which is smaller than a corresponding 4-channel ADC using four differential amplifiers and the associated bypass capacitors and anti-aliasing filter components. The LTM9012 has a gain of 20dB, achieving a 68.5dB signal-to-noise ratio (SNR) and a 79dB spurious-free dynamic range (SFDR). The amplifier and filtering circuitry inside the LTM9012-AB limits the input frequency to approximately 90MHz. Therefore, a 70MHz IF is suitable, rather than the higher IF often achieved with superheterodyne receivers in base station applications. However, this provides the most compact solution.

The LTM9012 represents a different approach to integration. Micromodule or SiP packaging allows a single die to be assembled on a laminate substrate along with a variety of passive components and molded to look like a regular ball grid array (BGA) integrated circuit (IC). In this case, the ADC is optimized for low power and good AC performance using a very small geometry CMOS process. The amplifiers are fabricated using a silicon-germanium (SiGe) process to maximize their performance. These amplifiers are traditional differential amplifiers, so the gain is set to 10V/V or 20dB with resistors. True op amp inputs simplify matching by isolating high frequency sampling interference from the signal path and allow single-ended signals to be paired internally with differential ADC inputs. Most monolithic ADCs with buffered front ends provide no gain at all, remain differential, and only provide isolation from interference. Also beneficial is anti-aliasing filtering, which limits the noise of the wideband amplifier. In terms of overall board space, since all reference and supply bypass capacitors are internal to the package, the overall system design can be very compact without compromising performance. This type of performance impairment often occurs when reference and supply bypass capacitors are placed too far or too close to the digital signals, which can disrupt the data conversion process. Finally, the substrate allows for a natural flow of pinouts: analog inputs on one side of the package, digital outputs on the other.

In this example, the number of active components is 5, with 4 SAW filters and 80 other small passive components (see Figure 3). The total area is about 43mm x 21mm = 903 mm2 ^, but not all of the area is used, so the effective area is about 700mm2 ^. Of course, this is one side of the board, and the design rules of a particular company may allow for a more compact layout. For power calculations, this example uses the LTC5569 as a dual-channel mixer, the AD8376 as a dual VGA, and the LTM9012-AB as a second-stage amplifier and a 4-channel ADC. The mixer is an active component that operates over a wide frequency range of 300MHz to 4GHz , so a single device can be configured to operate in any of the 700MHz to 2.7GHz cellular bands. The device has best-in-class power consumption and also has a rugged input that can withstand strong in-band blocking interferers without significantly degrading the noise figure. The total power consumption of the 4-channel system is 4.9W, which does not include the power that may be dissipated in the resistive divider.

Figure 3 : Example of superheterodyne receiver layout

Direct Conversion Architecture Example

For the four direct conversion channels, our only options are standalone I/Q demodulators, so four of these devices in a 5mm x 5mm QFN package are needed. Some devices, such as the LT5575, have integrated RF and LO baluns to minimize external component count. A little filtering is helpful, and of course some small bypass capacitors. For the low-pass filter, multi-section LC and RC circuits will do the job. For the gain stage, the LTM9012-AB is also suitable. As a 4-channel device, it only supports two direct conversion channels, so a second such device is also needed.

In this example, the number of active components is 6, and there are 84 small passive components, see Figure 4. The total area is approximately 27mm x 24mm = 648mm2 ^. For power calculations, this example uses the LT5575 I/Q demodulator and two LTM9012-ABs. The total power dissipation for the four channels is 5.1W, which does not include the power that may be dissipated in the resistive divider. However, the ADC samples at 125Msps, which is common, but may be more than needed for 10MHz. At 65Msps, the same functionality can be achieved with much lower ADC power consumption. Recalculating the power consumption gives a new total power dissipation of 4.6W.

Figure 4 : Example of direct conversion receiver layout

Feeling and reality

Not too many years ago, superheterodyne receivers used multiple mixers and multiple SAW filters per channel. The SAW filters back then might have been 25mm x 9mm in size. Passive core mixers required additional gain stages to compensate for insertion losses. This not-so-distant history leaves one with the impression that the difference in hardware complexity between superheterodyne and direct conversion receivers is large. As a percentage, the board area used for a superheterodyne receiver is 39% larger than that for a direct conversion receiver, which is a significant difference, but not so great when considering the actual PCB area. 39% of 903 mm2 is 352 mm2, about the size of a thumbprint. On a percentage basis, the ^difference in ^power consumption is not significant at all.

Of course, the perception that size and power are a huge drawback for superheterodyne receivers is relative to the overall size of the base station receiver itself. For a traditional rack-mounted system, a thumb-sized PCB area may not be a big deal. For a base station that is slim and fits in the palm of your hand, a thumb-sized PCB area is very large.

The reality is that integration continues, sometimes slowly, sometimes in leaps and bounds. Reductions in board space and power consumption may apply to one architecture more than the other. Recent examples of applications for the superheterodyne architecture are products such as the LT C5569 dual-channel active mixer. The author is not aware of any dual-channel I/Q demodulators for cellular base station applications, although such demodulators do exist for other applications in the lower frequency range. Recent examples of integration for both architectures are the LTM9012 quad ADC with integrated amplifiers. The device’s LVDS serial interface allows not only a smaller ADC, but also a smaller field programmable gate array (FPGA) or digital signal processor (DSP) than is used for a 4-channel ADC with a parallel interface. However, the direct conversion architecture still requires twice as many ADCs.

The examples discussed above assume that the performance requirements of a cellular base station are such that high-performance components are needed throughout the chain. The products used in the examples utilize optimized semiconductor processes, such as silicon germanium (SiGe) or complementary metal oxide semiconductor (CMOS) processes, which would not be possible to integrate with each other without optimization, or at least not without performance degradation. The performance requirements of certain size base stations may allow the use of highly integrated single-chip transceivers, such as femtocells. Improvements in the integration of blocks in such chips will allow such chips to be used in larger base stations. Here, both architectures encounter a roadblock: signal filters. The low-pass filters used in direct conversion receivers can be implemented in silicon. To date, the band-pass filters used in superheterodyne architectures have proven extremely difficult to implement in silicon. This is the reality today, but it is not necessarily a permanent obstacle. Perhaps one day, a technological breakthrough will occur and built-in highly selective band-pass filters will become feasible. Until then, the direct conversion receiver architecture has a significant advantage because the entire receiver chain can be integrated if performance allows.

in conclusion

The direct conversion receiver architecture for base stations is simpler than the superheterodyne receiver architecture, at least in terms of hardware. Recent products have allowed for the implementation of multi-channel superheterodyne receivers that are much smaller than before. Although still larger based on a percentage comparison, the difference may not be significant. Therefore, the superheterodyne architecture is expected to continue to be the receiver architecture of choice for cellular base stations.