Where zero IF wins: 50 percent smaller PCB footprint at 1/3 the cost – Part 1

Where zero IF wins: 50 percent smaller PCB footprint at 1/3 the cost – Part 1

Feature articles |
By Graham Prophet

Historically, this architecture has been withheld from applications that demand high performance. However with the demand for wireless growing around us and the rapidly congested spectra, a change is required in order to continue economically deploying radios in the infrastructure that supports our wireless needs.


Contemporary Zero-IF architectures can satisfy these needs as many of the impairments normally associated with these architectures have been resolved through a combination of process, design, partitioning and algorithms. New advances in ZIF technology challenges the current high performance radio architectures and introduces new products with breakthrough performance to enable new applications previously beyond the reach of ZIF. This article will explore the many benefits of ZIF architectures and introduce new levels of performance that they bring to radio designs.


Challenges of the Radio Engineer

The transceiver architect today (see note 1) is challenged by a growing list of demands driven by our ever increasing requirements for wireless devices and applications. This leads to the continual need to access more bandwidth. The designer has moved over the years from a single carrier radio to multi-carrier. As spectrum becomes fully occupied in one band, new bands are allocated; now there are more than 40 wireless bands that must be served. Because operators have spectrum in multiple bands and these resources must be coordinated, the trend is towards carrier aggregation; carrier aggregation leads to multi-band radios. This all leads to more radios, with higher performance, requiring better out-of-band rejection, improved emissions and dissipating less power.


While the demand for wireless is rapidly increasing, the power and space budget are not. In fact, with an ever increasing need to economize both in power and space, reducing both the carbon footprint and the physical footprint are very important. To achieve these goals, new perspectives on radio architectures and partitioning are required.




In order to increase the number of radios in a particular design, the footprints must be made smaller for each radio. The traditional way to do this is to progressively integrate more and more of the design onto a single piece of silicon. While this may make sense from a digital perspective, integration of analogue functionality for the sake of integration doesn’t always make sense. One reason is that many analogue functions in a radio cannot effectively be integrated. For example, a traditional IF sampling receiver is shown in Figure 1 below.


There are four basic stages to an IF sampling architecture: low noise gain & RF selectivity, frequency translation, IF gain and selectivity, and detection. For selectivity, SAW filters are typically used. These devices cannot be integrated and therefore must be off-chip. While RF selectivity is provided by piezoelectric or mechanical devices, occasionally LC filters are used for the IF filter. While LC filters may occasionally be integrated on monolithic structures, the compromise in both filter performance (Q and insertion loss) and the required increase in sample rate of the digitizer (detector) increase the overall dissipation.


Digitizers (analogue to digital converters) must be done on low cost CMOS processes to keep the cost and power reasonable. While they certainly can be fabricated on bipolar processes, this results in both larger and more power hungry devices which runs counter to optimization for size. Thus standard CMOS is the desired process for this function. This becomes a challenge for integration of high performance amplifiers, particularly the IF stage. While amplifiers can be integrated on CMOS processes, it is difficult to get the performance required from processes that are optimized for low power and low voltage. Furthermore, integrating the mixer and IF amplifier on chip require that the interstage signals be routed off-chip to access the IF and anti-alias filters prior to being digitized, sacrificing much of the benefit of integration. Doing so is counterproductive to integration as it increases the pin count and package size. Additionally, each time critical analogue signals pass through a package pin, a compromise in performance is made.

Figure 1. Traditional IF sampling receiver


The optimal way to integrate is to re-partition the system to eliminate the items that cannot be integrated. Since SAW and LC filters cannot be effectively integrated, the best option is to determine how to get rid of them by re-architecting. Figure 2 shows a typical Zero IF signal chain that achieves these goals by translating the RF signal directly to a complex baseband, completely eliminating the need for an IF filter and IF amplifiers. Selectivity is achieved by introducing a pair of low pass filters into the IQ baseband signal chain that can be integrated as active low pass filters instead of off-chip lossy fixed IF devices. Traditional IF SAW filters or LC filters are by nature fixed while these active filters can be electronically tuned, often from the hundreds of kHz range through hundreds of megahertz. Changing the bandwidth of the baseband allows the same device to cover a broad range of bandwidths without having to change a bill of material or switching between different fixed IF filters.



Figure 2. Typical zero IF sampling receiver


Although not intuitively obvious from the figure, zero IF receivers can also cover a very broad range of RF frequencies simply by changing the local oscillator. Zero IF transceivers provide a truly broadband experience with typical coverage continuously from several hundred megahertz up to around 6 GHz. Without fixed filters, truly flexible radios are possible, greatly reducing and possibly eliminating the effort required to develop band variations of the radio design. Because of the flexible digitizers and programmable baseband filters, zero IF designs not only deliver high performance, but also significant flexibility in adapting to a wide range of frequencies and bandwidths while maintaining nearly flat performance without the need to optimize analogue circuits (filters, etc) for each configuration – true software defined radio (SDR) technology.


This too adds greatly to the reduction of footprint by elimination of banks of filters for applications that must cover multiple bands. In some cases, the RF filter may be completely eliminated introducing a completely wideband radio that requires virtually no effort to change bands. By elimination of some devices and integration of others, the required PCB footprint for a Zero IF design is greatly reduced, not only simplifying the re-banding process, but also reducing the effort to change the form factor when required.


Smallest footprint


A direct comparison of the PCB area for each of these architectures (Figures 3 and 4) shows that for a dual Rx path, the respective PCB area for a reasonable implementation gives 2880 mm² (18 by 160 mm) for IF sampling and 1434 mm² (18 by 80mm) for zero IF sampling. Not counting the potential elimination of RF filters and other simplifications [note 2], the zero IF architecture offers the possibility of reducing the radio footprint by up to 50% as compared to current IF sampling technology. Future generation designs can potentially redouble these savings with additional integration.



Figure 3. Typical IF sampling layout




Figure 4. Typical Zero IF sampling layout



Lowest cost

From a direct bill of material point of view, the savings when moving from an IF sampling system to a zero IF architecture are 33%. Cost analyses are always difficult, however a thorough examination of Figure 1 and Figure 2 shows that many of the discrete items are eliminated including the IF and anti-alias filtering and that the mixer and baseband amplifiers are integrated. What is not obvious is that because zero IF receivers inherently offer out-of-band rejection not offered in traditional IF sampling architectures, the overall external filtering requirement are greatly reduced.


There are two contributors within the zero IF architecture that drive this. The first is the active baseband filter that provides both in-band gain and out-of-band rejection. The second is the high sample rate, low pass sigma delta converter used to digitize the IQ signals. The active filter reduces the out-of-band component while the high sample rate of the ADC moves the alias point out to a sufficiently high frequency that external antialiasing filtering is not required (because the active filter has sufficiently rejected the signals).



Figure 5. Active baseband filter & ADC


By applying the baseband signals to an active filter, Figure 5, high frequency content is rolled off. The ADC then digitizes and ultimately filters any residual output from the low pass filter. The cascaded system results are shown in Figure 6. This figure shows what a typical receiver performance might look like with the compound effect of an active filter and sigma delta ADC. Shown here is a typical 3 dB de-sense of both in-band and out-of-band power. Note the improvement in out-of-band performance without any external filtering.


For similar levels of performance, IF sampling receivers rely on discrete IF filtering such as SAW technology for selectivity and protection from out-of-band signals and to prevent aliasing of wideband signals and noise alike from aliasing back in band. IF sampling architectures must also be protected from other unwanted mixer terms including the half-IF term which drives additional RF and IF filtering requirements as well as restricts sample rates and IF planning. The zero IF architecture has no such frequency planning restrictions.




Figure 6. Typical Zero IF out of band rejection


Depending on the design and application, this native rejection reduces or eliminates external RF filtering requirements. This results in a direct savings by their omission as external RF filters can be relatively expensive depending on the type. Secondarily, removal of these lossy devices may allow the elimination of RF gain stages, saving not only cost but reducing power and improving linearity. All of these add to the savings delivered by re-partitioning and smart integration.


As noted, it is difficult to assess cost as this depends greatly on volume and vendor agreements. However a detailed analysis shows that zero IF architectures typically reduce the full system cost by up to 1/3 through the impact of integration, elimination and reduction in requirements. It is important to remember that this is system cost and not device cost. Because more functions are being placed in fewer devices, some device costs may increase while overall system cost are reduced.


Beyond bill of material costs, the integrated zero IF receiver addresses a few other areas. Because integrated systems reduce the number of devices in the system, assembly costs are lower and factory yields are higher. Because there are fewer discrete devices, alignment time is shorter. These items together reduce factory costs.


Because the zero IF receiver is truly wideband, engineering costs are reduced to re-band. IF frequencies must be carefully chosen in IF sampling systems, but with zero IF systems, there is no careful planning required. New bands may be added largely by changing the local oscillator. Additionally, because many application do not require an external RF filter when zero IF is used, further simplifications may result. Overall, cost savings can be substantial when considering a zero IF solution when the direct cost is considered alongside of the manufacturing and engineering costs outlined above.


Note 1; While this discussion primarily focuses on the receiver, this discussion applies to transmitters as well. For a transmitter, zero IF has been the accepted architecture of high performance for more than a decade.


Note 2; As examined here, the typical zero IF receiver also includes a full transmit path (AD9371) within the same package.


Watch for part 2 of this article…



If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles