X- and Ku- Band Small Form Factor Radio Design
by Brad Hall and Wyatt Taylor, Analog Devices, Inc.
In Part 1 of this article, we gave an overview of the super-heterodyne architecture as well as outlined a proposed new architecture called the High-IF Architecture which uses the AD9371 integrated transceiver to receive or transmit a tunable IF at a frequency between 3 and 6GHz. The X or Ku down/upconverter circuit translates between the higher frequency band and the IF. This provides many advantages over the traditional super-het approach.
This second part will provide a deeper review of the details surrounding the frequency planning and show some detailed examples of how a tunable IF grants the ability to work around interfering tones in the receiver as well as avoid spurious interference on the transmitter output. The discussion will finish with a description of a prototype system which was built up and tested.
RX Frequency Planning with the High-IF Architecture
One of the advantages of the High-IF Architecture is the ability to tune the IF. This can be particularly advantageous when trying to create a frequency plan that avoids any interfering spurs. An interfering spur can result when the received signal mixes with the LO in the mixer and generates an MxN spur that is not the desired tone within the IF band.
The mixer generates output signals and spurs according to the equation m×RF±n×LO where m and n are integers. The received signal creates an MxN spur that can fall in the IF band and in certain cases, the desired tone can cause a “crossover” spur at a particular frequency.
For example, if we observe a system designed to receive 12 GHz-16 GHz with an IF at 5.1 GHz, as in Figure 7, the MxN image frequencies that cause a spur to show up in band can be found with the following equation:
RF = ABS ( ( IF+-nxLO)/m)
In this equation, RF is the RF frequencies on the input of the mixer which cause a tone to fall in the IF. Let’s use an example to illustrate. If the receiver is tuned to 13 GHz, that means the LO frequency is at 18.1 GHz (5.1 GHz +13 GHz). Plugging these values into the equation above and allowing m and n to range from 0 to 3, we get the following equation for RF:
RF = (5.1+-nx18.1)/m, n=1,2,3, m=1,2,3
The results are in the following table (Table 1):
In the table (Table 1), the first row (highlighted yellow) shows the desired 13 GHz signal which is a result of a 1×1 product in the mixer. The other highlighted cells show potentially problematic in-band frequencies that can show up as spurs in band. For example, a 15.55 GHz signal is within the 12 GHz-16 GHz desired range. A tone at 15.55 GHz on the input mixes with the LO to generate a 5.1 GHz tone (18.1*2-15.55*2=5.1 GHz). The other non-highlighted rows can also pose a problem but due to their being out of band, they can be filtered by the input band pass filter.
The level of the spur is dependent on several factors. The main factor is the performance of the mixer. Since a mixer is inherently a non-linear device, there exist many harmonics generated within the part. How well the diodes inside the mixer are matched and how well the mixer is optimized for spurious performance will determine the levels on the output. A mixer spur chart is typically included in the datasheet and can help with determining these levels. An example of a mixer spur chart is shown in Table 2, for the HMC773ALC3B. The chart specifies the dBc level of the spurs relative to the desired 1×1 tone.
With this spur chart, along with an extension of the analysis that was done in Figure 8, we can generate a full picture of what MxN image tones may interfere with our receiver and at what level. A spreadsheet can be generated with an output similar to that shown in Figure 8.
In this chart, the blue portion shows the desired bandwidth. The lines show different MxN images and their levels. From this chart, it is easy to see what filtering requirements are needed prior to the mixer to meet interferer requirements. In this case, as previously mentioned, there are several image spurs that fall in band and cannot be filtered. We will now look at how the flexibility of the High-IF Architecture allows us to work around some of these spurs, which is something that the super-het architecture does not afford.
Avoiding Interferers in RX Mode
The chart in Figure 9 shows a similar frequency plan that ranges from 8 GHz -12 GHz with a default IF at 5.1 GHz. This chart gives a different view of the mixer spurs, showing the center tune frequency vs MxN image frequency as opposed to spur level as previously shown. The bold 1:1 diagonal line in this chart shows the desired 1×1 spur. The other lines on the graph represent the MxN images. On the left side of this figure is a representation with no flexibility in the IF tuning. The IF is fixed at 5.1 GHz in this case. With a tune frequency of 10.2 GHz, a 2×1 image spur crosses the desired signal. This means that if you are tuned to 10.2 GHz, there is a good chance that a nearby signal could block the reception of the signal of interest. The right plot shows a solution to this problem with flexible IF tuning. In this case, the IF switches from 5.1 GHz to 4.1 GHz near 9.2 GHz. This prevents the crossover spur from occurring.
This is just a simple example of how blocking signals can be avoided with the High-IF Architecture. When coupled with intelligent algorithms to determine interference and calculate new potential IF frequencies, there are many possible ways to make a receiver that can adapt to any spectral environment. It is as simple as determining a suitable IF within a given range (typically 3 GHz–6 GHz), then recalculating and programming the LO based on that frequency.
TX Frequency Planning with the High-IF Architecture
As with the receive frequency planning, it is possible to take advantage of the flexible nature of the High-IF Architecture to improve spurious performance of the transmitter. Whereas on the receiver side, the frequency content is somewhat unpredictable, on the transmit side it is easier to predict the spurious on the output of the transmitter. This RF content can be predicted with the following equation:
Where the IF is predefined and determined by the tuning frequency of the AD9371, the LO is determined by the desired output frequency.
A similar mixer chart as was done for the receiver channel can be generated on the transmit side. An example is shown in Figure 10. In this chart, the largest spurs are the image and the LO frequencies, which can be filtered out to desired levels with a band pass filter after the mixer. In FDD systems where spurious output may desensitize a nearby receiver, in-band spurs can be problematic and this is where the flexibility of the IF tuning can come in handy. In the example above, if a static IF of 5.1 GHz is used, there will exist a crossover spur on the output of the transmitter which will be near 15.2 GHz. By adjusting the IF to 4.3 GHz at a tune frequency of 14 GHz, the crossover spur can be avoided. This is depicted in Figure 11.
Design Example – Wideband FDD System
To show the performance that can be achieved with this architecture, a prototype RX and TX FDD system was built up with off the shelf Analog Devices components and configured for 12 GHz-16 GHz operation in the receive band and 8 GHz -12 GHz operation in the transmit band. An IF of 5.1 GHz was used to collect performance data. The LO was set to a range of 17.1GHz-21.1 GHz for the receive channel and 13.1GHz-17.1 GHz for the transmit channel. The block diagram for the prototype is shown in Figure 12. In this diagram, the X/Ku converter board is shown on the left and the AD9371 evaluation card is shown on the right.
Gain, noise figure and IIP3 data was collected on the receive downconverter and is shown in Figure 15. Overall the gain was ~20 dB, NF was ~6 dB and IIP3 was ~-2 dBm. Some additional gain leveling could be accomplished with the use of an equalizer or a gain calibration could be performed utilizing the variable attenuator in the AD9371.
The transmit upconverter was also measured, recording its Gain, OP1 dB and OIP3. This data is plotted across frequency in Figure 13. The Gain is ~27 dB, P1 dB ~22 dBm and OIP3 ~32 dBm.
When this board is coupled with the integrated transceiver, the overall specs for receive and transmit are as shown in Table 3.
Overall, the performance of the receiver is in line with a super-het architecture while the power is greatly reduced. An equivalent super-het design would consume more than 5 W for the RX chain. Additionally, the prototype board was fabricated without a priority to decrease the size. With proper PCB layout techniques as well integrating the AD9371 onto the same PCB as the downconverter, the overall size of a solution using this architecture could be condensed to just 4-6 square inches. This shows significant size savings over an equivalent super-het solution which would be closer to 8-10 square inches. Furthermore, size could be reduced further using advanced packaging techniques such as with a multi-chip module (MCM) or system-in-package (SiP). With these advanced techniques, size could be reduced down to 2-3 square inches.
In this two part article, we have shown a viable alternative architecture, the High-IF Architecture, which can enable a sizeable improvement in SWaP over the traditional approach. An overview of the super-heterodyne was given as well as a description of specifications of importance in receiver design. The High-IF Architecture was then introduced along with explanations of the advantages in terms of filtering requirements and level of integration to reduce overall part count. In part 2, we have covered the details surrounding how to go about crafting a frequency plan as well as a description of how the tunable IF can be utilized to avoid interfering signals on the receiver. On the transmit side, where the goal is to reduce output spurious, we have presented a way to avoid in-band spurious as well as an approach to predicting all of the output spurious products that may exist.
The enabler of this architecture is the leaps and bounds that the integrated direct conversion receiver has made recently. With the advent of the AD9371, higher performance is realizable with the advanced calibrations and high level of integration. This architecture will become particularly important in low-SWaP markets in the coming years.
Brad Hall is an RF system applications engineer with Analog Devices working in the Aerospace and Defense Business Unit in Greensboro, North Carolina. He joined Analog Devices in 2015. Previously he worked as an RF hardware design engineer working on signals intelligence systems. He received his Bachelor of Science in Electrical Engineering from University of Maryland in 2006.
Wyatt Taylor is a senior RF systems engineer with Analog Devices, located in Greensboro, North Carolina. He is focused on aerospace and defense radio applications, with a particular emphasis on integrated RF transceivers, small form factor microwave design, and software-defined radio (SDR). Formerly, Wyatt was an RF design engineer at Thales Communications Inc., and Digital Receiver Technology, Inc., in the Maryland area. Wyatt received his Bachelor of Science in Electrical Engineering and Master of Science in Electrical Engineering from Virginia Tech in Blacksburg, Virginia, in 2005 and 2006, respectively.