By Jan Whitacre, Agilent Technologies
London. City of literally thousands of “black cabs” driven by qualified London cabbies. To get their cab license, they must pass “the knowledge” — the ability to describe, by street name and turn, the route they would take from anywhere to anywhere else in the city. They train by spending months touring the city by bike or motor scooter using the A to Z street map book to familiarize themselves with street layouts, one-way systems and shortcuts.
So it is with measurement trends today. Traditional measurement science has been aimed at giving us the data (the map book), but, almost entirely regardless of the field of endeavour, that’s no longer enough. Today’s and tomorrow’s design and measurement tools need to deliver to their users the knowledge that the raw measurement data provides, reduced to a form meaningful in context and providing the user with the means to move forward. Integration often means that, even at the part level components are multi-functional, with inputs and outputs that may not even be in the same physical domain.
Let’s look at an example from the wireless LAN world. The new 802.11ac standard is in the process of ratification, and will bring much improved speed and capacity to our home and small office networks. Key changes are a much wider RF bandwidth, higher modulation density and more spatially multiplexed streams. Routers will be manufactured in the millions and will have a selling price in the order of $100, so taking only cents out of the bill of materials would make a huge difference to manufacturing cost and company profitability. One of the high-cost components is the RF amplifier, which now needs to work over a broader bandwidth (including split 80+80 MHz non-contiguous channels – see Figure 1), and be linear.
To reduce its cost, manufacturers will try to use less-expensive and, therefore, likely less-well-specified parts and correct for output linearity errors at its inputs. OK so far? But the amplifier has DIGITAL I/Q inputs and an RF output, so creating a correction matrix just became a cross-domain measurement challenge!
Digital predistortion to improve the linearity of power amplifiers typically requires generation and measurement of signals that are 3 to 5 times the bandwidth of the amplifier being linearized. Control software is used to generate a stimulus waveform which is downloaded to an RF signal generator and applied to the power amplifier. The amplifier’s response is captured using a signal analyzer and compared with the desired signal to create the predistortion matrix. The predistorted signal is then sent to the power amplifier and the response checked.
Figure 2 shows an example of a system that can determine the correction matrix.
Here’s another example: Operators are rolling out 4th generation cellular networks based on the 3rd Generation Partnership Project’s Long Term Evolution (3GPP LTE) standards. One of the ways to improve service, particularly at the edge of a cell’s range, is to use a technique known as beamforming at the base station transmitter site. Best suited to the Time Division Duplex (TDD) variant of LTE, where the uplink and downlink work at the same frequency, beamforming works by exploiting the interference patterns that exist whenever the same signal is transmitted from two or more spatially separated transmission points.
Using a linear array of base station (technically known as the eNB) antennas for transmission and reception, and by carefully controlling the relative magnitude and phase weightings applied to each information symbol copy transmitted on each antenna element, the resultant beam pattern is modified in real time to focus transmit energy and receive sensitivity in the direction of a specific mobile device (User Equipment or UE), and to minimize interference with a device communicating with an adjacent base station.
Optimal downlink transmission beam selection is primarily driven by some knowledge of the UE position within the cell. The eNB would typically estimate the optimum weightings through direct measurement of the received uplink reference signals observed across the eNB receiver array. This information can then be used to calculate the uplink Angle of Arrival (AoA) as well as decompose the channel characteristic matrix.
The illustration in Figure 3 shows eNB1 is communicating with target device UE1, with the eNB1 transmission using beamforming to maximize the signal power in the direction of UE1. At the same time eNB1 is attempting to minimize interference to UE2 by steering the power null location in the direction of UE2. Similarly eNB2 is using beamforming to maximize reception of its own transmission in the direction of UE2, whilst minimizing interference to UE1.
How do you check that newly-developed equipment gets it right? One of the main test challenges for beamforming is the need to verify and visualize the beamforming signal performance at the physical RF antenna array, in order to validate the eNB RF antenna calibration accuracy and baseband encoded beamforming weighting algorithm correctness.
The key requirement to making good measurements is system calibration. Figure 4 shows a typical system.
A correction wizard guides the system calibration process, prompting the user to connect the signal analyzer channel 1 measurement cable to the first output port of the two-way calibration splitter at the injection point represented by a dotted line. All the cross-channel characterization measurements will be made referenced to channel 1. The correction wizard is able to characterize the cross-channel corrections required to compensate the signal analyzer beamforming measurements for all mismatch effects inherent in the measurement cables, connectors, splitters, and attenuators.
From a development perspective, the use of multi-antenna beamforming transmission presents some specific test challenges including the need to verify correct implementation of the base station baseband receive/transmit algorithms used to generate beamforming weightings. In this case, measurement capability built into the network equipment itself, both at the base station and the mobile device, must be validated. Product development and network conformance testing must include the ability to stress this capability under varying operating conditions. Again, the ability to apply the measurement results and move forward depends on a clear understanding of the measurement concepts, the knowledge of overall system behaviour, and the accurate validation of calibration performance. It’s a complex problem that includes elements of RF, digital baseband and complex computational design elements, applied to a real-time, fluid environment. Success is measured by the content of calls from users: are the users (you and me) delighted with the performance of their new gadgets, or will they consign them to the technology scrapheap and move on to the competition?
So, is there a place in today’s and tomorrow’s world for the pure RF engineer? In summary, I’d say the answer is “no.” These two examples show a need for a much broader understanding of system behaviour that is by no means uncommon in the technology sector. There are examples of the same cross-domain scenarios in the avionics, automotive and covert communications sectors, to name just a few. Where previous generations of engineers could specialize in one domain (computing, digital signal processing, logic analysis, os RF, as examples), the engineering skills involved in today’s designs aren’t just from a single domain — they’re about the knowledge of how the domains interact: how to get from anywhere to anywhere else in the world of the massively integrated technology that is life in the 21st century.
this article to a friend!