11
Novelda Nanoscale Impulse Radar
James D. Taylor and Dag T. Wisland
CONTENTS
11.2 Overview of Novelda Impulse Radar
11.3 Novelda Nanoscale Impulse Radar CTBV Signal-Acquistion System
11.3.1 Signal Acquisition through CTBV Coding
11.3.1.1 Swept Threshold Sampling
11.3.1.2 Suprathreshold Stochastic Resonance Signal Detection
11.4 Novelda Radar Maximum Unambiguous Range Extension through Staggered Pulse Repetition Frequency
11.5 Novelda Radar—Speed and Power Reduction Innovations
11.6 Novelda Radar—Pipelined Data Output
11.7 Nobelda Radar Development Kits
11.9 Novelda Radar Case Study—Conclusions
11.1 Introduction
The Novelda Nanoscale Impulse Radar illustrates some technically advanced principles of ultrawideband (UWB) impulse radar design. Figure 11.1a shows the 5- × 5-mm package for the 2- × 2-mm complementary metal–oxide–semiconductor (CMOS) chip containing all of the essential radar subsystems [1,5]. The chip works with the radar assembly shown in Figure 11.1b, which includes a circuit board with antennas and serial peripheral interface (SPI) connections for digital computer control. The Novelda radar is a general-purpose radar that the user can program and configure for sensing applications at ranges up to 60 m with 4-mm spatial resolution. The system architecture has several unique processing features including the continuous time binary valued (CTBV) design paradigm. You will see the basic principles of the Novelda radar in many of the radars described in this book.
Like all radar systems, the Novelda radar emits electromagnetic (EM) waves and measures the backscattered (reflected) energy. This type of radar comprises transmitter circuits to generate an impulse signal, a receiver to collect the reflected signal, and special circuitry to collect the range data for target detection and imaging.
Radar systems generally require some type of a clock to regulate the process timing and measure the return-signal time delays. Synchronous clock radar systems have major design limitations with regard to power consumption and speed. Some integrated-circuit processors consume close to 50% of the power budget for clock distribution.
FIGURE 11.1
The Novelda nanoscale impulse radar illustrates the principles of UWB impulse radar design and signal acquisition. (a) A single 2- × 2-mm CMOS chip contains all the essential radar parts. (Adapted from Novelda AS, Novelda Nanoscale Impulse Radar, 2011, http://www.novelda.no/content/radar-ics) and (b) typical Novelda radar with antennas and circuit board. It has a range of 60 m and sample rate spatial resolution of 4 mm. Range resolution depends on the bandwidth of the signal. (Adapted from Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no. With permission from Novelda AS, © 2011.)
The Novelda Impulse Radar design uses the CTBV design paradigm to eliminate the use of the clock found in conventional designs. Instead, the Novelda radar architecture replaces the clock by sequencing computations using the inherent gate delays. This design presents major challenges and variations due to different processing methods. It also produces remarkable performance improvements in terms of increase in speed to the tune of several orders of magnitude and low power consumption [2].
11.2 Overview Of Novelda Impulse Radar
The Novelda radar uses the principles shown in Figure 11.2a, where the transmitted signal gets reflected from distant objects and returns to the receiver with a strength depending on the range and target reflecting area (radar cross section). To achieve the fine spatial resolution, the Novelda uses a Gaussian derived pulse of less than 1-ns duration, as shown in Figure 11.2b. (See Section 1.4 of Chapter 1 for a discussion on Gaussian pulse characteristics.) Radio waves can penetrate many solid materials and can reflect highly diminished signals, which requires increasing the receiver sensitivity through integration and process gain. The Novelda pulse spectrum and typical time-domain waveforms resemble those in Figure 11.2c and d [3,4].
The Gaussian pulse has a duration of less than 1 ns, and this requires extremely fast analog-to-digital converters (ADCs) in the receiver. Because the transmitted signal has limits on the emitted power, this may produce weak return signals that are slightly below noise levels. Similar to other UWB impulse radar designs described in this book, the Novelda radar uses integration of return range cells to recover reflected signals from noise. To understand the differences between the Novelda design and that of other systems, we need to compare continuous and strobed sampling concepts.
Continuous sampling: Most conventional UWB impulse radars continuously sample the reflected signal by capturing the analog received signal and storing it in a digital format for integration and further processing. Quite often, the radar may only capture return signals from a specific range. The sampler design depends on speed, power, and range resolution requirements. If the sample requires capturing a several-gigahertz UWB signal, then the sampler must sample at tens of picosecond intervals. This implies high clock speeds, requiring high amounts of power and silicon for storage memory.
FIGURE 11.2
Impulse radar fundamentals: (a) Impulse radars transmit a signal and detect the reflected signal from objects. Strength of the return signal depends on the range and target size. (b) Fine spatial resolution requires using a signal with duration of less than 1 ns (30-cm physical length). (c) A typical Novelda frequency spectrum. (d) The emitted Gaussian impulse has this format. (Adapted from Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no; Novelda AS, NVA6100 Preliminary Datasheet Impulse Radar Transceiver System, 2010, www.novelda.no. With permission from Novelda AS, © 2010.)
FIGURE 11.3
Conventional radar signal sampling compared with the Novelda nanoscale impulse radar strobed sampling technique. The Novelda signal-acquisition system does all sampling in a continuous time domain without the digital controls of conventional designs. This reduces the power consumption from clock circuits and uses delays in the elements to establish the sampling interval. (Adapted from Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no. With permission from Novelda AS, © 2011.)
Strobed sampling: The Novelda radar uses the strobed sampling concept shown in Figure 11.3. The radar receiver samples the backscattered electromagnetic energy after a given time offset. This offset represents the signal time-of-flight relative to the time of transmission or, in practical terms, the number of range increments to the target. Because there are 512 range bins in the Novelda radar (13-ns window), sampling becomes continuous when the pulse repetition interval (PRI) is 13 ns or less. For some sample rates, strobed sampling achieves the same effect as continuous sampling. Strobed sampling radar systems require highly accurate time offsets and high-speed sampling. Implementing these requirements with conventional synchronous digital technologies presents major challenges.
The Novelda radar operates in the continuous time domain without any clock. As shown in Figure 11.3, this moves the signal conversion point from discrete to continuous time unlike conventional systems. Because the processing keeps the received signal in the continuous time domain as long as possible, it enables distance measurement in the range of picoseconds (millimeters).
11.3 Novelda Nanoscale Impulse Radar CTBV Signal-Acquistion System
The conventional strobed receiver takes only a single point at a time. Searching for reflecting objects in the signal path requires a trace of the received signal over time to correlate a template or expected model. Some radar systems vary the point-in-time (range) strobe to collect range data over some desired range, for example, 4.5−5.1 m.
11.3.1 Signal Acquisition through CTBV Coding
The Novelda radar improves on the basic strobed sampler concept using CTBV coding. A high-speed 1-bit quantizer makes a good analogy to the CTBV coding system.
Figure 11.4 shows the CTBV-coded signal paths as dashed lines (----). A 1-bit quantizer converts the incoming signal to a binary sequence (11001011001 . . .). The coded signal is transmitted to a number of parallel samplers. As in the strobed sampler, the system delays the trigger signal by a time offset. The trigger signal connects to the parallel samplers, each of which has a slightly different time offset. The parallel samplers receive the incoming CTBV signal in a rapid sequence and produce an output readable as normal digital values.
The Novelda radar samples the incoming signal at a number of different ranges (typically 512) for each transmitted pulse. In this way, it works like many traditional radars set to detect only one range and operating in parallel. Because the CTBV sampler samples only single bits due to the single-bit quantizer, this gives little information.
The improvement comes because the Novelda radar recreates the incoming analog signal as digital values for use in a computer by two different methods: swept threshold sampling and suprathreshold stochastic resonance signal detection. These methods depend on the amount of noise in the system.
FIGURE 11.4
The Novelda radar CTBV signal-acquisition system. This uses the suprathreshold stochastic resonance phenomenon that uses naturally occurring system noise to help signal detection and increase the signal-to-noise ratio. (Courtesy of Novelda AS, Norway, © 2011.)
11.3.1.1 Swept Threshold Sampling
Instead of using a single threshold in the quantizer, the signal goes through a process called swept threshold sampling, which shares some similarities to flash or direct analog-to-digital conversion.
The flash (or direct conversion) digitizer method uses a bank of comparators to sample the input signal in parallel, and each comparator fires at its decoded voltage range. This generates an output indicating a code for each voltage range. While this direct analog-to-digital conversion can provide gigahertz sampling rates, it has a resolution limited to 8 bits or less because the number of comparators needed, (2N − 1), doubles with each additional bit, N. Flash ADC devices have other disadvantages including large physical size, high input capacitance, and high power dissipation.
The Novelda radar swept threshold converter uses a single quantizer instead of a voltage comparator bank. The sampling procedure repeats while the threshold sweeps over the range of interest. For each threshold step, the conversion results accumulate in a digital counter. In a noiseless environment, the system resolution depends on the step size of the threshold sweep. Figure 11.5 shows the operation of the swept threshold sampler.
After the completed sweep, the counter indicates the converted value. For example, take the case of the sampler sweeping the threshold from 0 to 1 V in steps of 0.1 V. For an input of 1 V, this produces a conversion value of 100; for example, a 0.6-V input converts to 60 and a 0-V input converts to 0. For noisy signals, we can repeat the procedure several times and get processing gain in the form of averaging. However, recovery of weak signals may require substantial averaging.
In the case of swept threshold sampling, we use simple digital counters as digital averagers. The long signal integration time for digital conversion will limit comparable signal-recovery system performance. Lossless digital counters have limits set only by the number of bits.
The swept threshold sampling process opens the possibility to add a large processing gain early in the Novelda radar signal-acquisition process [1]. Figure 11.6 shows the logical process used in the radar chip.
11.3.1.2 Suprathreshold Stochastic Resonance Signal Detection
The Novelda radar uses the suprathreshold stochastic resonance phenomenon to recover weak signals. This concept exploits the system noise to increase the signal-to-noise ratio of the received signal [1].
Stochastic resonance describes a phenomenon by which signals too weak to pass through a nonlinear system can pass through it by adding noise to the input. The added noise will make the weak signal pass through a nonlinear system from time to time. Figure 11.7 explains the stochastic resonance sampler, sampling process, and simulated results. This stochastic resonance sampler uses the same circuit as the swept threshold sampler but sets the threshold VT equal to the input signal level of 0.5 V. In the noiseless signal sampling case, the thresholder output would remain equal at each repeated sampling and make it the equivalent of a 1-bit sampler, which detects whether the signal exceeds or falls below the DC level. In the case of the noisy signal, Vin(τ) will sometimes exceed the threshold. For the ideal case, where Vin(τ) = 0.5, a noisy signal Vin(τ) will exceed the threshold in 50% of the samples. In an ideal case of Vin(τ) = 0.65, the signal will exceed the threshold a greater percentage of the time and depend on the amount of noise in the input signal. For the case with 93% probability of Vin(τ) exceeding the threshold, 1000 samples will result in a counter value of approximately 930. By knowing the probability density function (PDF) of the noise, we can use the counter value to calculate the best guess for the ideal Vin(τ) or Vin recovered(τ). We will call the noise in this recovered value σN recovered.
FIGURE 11.5
The Novelda nanoscale impulse radar swept threshold sampler. (a) Sampler block diagram. (b) The sampler compares the received backscatter signal Vin(t) = ideal Vin(t) + noise with a threshold VT and samples the resulting unclocked digital signal at time τ after the radar pulse emission. Strobed sampling means taking a single sample at a given interval τ. Repeating the sample several times while sweeping the threshold recovers the input signal amplitude. (Hjortland, H.A., Wisland, D.T., and Lande, T.S., Thresholded samplers for UWB impulse radar, ISCAS 2007 IEEE International Symposium on Circuits and Systems, May 27-30, 2007, 1210-1213, © 2007, IEEE.)
To evaluate the possibilities of stochastic resonance sampling, a simulation compares the results of swept threshold and analog average sampling. Figure 11.7 shows the number of samples needed to get desired σN recovered values of 0.001, 0.00316, and 0.01 plotted against the input noise σN. The plot shows a surprising result that both stochastic and swept threshold samplers perform very close to the analog sampler at high input signal noise levels. Both the stochastic and swept threshold samplers offer good choices for the Novelda radar design.
FIGURE 11.6
The Novelda nanoscale impulse radar sweep controller principle of operation. The chip implements the processes of Figure 11.5 by following this general diagram. (From Novelda AS, NVA6100 Preliminary Datasheet Impulse Radar Transceiver System, 2010, www.novelda.no. With permission from Novelda AS, Norway, © 2010.)
11.4 Novelda Radar Maximum Unambiguous Range Extension through Staggered Pulse Repetition Frequency
Simple radar systems operating at a constant pulse repetition frequency (PRF) have a maximum unambiguous range (MUR) based on the reflection time from one pulse to the next. For example, the 100-MHz PRF of the Novelda radar has a MUR of 1.5 m. If the time of a reflection exceeds the pulse interval, then it can appear as a closer range target.
The Novelda radar overcomes this MUR problem by using a staggered PRF. This introduces a certain amount of randomness to the pulse transmission interval. Under these conditions, the second and subsequent reflections will appear in the receiver at slightly different times relative to the most recently transmitted pulse. Because of the range increment averaging of the receiver, these uncorrelated pulsed will cancel out instead of appearing as ghost reflections.
11.5 Novelda Radar—Speed and Power Reduction Innovations
Reduction in energy consumption presents a major challenge in microelectronic design. Lower energy consumption gives a longer battery life and less heat dissipation, which enables smaller devices and new applications. The total heat dissipation requirement depends on the power consumption and time. For typical applications, the Novelda radar works with an external controller unit for further signal processing, communications with other components, and so on. Modern microcontrollers can reduce power consumption by using sleep modes, which turns-off peripherals, and by reducing clock speed or even completely halting the processor core while waiting for an input change or the event to occur.
FIGURE 11.7
Stochastic resonance sampling. (a) Block diagram of the sampler. (b) Stochastic resonance sampling process. (c) Comparison of the stochastic sampler with swept threshold and analog average methods. (Hjortland, H.A., Wisland, D.T., and Lande, T.S. Thresholded samplers for UWB impulse radar, ISCAS 2007 IEEE International Symposium on Circuits and Systems, May 27-30, 2007, 1210-1213, © 2007, IEEE.)
The input stage consumes the most power (and, of course, PRF independent), so speed reduction does not help reduce power consumption very much. However, the input-stage power reduction gives great power savings.
The Novelda radar has a built-in sweep controller, which enables it to operate almost independent of the external controller. The conventional synchronous controllers found in other systems turn on at specified intervals and wait for the data availability, which could take place any time from a couple of microseconds to a millisecond. The asynchronous controller requests a sample; the radar chip will then capture and process the incoming signal and notify the master control when the new data become available. This eliminates the power consumed by a synchronous controller polling for an asynchronous event. The user can configure the Novelda radar chip to turn off power-hungry circuits when not needed. Figure 11.8 shows the Novelda power consumption cycle and an asynchronous controller.
FIGURE 11.8
The Novelda nanoscale impulse radar asynchronous controller turns the chip components on and off based on signal availability. This reduces power-consumption and heat-dissipation problems by applying power only when needed to check for events such as a received signal. Compare this with the normal synchronous controllers which turn chip components on and off at preset intervals for events such as received-signal availability. (Adapted from Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no. With permission from Novelda AS, © 2011.)
11.6 Novelda Radar—Pipelined Data Output
All collected radar range data must transfer for further processing and conversion to useful information or display as range, images, warnings, and so on. The Novelda nanoscale impulse radar interfaces with a computer through an SPI, so reading the sampled frame back to the master controller can take up to several hundred microseconds. To get around this problem, the Novelda radar uses double buffering of the output buffers, which store the sampled frame data. This permits pipelining the sample readout operations by notifying the master when a new frame becomes available and immediately copying the result to a second on-chip buffer. This resets the primary result buffer and starts processing a new sample. It can then continue to read the next sample frame while the radar chip prepares the next frame in the background. Figure 11.9 shows the basic concept behind output pipelining.
11.7 Nobelda Radar Development Kits
Novelda provides an off-the-shelf general-purpose impulse radar sensor in which the user can put a wide variety of applications, such as the ones shown in Table 11.1. Figure 11.10 shows a typical development kit with sinuous antennas.
FIGURE 11.9
The Novelda radar increases data throughput by pipelining. (a) Conventional data output with a serial peripheral interface (SPI) can take up to several hundred microseconds and requires a data capture, notify, and read process. This limits the sampling speed to the data transfer rate by keeping the sampler waiting and idle. (b) The Novelda radar uses a double buffering system which allows the sampler to capture data and move it to a separate input/output buffer to await reading. The sampler can then proceed to capture a new range data sample while the previous sample waits to transfer to the control computer. (Adapted from Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no. With permission from Novelda AS, © 2011.)
TABLE 11.1
Potential Applications of Novelda Nanoscale Impulse Radar
FIGURE 11.10
The Novelda nanoscale impulse radar developmental kit comes with sinuous antennas and appropriate processors. The kits come in frequency ranges from 6.0 to 9.1, 0.450 to 3.5, and 0.85 to 9.6 GHz. Accompanying operating interface control software and sample applications help the user operate the radar and modify it for special applications. (Adapted from Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no. With permission from Novelda AS, © 2011.)
TABLE 11.2
Specifications of Novelda Nanoscale Impulse Radar Chip
The Novelda impulse radar uses the two different chip configurations shown in Table 11.2. Both chips have the same characteristics except for the frequency ranges. Figure 11.11 shows the block diagram of the NVA 6100 impulse radar chip.
A developer can order one of the three different kits described in Table 11.3 and get a ready-to-use impulse radar for prototyping and concept development. All kits have an Atmel microcontroller (AT91SAM7S) for a communications hub between the radar chip SPI and the computer. The kits communicate with the host computer through a USB connector, which can also power the circuit boards.
The development software includes RadarScope, a graphical user interface (GUI) designed to help new users quickly learn the technology. Additional features include Radarlib, a Windows-C library complete with an easy-to-use applications programming interface (API) and example applications for popular environments such as MATLAB®. The user can program the microcontroller through the USB, in addition to a JTAG access port that requires an external JTAG programmer and a software.
FIGURE 11.11
The Novelda NVA6100 radar 2- × 2-mm CMOS chip comes in a 5- × 5-mm package. When combined with antennas, control circuitry, and a computer, it provides a complete impulse radar. (From Novelda AS, NVA6100 Preliminary Datasheet Impulse Radar Transceiver System, 2010, www.novelda.no. With permission from Novelda AS, © 2010.)
TABLE 11.3
Novelda AS Radar Development Kits
11.8 Novelda Radar Antennas
Antennas can have a great influence on the radar system’s performance. Novelda AS provides antennas for development kits with a narrow beam width and small size. Figure 11.12 shows a 3D model of the sinuous antenna with a dielectric lens for increased directivity. Table 11.4 gives the antenna technical specifications and characteristics.
FIGURE 11.12
Novelda impulse radar antennas come in three sizes as described in Table 11.4. (a) Typical antenna shown with cover and connector. (b) The antenna uses a sinuous antenna combined with a dielectric lens for a 40° angle and 6-dBi typical gain. (From Andersen, N., and Land, T.S., Nanoscale Impulse Radar, Novelda white paper, 2010, www.novelda.no. With permission from Novelda AS, © 2011.)
TABLE 11.4
Novelda AS Impulse Radar Antenna Characteristics
| Development Kit | Frequency Range (GHz) | Size | Angle | Typical Gain [dBi (flat)] |
| NVA R620 | 0.5–3.0 | 120 × 120× 22 mm (without lens) 67 mm (with lens) | 40° opening angle with lens (80° without) | 6 |
| NVA R630 | 3.0–6.0 | 50 × 50 × 17 mm (without lens) 55 mm (with lens) | 40° opening angle with lens (80° without) | 6 |
| NVA R640 | 0.85–9.6 | 45 × 45 × 14 mm (without lens) 45 mm (with lens) | 35°-40° opening angle with lens (65°-80° without) | 6 |
Note: Information current as of September 2011. Contact Novelda AS for latest specifications.
Source: Novelda AS, Novelda Nanoscale Impulse Radar, 2011, Online: http://www.novelda.no/content/radar-ics; Andersen, N. and Land, T.S. Nanoscale Impulse Radar, Novelda white paper, 2010. Online: www.novelda.no (accessed on April 20, 2011).
11.9 Novelda Radar Case Study—Conclusions
The Novelda impulse radar provides a ready-to-use high-resolution range sensor that can connect to personal computer and adapt to a wide variety of practical applications. It illustrates several important developments in microcircuit-based radar design:
Extreme miniaturization in the form of a 2- × 2-mm CMOS chip
The use of range bins to collect and integrate reflected signals to get extended ranges with low emitted power
The use of the swept threshold sampler to collect return data
Improved detection by use of the stochastic resonance, which provides signal detection on par with analog averaging and swept threshold detection in the presence of noise
Asynchronous data collection from double-buffered outputs to decrease power consumption
Acknowledgments
My (James D. Taylor) special thanks to Dr. Dag T. Wisland, CEO of Novelda AS, Kviteseid, Norway, for agreeing to coauthor this chapter. He provided the specialized technical information regarding the Novelda Nanoscale Impulse Radar and reviewed my work for technical accuracy.
References
1. Novelda A.S.. Novelda Nanoscale Impulse Radar, 2011. Online: http://www.novelda.no/content/radar-ics (accessed on April 14, 2011).
2. Andersen, N. and Land, T.S. Nanoscale Impulse Radar, Novelda white paper, 2010. Online: www.novelda.no (accessed on April 20, 2011).
3. Novelda A.S. NVA6100 Preliminary Datasheet Impulse Radar Transceiver System, 2010. Online: www.novelda.no (accessed on April 20, 2011).
4. Hjortland, H.A., Wisland, D.T., and Lande, T.S. Thresholded Samplers for UWB Impulse Radar, ISCAS 2007 IEEE International Symposium on Circuits and Systems, May 27-30, 2007, pp 1210-11213.
5. Løvhaug, N.R. and Lie, E. Unique processor chips with tiny radar transceivers, R&D Magazine, May 6, 2011. Online: http://www.rdmag.com/News/2011/05/Information-Tech-Electronics-Unique-processor-chips-with-tiny-radar-transceivers/ (accessed on April 5, 2011).