There’s a popular YouTube video of a 1959 Chevy Bel Air crashing into a 2009 Chevy Malibu in slow motion. The Malibu does surprisingly well, while the Bel Air’s disintegration is not so kind to its crash test dummy. Obviously, we’ve learned a lot in the past 50 years about making crashes survivable.
Yet we’ve only recently begun designing-in electronics that make many kinds of crashes avoidable—even though drivers haven’t been getting any smarter—such as lidar-based adaptive speed control and rollover sensing coupled to automatic traction control. Examples of those two systems, as well as a better way of handling load-dump transients on electronics buses, illustrate the way chip designers think about implementing active automotive safety systems.
Adaptive Speed Control
Adaptive speed control systems work much like the simple control systems that have been used for decades—the driver sets the car’s cruise control for the desired speed, and the vehicle maintains it until it’s interrupted. The difference appears whenever the vehicle is overtaking a slower vehicle.
In that case, the adaptive speed control reduces the vehicle’s speed below the set value to match the speed of the vehicle ahead while maintaining a safe distance. To achieve that level of safety, light detection and ranging (lidar) has evolved from a tool used by police to catch speeders to the subsystem that detects the presence of other vehicles and measures the distance between us and them.
Lidar subsystems may use continuous-wave (CW) or pulsed signals. CW systems detect the phase-shifted echo of the transmitted signal, using a phase comparator in the receiver. The phase shift signals distance, and the rate of change corresponds to the rate of closure.
Pulsed systems calculate the time of flight (TOF) of short light pulses to determine distance to the car ahead and the rate of closure. Generally, CW systems cost too much to implement for use in automotive applications, so lidar systems using a pulsed laser predominate.
The common elements of these systems include a power supply, an electrical signal source, a power amplifier, and a transmitter to send out a signal, plus a receiving sensor, an amplifier, a signal conditioner, and a high-speed analog-to-digital converter (ADC) to deliver a digitized version of the received signal to a buffer memory, from which a DSP, FPGA, or microcontroller recovers the data for processing (Fig. 1). Texas Instruments uses an ADC that buffers its own output, allowing the data-conversion portion of the IC to be shut down while the DSP or FPGA digests the data from the previous pulse.
The size of the gap between cars that can be accommodated depends on laser output power, beam width and alignment, atmospheric characteristics (e.g., fog), target reflectivity, and receiver sensitivity.
While laser driver design is reasonably straightforward, the receiver has several critical design requirements. For instance, the receiver circuit designer may use any of three types of detector: silicon PIN detectors (“PIN” refers to the semiconductor stack, a sandwich of P-type, intrinsic, and N-type material), silicon avalanche photo diodes (APDs), or (unlikely) photomultiplier tubes. APDs offer the best combination of high speed, high sensitivity, and ruggedness.
Assuming it uses an APD, the receiver device converts light pulses reflected from the car ahead to current pulses, which a transimpedance amplifier converts to voltage pulses. This requires another design decision. A suitable transimpedance amplifier should have high gain, high input impedance, ultra-low voltage and current noise, and low input capacitance.
In a typical design, the transimpedance amplifier’s voltage output is further amplified and may undergo further signal conditioning before the ADC digitizes it. To be effective across a range of inter-vehicle distances, the analog front-end (AFE) circuitry requires at least 100 dB of dynamic range, which implies some kind of variable gain amplifier (VGA) as the last analog stage. Another design decision concerns whether the ADC input is differential or single-ended, as this implies variability back up the signal chain in terms of how the signal is handled.
The accuracy of TOF measurements, the factor at the root of the distance between vehicle calculations, depends on both the pulse width of the laser blip and the speed and accuracy of the ADC. In terms of sample rate, given c, the speed of light, the minimum sample rate is simply c divided by the required resolution.
For automotive lidar, the accuracy requirement for distance measurements is roughly ±3 feet. With that assumption, the inter-vehicle distance measured must consider a round trip for the laser blip, meaning the needed measurement resolution is twice 3 feet, or 6 feet. Using 3 × 108 meters per second (9.84 × 108 feet/second) for c, the minimum ADC sample rate must be (9.84 × 108)/6, or 163.9 Msamples/s, meaning the sample interval is on the order of 6.1 ns.
For this kind of application, there are some interesting features in Texas Instruments’ 200-Msample/s ADC08B200A ADC, including a 1-kbyte on-chip buffer. It also has an on-chip clock multiplier so 200 Msamples/s can be obtained with an external clock rate as low as 25 MHz.
Incorporating the buffer within the ADC means that the FPGA or DSP can be smaller than it would have to be if it included the buffer. It also reduces the bill of materials by one standalone FIFO.
Integrating Front End and Data Converter
Not surprisingly, TI offers an integrated solution for the lidar aspect of adaptive speed control, the ADC08B200. In addition to the ADC, the ADC08B200 has a clock multiplier that allows the use of a low-frequency clock oscillator. The clock multiplier can multiply the input clock frequency by 1, 2, 4, or 8. This permits the use of a clock source as low as 25 MHz to obtain 200-Msample/s operation. The external clock source could even be the same as the clock used for the FPGA/DSP/microcontroller.
Also, the ADC’s 1-kbyte buffer relaxes the speed and complexity requirements of the FPGA, DSP, or microcontroller. The buffer can be read at any desired rate up to 200 MHz. It can be bypassed as well, in which case the data is continuously streamed out at the ADC sample rate. The converter function can be powered down while the buffer is being read to save power.
Analog Devices provides several different types of microelectromechanical-systems (MEMS) devices for vehicle safety. Its efforts began several decades ago with accelerometers for airbag systems. Today, its MEMS gyros are used to prevent certain types of crashes by modulating brake pressure when the vehicle is in danger of rolling over. If the gyro senses that the car is spinning out of control, differential braking is engaged to bring it back onto the straight-and-level.
Such gyro systems also are used in dead-reckoning functions in GPS systems that continue to provide positioning information when satellite signals are temporarily lost.
According to Analog Devices, the electronics and mechanical structures of the gyros are integrated on the same substrate. The system comprises a mechanical sensor structure and two sets of electronics. One of these sets drives a vibration in a resonator structure that undergoes a Coriolis effect. The other detects Coriolis force-induced displacements in a capacitor structure.
In more detail, in such a MEMS gyro, a frame containing a resonating mass is tethered to the substrate by springs. The mass is driven at its resonant frequency in one direction (up and down in this case), and the springs are mounted at 90° relative to the resonating motion.
In a vehicle, when the resonating mass moves away from the center of rotation, the Coriolis effect accelerates it to the right. This exerts a leftward reaction force on the frame. When it moves toward the center of the rotation, there is a corresponding force to the right. That movement changes the distance between interdigitated sense fingers in the capacitor structure. The change in capacitance is read as an angular rate of change. (Fig. 2)
The capacitor is a planar structure that consists of a number of interwoven fingers. An electromechanical oscillator drives the mechanical structure at its resonant point—about 15 kHz.
The challenge in designing the detection electronics is exquisite. Displacements in the actual resonator can be on the order of 10 µm, but the physical displacement in the capacitor is on the order of only 1 Angstrom—about the dimension of a hydrogen atom. Electronically, this results in an actual capacitance change of about 90 aF (attoFarads). But noise means the electronics have to deal with a capacitive change of about 12 zF (zeptoFarads) based on about 16 Fermis (16 × 10-15 m) of movement.1
Although many readers will be familiar with Coriolis effects in spinning gyroscopes and the Foucault pendulums found in the lobbies of almost every science museum, it’s something of a hand-wave to describe these devices by saying, “It’s just like a tiny Foucault pendulum.” Analog Devices explains the process in detail in Chapter 3 of its Linear Circuit Design Handbook.2
Gyroscopes are used to measure angular rate—how quickly an object turns. A combination of three gyros at right angles in space provides information about angular motion in three dimensions: yaw, pitch, and roll.
MEMS gyro sensitivity is measured in units of millivolts per degree per second (mV/°/s). The Analog Devices analog ADXRS300, for example, is rated at of 5-mV/°/s sensitivity and outputs 1.5 V for a full-scale input of 300°/s. The company’s more sophisticated products may integrate three axes and provide self-calibrated digital output.
Changes in capacitance are tiny, implying that noise is a significant challenge. Situating the electronics, including amplifiers and filters, on the same die as the mechanical sensor is essential to allow the differential signal, which alternates at the resonator frequency, to be extracted from the noise by correlation.
There are two sources of noise: random atomic vibration and impacts from air molecules. Surprisingly, it turns out to be better to leave the structure open to outside air, even at the cost of additional random noise because the air helps cushion the internal structure from shock.
Some Analog Devices gyros use a trick that makes it possible to reject shocks of up to 1000 g. They use two resonators to differentially sense signals and reject common-mode external accelerations that are unrelated to angular motion.
The two resonators are mechanically independent and driven 180° out of phase. As a result, they measure the same magnitude of rotation, but their outputs are in opposite sense. This cancels non-rotational signals that affect both sensors and makes it possible to use the difference between the two signals to measure angular rate.
Power-Bus Transient Protection
Adaptive speed control and rollover protection are safety issues. Chipmakers’ innovation in automotive products extends to more basic systems, though, like the power buses that run those occupant-protection electronics along with less critical systems such as infotainment and navigation.
One simple problem that requires not-so-simple design engineering is protecting the devices on the power buses from voltage transients. Linear Technology’s LTC6802 and LTC6803 monitor the characteristics of individual batteries in battery stacks in hybrids and electric vehicles (EVs), which provide hundreds of volts for the vehicles’ traction motors (Fig. 3) (see “Automotive Applications Benefit From Advanced ICs" and "High-Voltage Electric/Hybrid Vehicle Battery Monitor Chip Evolves”).
Battery stacks in cars are a relatively new development, but there’s an old problem that recently got a pair of new solutions, courtesy of Linear CTO Robert Dobkin. The new ICs provide a way to deal with big power-bus transients with a device that has a very small footprint.
Ever since the dawn of the semiconductor age, car makers have had to be concerned about voltage spikes. As vehicles have added fuel-injection control, airbags, satellite radio, GPS, ABS, and traction control, these concerns have only grown. Now that autonomous electronics are on the verge of taking over lane control, pedestrian avoidance, and relative position-keeping, as the vehicle hurtles down the highway, maintaining a clean power bus is more imperative than ever.
Prevention would be the better cure, but that’s more a hope that an achievable reality. Inevitably, when power flows over long cable runs, transients occur whenever there are load steps. Sudden, negative-going current changes on long wire runs result in positive-going voltage spikes. But the most dreaded event is the “load dump,” associated with corroded battery terminals.
Load dumps have been known to create voltage surges of as much as 125 V that remain elevated for hundreds of milliseconds. The Society of Automotive Engineers (SAE) characteristic load dump profile includes a 5-mA rise time of 5 ms and an exponentially decaying fall time with a 200-ms time constant.
Historically, engineers have attacked the problem with capacitors, transient-voltage suppression (TVS) diodes, and even fuses. But as the density of electronic devices within the vehicle has grown, the space available to install discretes as protection devices has shrunk.
Linear Technology already had an active device for dealing with the overvoltage problem. As Dobkin and his top designers spoke with the power engineers at automakers, though, it became clear that something better was needed. Released in 2007, the company’s LT4356 Surge Stopper operates from 4 to 80 V and protects against reverse-voltage transients on its input pins to magnitudes of –60 V.
All of the LT43xx devices are linear regulators for automotive use, intended to feed external devices like dc-dc converters. They clamp voltage surges to a value set by an external resistor divider, while remaining unaffected by the surge. Linear’s original LT4356 could suppress surges greater than 100 V. Its Achilles heel was the positioning of the current-sense resistor—upstream of the external pass transistor. (Fig. 4a) Over-current protection, then, had to be disabled if the device was required to protect against transients greater than 100 V.
The company realized the solution in two new products: the LT4363 high-voltage surge stopper with over-current protection, and the LTC4366 high-voltage floating surge stopper.
For designers who were comfortable with the LT4356, the LT4363 simply moves over-current sensing downstream of the pass FET, enabling over-current protection while withstanding large voltage transients (Fig. 4b). But like the LT4356, its absolute maximum rating is 100 V. So, the input must be protected from high-voltage transients greater than 100 V with a resistor and TVS diode.
The LT4366 is more interesting (Fig. 5). It shows what happens when an engineer like Dobkin revisits an old idea. (In fact, the basic evolution of the part took place on a series of cocktail napkins on a long airline flight.) “What if we allowed the part to float?” Dobkin asked himself. It became apparent that a floating part could be used in systems that operate continuously at voltages above 100 V and require protection against voltage transients even greater than 200 V.
In a typical application, external voltage-dropping resistors allow the part to float up with the supply, isolating it from any high-voltage surges. Thus, the upper limit on the operating voltage is limited only to the availability of the necessary high-value resistors and a MOSFET that can handle the power dissipated during voltage regulation. Because it isn’t tied to ground, the LTC4366 can handle much larger transients than its non-floating relatives—with some additional functionality.
The LTC4366 operates in three different modes: start, run, and regulate.
In start mode, a 15-μA trickle current flows through RIN. Half of this charges the gate, and the other half is used for bias current. As the GATE pin charges, the external MOSFET brings up the OUT pin until the part enters run mode. At that point, the output is high enough to operate a charge pump. The charge pump is then used to fully charge the gate 12 V above the source voltage.
While it is in regulate mode, to protect the load from an input-supply overvoltage, an overvoltage regulation amplifier is referenced to the output through a 1.23-V voltage reference. If the drop across the upper feedback resistor exceeds 1.23 V, the regulation amplifier then pulls down the gate to regulate the drop back to 1.23 V.
Why did Linear create two new parts? It might seem that the LTC4366 would be a fitting companion for the LT4356. There could be one downside to using the LTC4366—cold crank, which is a problem with the starter motors for big, internal-combustion engines in conventional cars. Under very cold conditions, the voltage drop associated with starting the engine can pull the battery bus down to as little as 4 V. In this case, it isn’t the time constant of the transient that is the problem, but its duration.
Specs for the LT4356 and LT4363 include a guaranteed 4- to 80-V operating range. Where cold crank is an issue in new designs, the LT4363 has advantages. But remember that this is the automotive IC business where the payback for low margins is long product life cycles. Linear will be shipping LT4356 devices for old design-ins for many years yet.
On the other hand, the operating range of the LTC5366 extends from greater than 500 V down to 9 V. A slightly higher bottom end is probably outweighed by the much higher top end in hybrids and EVs. Even in conventional cars and trucks, the impact of the higher cold-crank spec depends on what’s being powered. It might not be suitable for engine-control units and the like, but it would not be a problem on power buses for applications such as infotainment and GPS.
- For more, see “Single-chip Surface-Micromachined Integrated Gyroscope with 50°/Hour Root Allan Variance” at http://ieeexplore.ieee.org/Xplore/login.jsp?url=http%3A%2F%2Fieeexplore.ieee.org%2Fiel5%2F7787%2F21395%2F00992288.pdf%3Farnumber%3D992288&authDecision=-203.
- Analog Devices, Chapter 3, Linear Circuit Design Handbook, www.analog.com/library/analogDialogue/archives/43-09/EDCh%203%20sensors.pdf.