Hybrid-electric vehicles must operate reliably under harsh conditions that demand special qualities on the part of power-conditioning electronics.

Consider the electrical system for a hybrid-electric vehicle (HEV). Because it powers traction motors, the electrical system on hybrids can easily reach several hundred volts and handle power levels into the kilowatt range or higher. Hybrids typically use dc-dc converters to boost the voltage from a high-voltage battery high enough to power the traction motor and also for reducing voltages to around 14 V for ordinary vehicular electronics.

Authored by:
Zhang Bin, Harold Tisbe
Isolation Products Div.
Avago Technologies
San Jose, Calif.
Edited by Leland Teschler
leland.teschler@penton.com
Key points:
• Inside hybrid-electric vehicles, optocouplers are used to insulate high voltages and isolate unwanted signals.
• High CMR performance and isolation voltage are highly desirable in the inverters that power vehicular electric motors.
• Worldwide regulatory standards such as those of the IEC, UL, CSA, and EN now dictate electrical system performance in hybrid vehicles.
Resources:
Avago Technologies, www.avagotech.com
ECE Regulation No. 100, 1995, www.unece.org
ISO 6469-3 Electric road vehicles –Safety specifications – Part3: Protection of persons against electric hazards, 1st Edition 2001, www.iso.org
Paul Gray, Robert Meyer, “Analysis and Design of Analog Integrated Circuits”, www.amazon.com/Analysis-Design-Analog-Integrated-Circuits/dp/0471321680
UL 840 Insulation Coordination Including Clearances and Creepage Distances for Electrical Equipment, www.ul.com
UL 2202 Standard for Electric Vehicle (EV) Charging System Equipment, 1st Edition, 1998, www.ul.com

In HEVs, both ac-dc and dc-dc converters are key building blocks. For example, Prius models made after 2003 have a main battery voltage of about 200 V. Dc-dc converters step down the main battery voltage to 12 V for charging the auxiliary battery and for powering auxiliary circuits. There are additional step-down conversions to 5, 15, and 24 V for various other operations.

On plug-in hybrids, ac-dc converters take power from the utility grid for battery charging. These chargers are typically high-power converters ranging from a few hundred watts up to 2 kW with output voltage between 48 and 300 V. Because severe high voltages are present, it is imperative to add safety insulation against shock hazards. Designers must build in galvanic isolation between the primary and secondary of both ac-dc and dc-dc converters because of the presence of hazardous high voltage (above 25 Vac or 60 Vdc). The standards and regulations for HEV are still being worked out, but there are already a handful of international and national standards that apply. They include the Federal Motor Vehicle Safety Standards (FMVSS) FMVSS 305 standard which applies for vehicles that use more than 48 V for propulsion power. It is enforced by the National Highway Traffic Safety Administration (NHRSA) of the U. S. Dept. of Transportation. It’s designed to avoid fatalities and injuries during a crash caused by electrolyte spillage from propulsion batteries, intrusion of propulsion battery-system components into the occupant compartment, and electric shock. The requirement states the isolation barrier between battery and exposed conductive part should maintain 500 Ω/V before and after the crash impact.

ECE-R 100 is another regulation set forth by the United Nations Economic Commission of Europe (UNECE) and is widely adopted in the EU. Its scope covers electric road vehicles with maximum design speed exceeding 25 km/hr. It sets isolation resistance between any exposed conductive part and each polarity of the battery at 500 Ω/V minimum under normal operation and post impact.

ISO 6469-3 and IEC 61851-21 are safety standards for protection against electrical shock. The IEC document covers electrical devices and equipment and the ISO regulation deals with all other technologies. Hybrid-electrical vehicles, having both combustion engines and electrical motors, come under both standards. According to ISO 6469-3 insulation resistance should exceed 5k Ω/V for double/reinforced insulation. In addition, insulation should withstand a dielectric breakdown test at two times the constantly increasing working voltage (3,250 Vrms or 3,750 kVrms) for 1 min, at a frequency between 50 and 60 Hz.

Underwriters Laboratories (UL) standard UL 2202 and SAE International (SEA) standard SAE J2344 are national U. S. standards that apply. They are also widely followed by many foreign automotive manufacturers that deal with the U. S market. UL 2202 has detailed specifications on insulation between primary and secondary circuits; hence it is often followed by electrical designers for chargers and dc-dc converters in electrical vehicles.
UL 2202 is the standard for both onboard and off-board charging systems connected to a main voltage of 600 Vac or less for recharging storage batteries in over-the-road electric vehicles. Insulation spacing (that is, through-air (clearance) and oversurface (creepage) are in section 21 of the standard. However, the standard also permits alternative spacing in accordance to UL 840, “Insulation Coordination Including Clearances and Creepage Distances for Electrical Equipment.” UL 2202 has a single-spacing specification for the intended voltage range, regardless micro environment, while UL 840 allows reduced spacing under certain circumstances. As a result, the spacing required by UL 840 is smaller than that of UL 2202.

It is interesting to note that, for clearance, UL 2202 meets UL 840’s requirement for overvoltage Cat IV, and doubles the spacing requirement for Cat III. Similarly, the creepage distance of UL 2202 is about the same as that of UL 840 for pollution degree 4, and is twice the distance for pollution degree 3. Essentially insulation spacing provided by UL 2202 has double the insulation requirement of UL 840 for overvoltage category III, pollution degree 3 with material group IIIa/b, a condition that satisfies most of the applications for onboard and off-board chargers.

In isolated ac-dc and dc-dc converters, optocouplers often provide error amplifier feedback because they are inexpensive, are relatively small, and can transmit dc signals. Transformers can provide the same sort of isolation; however, they need complex signal conditioning circuits to transmit dc signals.
For closed-loop control, the system-loop transfer function should meet the Nyquist criteria for stability. Where loop gain rolls off to 0 dB at the cut-off frequency, fc, phase delay at fc should be less than 45° to provide enough phase margin and a “well-behaved” dynamic response.

For a sampling system the closed-loop bandwidth, which is also the loop cut-off frequency, should be bounded and not exceed fs/(2πD), where D = the maximum duty cycle, and fs = the pulse-width-modulation (PWM) switching frequency. At fc =
fs/(2πD), error amplifier gain may be high enough to cause the amplified output ripple voltage to drive the error amplifier into saturation, necessitating a further reduction in fc. In practice, it’s possible to see a cut-off frequency up to fs/4 but much lower values are common, often because of the bandwidth limitation of both error amplifiers and phototransistor optocouplers.

HEV applications often use a high PWM frequency to keep down the size and weight of power transformer. A typical PWM switching frequency of 200 kHz would use a cut-off frequency between 10 and 50 kHz.
Closing the loop requires a good understanding of the loop-transfer function, which includes the signal-transfer path from Vo to iLED, iLED to Vc, and Vc to Vo. The transfer function from Vc to Vo is determined by the PWM control methodology and circuit topology, which literature on the topic explains well. The transfer function from ILED to Vc must be analyzed when designing a compensation network that lets the loop meet the Nyquist criteria in most unfavorable operating conditions.

Feedback circuits often use a common emitter topology partly because its inverting logic is necessary for the start-up of the PWM controller. The small signal model of this circuit contains an LED (light-emitting-diode) pole related to LED input capacitance, approximately 100 to 200 pf including both depletion and diffusion capacitance. Small signal models generally neglect this capacitance. At relatively low bias current of 1 mA (vT/ILED), the small-signal input resistance of the LED is on the order of 25 Ω, resulting in a pole at about 60 MHz which, therefore, is not a dominant pole. Neither the output impedance of the error amplifier nor the external driving resistor could affect the pole because the LED input resistance appears in parallel with its capacitance.

Based on the small signal model the ILED to Vc transfer function can be written as:

where α = ib/iLED, photodiode responsivity; ib = photodiode current driving into base, A; cpd = photodiode junction capacitance, f; gm = transistor transconductance; cu = collector to base capacitance, f; rπ = transistor base small-signal input resistance, Ω; RL = optocoupler pull-up resistance, Ω; and ic = collector current, A.
Given
                       β = gmrπ   (2)
CTR = ic/iled = (ic/ib)(ib/iled)=βα=αgmrπ  (3)
we have

The transfer function (4) has dc transimpedance gain, CTR × RL, one zero at a much higher frequency than the frequency of interest and, thus, is ignored, and two poles. Assuming two poles in (4) are widely separated we have:
p1 = - 1/(rπ(cu(gmRL + 1 +RL/rπ) + cpd))  (5)
p2 = - (1/(RLcu ) + 1/(rπcpd)+1/(RLcpd)+gm/cpd)  (6)

Equation (5) says that p1, the dominant pole of ACPL-M43T, is inversely proportional to the small-signal base-emitter junction resistance rπ as well as the sum of Miller capacitance, cu(gmRL + 1 + RL/rπ), and photodiode capacitance cpd. Because of the large area of the photodiode, cpd is on the same order of magnitude as Miller capacitance.

The base-emitter junction resistance rπ decreases with rising collector current according to the characteristic equation  (7).

rπ = vT/ib = βvT/ic  (7) where vT = the thermal voltage, V.


Base-collector junction capacitance, cu, rises with decreasing collector voltage, vc, due to the reduced biasing voltage across the collector and base junction. As a result, cu rises with increasing collector current in the common emitter configuration. The compounded effect of collector current on the dominant pole, p1, is evident in the accompanying figure, where RL = 1 kΩ, Vcc = 5 V, and Ta = 25˚C. The bandwidth reduction at low-bias current is because of the reduction of rπ. At high biasing current, the increase of cu dominates because of the lower base-collector bias voltage (~ 0.2 V at 4.5 mA for RL = 1 kΩ, Vcc = 5 V).

A plot of bandwidth across temperature indicates the worst-case condition happens at high temperatures with a typical bandwidth of 600 kHz. At a high temperature, there is a rise in both current gain, β, and thermal voltage, vT, in equation (7), resulting in higher rπ and reduced bandwidth. With a β variation up to ± 30%, the resulting worst-case bandwidth exceeds 400 kHz, which is still a decade away from the frequency of interest. Hence, the optocoupler pole can be ignored for loop compensation.

In contrast, a four-pin phototransistor found in many consumer power supplies is much slower because of the large photodiode capacitance between base and collector, multiplied by large voltage gain as Miller capacitance. As a result, the typical bandwidth of a phototransistor is on the order of 10 kHz. Significant phase delay is introduced at the frequency of interest and zero compensation must be designed to cancel the phototransistor pole to get a larger bandwidth and good dynamic performance.

However, the pole location changes with temperature, biasing current, and depending on the part. It is difficult for a zero implemented by a passive component to track the pole position at all conditions. There must be a compromise in loop compensation to ensure good phase margin, resulting in much lower closed-loop bandwidth and poorer dynamic performance.

CTR across temperature
As indicated in equation (4), CTR directly affects dc-loop gain and cut-off frequency. The worst-case phase margin is at high CTR where a second loop pole, p2_loop, is getting close to the cut-off frequency, fc_high. When CTR is low, both dc gain and loop bandwidth drop, such that the static and dynamic-loop regulation deteriorates. It is important for CTR variations to be small and well bounded so power converters behave properly at worst-case conditions.

For optocouplers, CTR is the product of light output (LOP) generated from LED photodiode current caused by incident light and the β of the transistor. Although LOP and β vary from part to part, optocouplers such as the ACPL-M43T can hit a max-to-min CTR span ratio of less than two, which is well within 1 CTR bin of a typical phototransistor. When temperature rises, the LOP of the LED drops but the transistor β rises. This action of the LOP and β helps to narrow the CTR variation across temperature. The temperature coefficient of β dominates at cold temperatures, reducing CTR. At high temperatures, the LOP coefficient dominates, which results in CTR bending lower as well.

Another common concern for optocouplers is the degradation of CTR over the life of the part as seen in many phototransistor-type optocouplers. However, advances in LED fabrication and assembly have resulted in parts which exhibit less than 10% CTR drift after 5,000 hr of accelerated life time stress under 150˚C, and 20 mA of LED driving current. The ACPL-M43T is an example of such advances.

Phototransistors for many consumer power supplies have tight binning at room temperature. But manufacturers typically don’t characterize their variations across temperature, biasing current, and lifetime, so these factors are unpredictable. While this uncertainty is acceptable for such consumer applications as phone chargers and computer power supplies, it should be a concern for automotive applications where deteriorating performance in electrical vehicles could potentially cause system malfunctions and accidents.

© 2010 Penton Media, Inc.