Current sensing has long been an important function implemented by battery management systems (BMS), modules which monitor and protect high-capacity batteries. In both lithium-ion and sealed lead-acid battery types, current measurements are used to protect the battery against abuse and ensure its safe use by providing for emergency shut-down in over-current conditions. For protection and safety functions alone, the accuracy of the current measurements can be at a fairly low level. The system designer may specify the over-current conditions conservatively, so that even if the current sensor severely underestimates the current, the safe shut-down threshold is not crossed.
Now, however, the requirements for current sensing are becoming much more stringent in certain applications. Car manufacturers in particular are working furiously to improve the performance and consumer appeal of electric vehicles (EVs). Range anxiety is one of the biggest impediments to consumer adoption of EVs, and so the accuracy of an EV’s “fuel gauge”—that is the State of Charge (SOC) reading showing how much energy is available for use—is of critical importance to the driver. Accurate SOC measurements also enable the BMS to optimize operation for long cycle life, in EVs and in industrial equipment, by maintaining the SOC at between 0% and 80%.
The accuracy of the fuel gauge depends absolutely on the accuracy of the BMS’s current measurements. And as this article will show, precision analogue circuitry and an appropriate architecture can provide much higher levels of accuracy than are commonly achieved in today’s BMS.
The Use of Current Measurement in li-ion BMS
Today, the two common types of lithium-ion batteries in high cycle-life applications (that is, applications in which the battery is charged and discharged many times) are lithium iron phosphate (LFP) and lithium-titanate (LTO). In these batteries, as well as in other high energy-density li-ion batteries, accurate charge integration over time is considered to be the most important method of measuring SOC.
This is because, in LFP and LTO battery types, the output voltage stays remarkably constant over most of the discharge cycle. This means that open-circuit voltage (OCV) measurement, when the battery is 'at rest', is a poor indicator of SOC. What is more, some applications for LFP and LTO batteries maintain the battery in constant use, either charging or discharging at an intermediate SOC (between fully charged and fully discharged). This is a typical use case, for instance, for renewable energy equipment such as wind turbines and solar panel arrays, and in this type of equipment, OCV conditions hardly ever occur anyway.
Charge integration is the process of counting coulombs out (as the battery is discharged), and counting them back in (as it is charged). Starting from a known charge capacity and from the fully charged state, such coulomb counting can provide a measurement of SOC over a theoretically infinite number of charge/discharge cycles.
But clearly its accuracy is wholly dependent on the accuracy of the current measurement circuit, over the entire positive and negative signal swing, and with precise detection of the zero current point.
Maintaining high accuracy at intermediate and low currents is of particular importance, to cater for the typical operating profile of an EV or of an energy buffer in power generation equipment. This new requirement for ultra-high accuracy in all conditions is beyond the capabilities of the current-sense circuits that were adequate for earlier generations of BMS.
In automotive applications today, one of two techniques for current sensing is normally adopted.
- Shunt interface solutions measure the voltage drop across a precision shunt resistor. Typically, however, these interfaces are unable to offer a perfect zero offset, leading to inaccuracies at low and medium current levels, and hampering the correct detection of the zero current point.
- Magnetic current sensors also feature a troublesome offset, and also offset drift over temperature, hysteresis, and sensitivity to interference from external magnetic fields. This combination of drawbacks makes the magnetic current sensor unsuitable for serious consideration in EV BMS and other applications requiring high accuracy.
Extremely elaborate, complicated and application-specific software compensation would be required, then, to correct the offset of conventional shunt interface solutions.
Far easier, then, to use a shunt interface which is perfectly offset free over the automotive temperature range of −40°C to +125°C. (That is, for a ±5-bit resolution sensor interface the LSB is markedly less than 1, giving an essentially zero-LSB offset.) This was the starting point for ams in the development of its AS8510 shunt current-sensing interface IC. When paired with an offset-free shunt resistor, this sensor interface is able to offer an almost perfectly linear output over the entire operating temperature range and over the full signal range above and below ground. It supports a fast sampling rate of up to 8 kHz.
Using such a device, the developer of a BMS for LFP or LTO batteries is able to achieve high accuracy far more easily, because the only compensation required is for the drift of the shunt, gain stage and ADC reference over temperature and life. Typically these give rise to a combined error of less than 0.5%, so in many applications no compensation may be required. As for any electronic circuit with a digital output, around the zero-current point only, a trivial amount of error will also be attributable to the limits of the ADC’s resolution, and to noise.
Multiple Functional-Safety Features
As well as accurate current measurements, a current sensor interface in automotive applications should also support the requirement for compliance with the ISO26262 functional safety standard.
The AS8510 is a two-channel data-acquisition device. Its two identical channels each have a chopper, programmable amplifier, integrating-type 16-bit sigma-delta ADC, a de-chopper, and filter options independently configurable for each channel. The digital representation of the device’s current measurements are transmitted through a serial peripheral interface to an external microcontroller. By adding an external chopper and level-shifter components, the AS8510 can also be made to measure currents on the positive power rail (the battery’s high side) with the same zero offset.
The redundancy inherent in the twin signal-conditioning channels supports automotive functional safety requirements. Other features also help BMS manufacturers to meet the obligations of ISO26262 compliance, such as the programmable current source freely connectable to the input pins (facilitating fault testing), external access to the ADC’s reference voltage, and independent configuration of each channel.
How the System Designer Can Optimize for the Highest Accuracy
As was described above, the main components of the error in an AS8510-based current measurement system are the temperature and lifetime drift of the shunt resistor, gain stage, and reference voltage.
In fact, the temperature dependency of the shunt and the ADC reference voltage causes marked deviations from nominal values only at temperatures below −20°C and above 65°C. Between these temperatures, the temperature drift of the reference and gain stage are fully specified by ams, and typical implementations using the AS8510 are able to achieve system accuracy of 0.8% over the full temperature range when subject to a single-point end-of-line calibration.
If required, though, how can the designer achieve even tighter accuracy?
Fortunately, the AS8510 includes an internal temperature sensor, and because it operates at very low power, internal self-heating is negligible. This means that the integrated temperature sensor can provide readings for use in end-of-line compensation, both of the internal signal path and also external devices such as the shunt, provided they are thermally coupled to the AS8510.
The biggest sources of error in the sensor interface are the gain stage’s drift over temperature, and the ADC reference’s drift over temperature. These errors can therefore be eliminated if the temperature drift characteristics are known.
These characteristics can be determined by calibration. Applying a reference current at various temperatures, the user can cancel the effect of drift in software, drawing on the internal temperature sensor’s readings (that is, curvature compensation of the drift characteristic as a function of temperature).
Alternatively, the ADC reference has a generic curvature with a known polynomial function, described in the AS8510’s data sheet. This error source can also be eliminated by software with the help of the internal temperature sensor’s readings.
With multi-point calibration, shunt sensor systems can achieve extremely high accuracy: With temperature drift eliminated, the error is largely determined by the shunt’s drift over lifetime, which is typically around 0.2%.
Interestingly, the availability of an internal temperature sensor also enables the use of metal tracks in a PCB to replace the shunt, since the large temperature coefficient of the metal (copper and aluminium are rated at around 4,000ppm/K) may be compensated with the help of the temperature sensor’s readings. Hardware evaluations have shown that the thermal conduction from the metal track to the IC is good enough to support calibrations that achieve system accuracy of around 2% even for large and pulsed currents, and even better accuracy for small currents.
Shunt resistors are intrinsically linear and offset-free. By combining a shunt in the 50-100 μΩ range with a highly linear, offset-free, and high-resolution signal-conditioning device, current sensor systems can be designed which are able to provide high accuracy when measuring signals from mA up to kA virtually without energy losses. Most remarkably, such a system can offer high accuracy over the entire signal range down to zero current because of its true zero offset and high linearity.