Testing Lead Acid Batteries

Many manufacturers of battery testers claim to measure battery health on the fly. These instruments work well in finding battery defects that involve voltage anomalies and elevated internal resistance, but other performance criteria remain unknown. Stating that a battery tester based on internal resistance can also measure capacity is misleading. Advertising features that are outside the equipment’s capabilities confuses the industry into believing that multifaceted results are attainable with basic methods. Manufacturers of these instruments are aware of the complexity involved, but some like to add a flair of mystery in their marketing scheme, similar to a maker of a shampoo product promising to grow lush hair on a man’s bald head. Here is a brief history of battery testers for lead acid and what they can do.
The carbon pile, introduced in the 1980s, applies a DC load of short duration to a starter battery, simulating cranking. The voltage drop and recovery time provide a rough indication of battery health. The test works reasonably well and offers evidence that power is present. A major advantage is the ability to detect batteries that have failed due to a shorted cell (low specific gravity in a cell due to high self-discharge). Capacity estimation, however, is not possible, and a battery that has a low state-of-charge appears as weak. A skilled mechanic can, however, detect a faulty battery based on the voltage signature and loading behavior. To do a CCA pass/fail test, load a fully charged starter battery with half the rated CCA value for 15 seconds. To pass, the voltage must stay above 9.6V at 10º C (50º F) and higher. Colder temperatures cause a large voltage drop.
The AC conductance meters appeared in 1992 and were hailed as a breakthrough. The non-invasive method injects an AC signal into the battery to measure the internal resistance. Today, these testers are commonly used to check the CCA of starter batteries and verify resistance change in stationary batteries. While small and easier to use, AC conductance cannot read capacity, and the resistive value gives only an approximation of the real CCA of a starter battery. A shorted cell could pass as good because in such a battery the overall conductivity and terminal voltage are close to normal, even though the battery cannot crank the motor. AC conductance testers are common in North America; Europe prefers the DC load method.
Critical progress has been made towards electrochemical impedance spectroscopy (EIS). Cadex took the EIS technology a step further and developed battery specific models that are able to estimate the health of a lead acid battery. Multi-model electrochemical impedance spectroscopy, or Spectroäfor short, reads battery capacity, CCA and state-of-charge in a single, non-invasive test.Figure 1 illustrates the Spectro CA-12 handheld battery tester.

Figure 1: Spectro CA-12 battery tester Compact battery rapid tester displays capacity, CCA and state-of-charge in 15 seconds. Courtesy Cadex

The Spectro CA-12 handheld device, in which the Spectro™ technology is embedded, excites the battery with signals from 20–2000Hz. A DSP deciphers the 40 million transactions churned out during the 15-second test into readable results. To check a battery, the user simply selects the battery voltage, Ah and designated matrix. Tests can be done under a steady load of up to 30A and a partial charge, however, if the state-of-charge is less than 40 percent, the instrument advises the user to charge and retest.
The Spectro method is a further development of EIS, a technology that had been around for several decades. What’s new is the use of multi models and faster process times. Cost and size have also shrunk. Earlier models cost tens of thousands of dollars and traveled on wheels. The heart of Spectro is not so much the mechanics but the algorithm. No longer do modern EIS devices accompany a team of scientist to decipher tons of data. Experts predict that the battery industry is moving towards the multi-model EIS technology to estimate batter performance
Nowhere is the ability to read capacity more meaningful than with deep-cycle batteries in golf cars, aerial work platforms and wheelchairs, as well as military and naval carriers. Getting a readout in seconds without putting the vehicles out of commission allows for a quick performance check on a suspect battery before deployment in the field. Figures 2, 3 and 4 show typical battery problems and how modern test technologies can detect them.

	Figure 2: Low charge Drive is sluggish; Spectroäreads low SoC. Capacity estimation is correct in spite of low charge.
	Figure 3: Low capacity Battery has good drive but short runtimes. Spectroäreads good impedance but low capacity.
	Figure 4: Faulty set Spectroäfinds low performing and shorted blocks in a string. Good batteries can be regrouped and reused. All figures Courtesy of Cadex

Matrices

Measurement devices, such as the Spectro CA-12, are not universal instruments capable of estimating the capacity of any battery that may come along; they require battery specific matrices, also known as pattern recognition algorithm. A matrix is a multi dimensional lookup table against which the measured readings are compared. Text recognition, fingerprint identification and visual imaging operate on a similar principle in that a model exists, with which to equate the derived readings.
This book identifies three commonly used measuring methods. The principle in all is to take one or several sets of readings and compare them against known reference settings or images to disclose the characteristics of a battery. The three methods are as follows.

Scalar:     The single value scalar test takes a reading and compares the result with a stored reference value. In battery testing this could be measuring a voltage, interrogating the battery by applying discharge pulses or injecting a frequency and then comparing the derived result against a single reference point. This is the simplest test, and most DC load and single-frequency AC conductance testers use this method.

Vector:    The vector method applies pulses of different currents, or excites the battery with several frequencies, and evaluates the results against preset vector points to study the battery under various stress conditions. Typical applications for this one-dimensional scalar model are battery testers that apply multi-tier DC loads or inject several test frequencies. Because of added complexity in evaluating the different data points and limited benefits, the vector method is seldom used.

Matrix:    The matrix method scans a battery with a frequency spectrum as if to capture the image of a landscape and compare the imprint with a stored model of known characteristics. This multi-dimensional set of scalars, which form the foundation of Spectroä, provides the most in-depth information but is complex in terms of evaluating the data generated. With a proprietary algorithm, the Spectroätechnology is able to estimate battery capacity, CCA and SoC.

Matrices are primarily used to estimate battery capacity, however, CCA and state-of-charge also require matrices. These are easier to assemble and serve a broad range of starter batteries. While the Spectroämethod offers an accuracy of 80 to 90 percent on capacity, CCA is 95 percent exact. This compares to 60 to 70 percent with battery testers using the scalar method. Service personnel are often unaware of the low accuracy; verifications are seldom done, as this would involve several days of laboratory testing.
A further drawback of scalar battery testers is obtaining a reading that is neither resistance nor CCA. While there are similarities between the two, no standard exists and each instrument gives different values. In terms of assessing a dying battery, however, this method is adequate as it reflects conductivity. The larger disadvantage is not being able to read capacity. Table 5 illustrates test accuracies using scalar, vector and matrix methods.

Measuring units	Scalar Single value	Vector One-dimensional set of scalars	Matrix Multi-dimensional set of scalars
CCA	60–70% accurate		90–95% accurate
Capacity	N/A		80–90% accurate
SoC	Voltage-based; requires rest after charge and discharge		90–95% accurate (with new battery)

Table 5: Accuracy in battery readings with different measuring methodsScalar and vector provide resistance with references to CCA on starter batteries. The matrix method is more accurate and provides capacity estimations but needs reference matrices.
To generate a matrix, batteries with different state-of-health are scanned. The more batteries of the same model but diverse capacity mix are included in the mix, the stronger the matrix will become. If, for example, the matrix consists only of two batteries, one showing a capacity of 60 percent and the other 100 percent, then the accuracy would be low for the batteries in between. Adding a third battery with an 80 percent capacity will solidify the matrix, similar to placing a pillar at the center of a bridge. To cover the full spectrum, a well-developed matrix should include battery samples capturing capacities of 50, 60, 70, 80, 90 and 100 percent. Batteries much below 50 percent are less important because they constitute a fail.
It is difficult to obtain aged batteries, especially with newer models. Forced aging by cycling in an environmental chamber is of some help; however, age-related stresses from the field are not represented accurately. It also helps to include batteries from different regions to represent unique environmental user patterns. A starter battery in a Las Vegas taxi has different strains than that of a car driven by a grandmother in northern Germany.
Different state-of-charge levels increase the complexity to estimate battery health. The SoC on a new battery can be determined relatively easily with impedance spectroscopy, however, the formula changes as the battery ages. A battery tester should therefore be capable of examining new and old batteries with a charge level of 40 to 100 percent. With ample data, this is possible because natural aging of a battery is predictable and the scanned information can be massaged to calculate age. This is similar to face recognition that correctly identifies a person even if he or she has developed a few wrinkles and has grown gray hair.
Simplifications in matrix development are possible by grouping batteries that share the same chemistry, voltage and a similar capacity range into a generic matrix. This simplifies logistics; however, the readout is classified into categories rather than numbers. Figure 6 illustrates the classification scheme of Good, Low and Poor. Good passes as a good battery; Low is suspect and predicts the end of life; and Poor is a fail that mandates replacement.