# Getting the Most Out of the MAX14920/MAX14921 High-Accuracy Battery-Measurement AFEs

Abstract: The MAX14920/MAX14921 is a high-accuracy battery-measurement analog front-end (AFE) ideal for cell balancing and battery-measurement applications. The performance of the system in which the MAX14920/MAX14921 is used, however, depends highly on the components that support and surround it. This application note provides a framework for, and examples of, selecting system architectures and building blocks that meet diverse accuracy and cost requirements.

*EDN*, November 19, 2013.

## Introduction

## Building a System that Meets Your Needs

- Select an architecture.
- Determine important parameters.
- Select system components.

**Table 1**shows the relative cost, maximum expected six-sigma error, and maximum expected three-sigma error of the architectures covered by the following analysis. The six-sigma error denotes the maximum error statistically expected on 99.99966% of all boards built using the respective architecture, while the three-sigma error denotes the maximum error statistically expected on 99.73% of all boards built using the same architecture.

Table 1. Cost/Performance Comparison of Architectures | |||

Architecture | Relative Cost | Three-Sigma Error | Six-Sigma Error |

Accuracy Optimized | High | 1.087mV | 1.714mV |

Accuracy-Optimized, Cost-Reduced | Medium-High | 5.014mV | 7.305mV |

Cost-Optimized, Accuracy-Enhanced | Medium-Low | 17.632mV | 31.154mV |

Cost-Optimized | Low | 161.849mV | 251.307mV |

## Accuracy-Optimized Architecture

*Figure 1. Accuracy-optimized architecture.*

### Architectural Analysis

**Figure 1**, has the potential to provide the greatest accuracy of any of the architectures because it offers flexibility for individual selection of both the main components that contribute to system precision: the analog-to-digital converter (ADC) and the reference. Much like ADCs that are external to the microcontroller, references that are external to the ADC are more suited for high-accuracy systems. External references have better initial accuracy and temperature coefficient values than references internal to the ADC. This enhances both room temperature accuracy and accuracy over the whole operating temperature range of the system. By selecting an external reference and an ADC separately, however, additional cost is incurred, and the designer must marginally increase the price of the system as compared to other architectures to achieve the greatest accuracy.

### Important Parameters

- ADC
- Offset error
- Gain error
- Integral nonlinearity (INL)
- Gain-error temperature coefficient
- Offset-error temperature coefficient

- External Reference
- Initial output voltage accuracy
- Output voltage temperature coefficient

### Component Selection/Example

**Figure 2**. A complete schematic, layout, and bill of materials can be found in the data sheet for the MAX14921 Evaluation Kit.

*Figure 2. Maxim's evaluation kit solution for the MAX14921.*

**Figure 3**shows actual cell measurements taken in the lab on the MAX14921EVKIT over temperature. Measurements were taken at various cell voltages and show a maximum change in error of only 0.009% (368µV) over temperature, with a maximum cell voltage measurement error of 0.017% (696µV).

*Figure 3. MAX14921 system measurement error over temperature.*

*Figure 4. Error calculations for accuracy-optimized architecture.*

## Accuracy-Optimized, Cost-Reduced Architecture

*Figure 5. Accuracy-optimized, cost-reduced architecture.*

### Architectural Analysis

**Figure 4**an excellent solution for systems requiring high accuracy within a budget.

### Important Parameters

- Offset error
- Gain error
- Integral nonlinearity (INL)
- Gain-error temperature coefficient
- Offset-error temperature coefficient

### Component Selection/Example

*Figure 6. Error calculations for accuracy-optimized, cost-reduced architecture.*

## Cost-Optimized Architecture

*Figure 7. Cost-optimized architecture.*

### Architectural Analysis

- The on-board ADC typically has very low accuracy performance.
- A 3.3V microcontroller may have a reference around 1.195V. As a result, the integrated ADC can only accept a 1.195V full-scale voltage, but most cell-stack monitoring applications require a full-scale voltage of 4V or higher.

### Important Parameters

_{1}= 1MΩ requires R

_{2}= 2.347MΩ. Such precise values are not available, so the designer must settle for a nearby value that is purchasable. In this case both 2.32MΩ and 2.37MΩ are available. It is advised to choose an actual value higher than the calculated value to keep the full-scale voltage of the divided signal below the 1.195V reference. If a value lower than the calculated value is selected, the full-scale voltage will exceed the reference causing loss of data. With R

_{1}= 1MΩ and R

_{2}= 2.37MΩ, neglecting all other factors, the voltage-divider introduces 8.05mV of error.

- The errors introduced by the voltage-divider as mentioned above are introduced at the input to the ADC. This means that the error is included in the value output from the ADC. This value is then multiplied by the IN/OUT ratio. As such, the true error introduced at the output by the voltage-divider is the IN/OUT ratio (3.35 in this case) multiplied by the sum of the imperfect resistor error value and the maximum error due to resistor tolerance. In the above example, using 0.1% tolerance resistors, this means that the 9.73mV of error on the input becomes 9.73mV x 3.35 = 32.6mV on the output.
- The IN/OUT multiplier also amplifies the ADC and reference errors. Thus, if the total error introduced by the ADC and reference is 1mV, the error from the ADC after the multiplier is 1mV x 3.35 = 3.35mV.

**Note:**The high value of R

_{2}= 2.37MΩ along with the 10pF (max) input capacitance to the ADC forms a RC time constant that must be accounted for in the system. Waiting 5 times the RC value, or about 120µs in this example, before beginning the ADC conversion allows the input capacitance to charge before sampling the signal.

### Component Selection/Example

*Figure 8. Error calculation for cost-optimized architecture.*

## Cost-Optimized, Accuracy-Enhanced Architecture

*Figure 9. Cost-optimized, accuracy-enhanced architecture.*

### Architectural Analysis

### Important Parameters

_{1}and R

_{2}in the voltage-divider. Refer to the "Cost-Optimized Architecture" section for analysis of the impact of the voltage-divider on the accuracy.

### Component Selection/Example

_{1}= 1MΩ yields a calculated resistor value of R

_{2}= 212.121kΩ. Since 210kΩ and 213kΩ resistors are the closest available, the 213kΩ resistor will be selected to avoid exceeding the reference voltage at full-scale. With these resistor values on the voltage-divider, the maximum error introduced by the voltage-divider is 3.55mV.

**Note 2:**For a low-cost 3.3V voltage reference with higher accuracy than the MAX6034B, Maxim recommends the MAX6034A voltage reference.

*Figure 10. Error calculations for cost-optimized, accuracy-enhanced architecture.*

## Conclusion

## Appendices

### Appendix 1 - Layout Matters

### Appendix 2 – Calculating Error Using "Error Measurements" Spreadsheet

#### MAX1492X Error Calculations Worksheet

*Figure 11. MAX1492X error calculations worksheet.*

*Figure 12. Rows 2 to 8.*

#### Repeat Interval

#### Sampling Time

#### Hold Time

*Figure 13. Timing diagram.*

#### Full Derating of Capacitor

### Total Error Calculations Worksheet

*Figure 14. Total error calculations worksheet.*

*Figure 15. Application-specification cells.*

_{1}into cell H3, and the tolerance of the resistors into cell G3. The spreadsheet automatically obtains the "OUT" value in cell F3 from the Reference Voltage cell (B4). Once this data has been entered, the equation in cell J3 will give the calculated R

_{2}value. Based on availability, the chosen R

_{2}value will be different from the calculated value. The actual value should be entered into cell K3, and when that is done, cell M3 outputs the maximum error introduced by the voltage-divider. This value may be negative, reflecting the fact that the error introduced by the voltage-divider causes error below the maximum reference value. The voltage-divider error value is automatically calculated by cell E32 and entered into the final calculation.

*Figure 16. Voltage-divider error calculator.*

**Figure 17**). Each section has relevant specifications for each component as described above. Because there are many approaches to specifying error in particular instruments, there are multiple options for each section. Each category of data has a pulldown menu in column F to allow change of units. For instance, cell F17 allows the user to select percent accuracy or accuracy as measured in mV depending on how the data sheet specifies this characteristic (

**Figure 18**).

*Figure 17. Component error calculations.*

*Figure 18. Selecting unit of measurement.*

**Figure 19**.

*Figure 19. Mutually exclusive parameters.*

*Figure 20. Mutually exclusive parameters error message.*

**Figure 21**.

*Figure 21. Total unadjusted error message.*

*Figure 22. Component error calculations.*