X3100 Current Measurement Accuracy

APPLICATION NOTE

AN142 Rev 0.00 Page 1 of 8

May 5, 2005

AN142

Rev 0.00

May 5, 2005

Description

This test measures the accuracy of the current sense

amplifier circuit of the X3100. In the test, a known current is

applied through a sense resistor to generate a voltage

between the current sense inputs of the X3100. The analog

output voltage of the X3100 is measured. This value

(adjusted by the amplifier gain) is compared to the input

voltage. The collected data includes the values from five

devices from three different lots. The data also includes tests

at 0°C, 25°C and 50°C and at each gain setting.

Test Setup

This test applied a current across the sense resistor to

develop a voltage ranging from approximately 200uV to

100mV. The current source was forced through a 20.37

resistor that connected across the VCS1 and VCS2 pins.

The 20.37 resistor was chosen to improve the accuracy of

the resistance measurement. In an actual battery pack, a

resistor in the range of 25m to 50m would be used. At

200uV, the minimum current measured would be 4 - 8mA. In

this current measurement test, the current was forced in

specific discrete increments, such as 10uA, 20uA, etc.

However, since the resistor was 20.37s, the applied

voltage is represented in less recognizable increments.

For each applied voltage the output was recorded, first in

one direction, then in the other. This involves the changing of

the analog multiplexer setting in the X3100. While this

procedure cancels the majority of the offset, as the data

shows, a small offset remained even after this procedure. In

gathering the data with this test, the output was sampled 20

times for each setting. The high and low values were

discarded and the remaining samples were averaged. This

procedure was used to reduce the effect of external noise on

the test environment.

Results

The data for this test is available in raw form in an Excel

spreadsheet. This spreadsheet includes data for each

device, data for each lot, and data for all devices. Each set of

data includes results for the three temperature settings and

the four gain settings. The data was copied to other files for

the purposes of generating graphs.

For the result summaries there are raw error measurements

and error measurements assuming both an offset correction

and a gain correction. The offset correction was derived by

taking the average offset over all devices, temperatures and

lots. This offset value was then added to each of the

measured outputs at each gain. After this correction, the

gain was computed for each input voltage. These gain

values were averaged over the devices, temperature and lot.

This average value was then factored into the results.

Vcs1

Vcs2

Vss

X3100

20.37



I = 10uA, 20uA, 30uA...5mA

FIGURE 1. CURRENT SENSE MEASUREMENT

X3100 Current Measurement

Accuracy

AN142 Rev 0.00 Page 2 of 8

May 5, 2005

On the following charts, errors are calculated as follows:

Raw error:

Average error:

Average error after offset and gain adjust:

Worst case error at 0°C, after offset and gain adjust:

Worst case error at 50°C, after offset and gain adjust:

Where:O = Output

I = Input

G = Gain

avg = Average

std = Standard Deviation

off = offset

OG = Offset/Gain

(x) = at x degrees C

Overall Performance

The following set of graphs shows the performance of the

X3100 over temperature and across three lots of five devices.

The graphs will show the basic input/output curve, raw error,

and error after correction for offset and gain.

The below data shows that the error at the x10 gain is the

lowest. Using a 20m current sense resistor, and considering

3 standard deviations from the worst case temperature (over 3

manufacturing lots), the X3100 output is accurate to less than

1.4% at 300mA of load current and above. At a gain of 25, the

X3100 has less than 5% error at a load current of 50mA. This

error declines at higher currents to about the same as the x10

gain. At all gains, the error increases as the input current

decreases.

Comments

The AO pin on the X3100 is biased to a nominal 2.5V when

reading the voltage across the VCS1 and VCS2 pins. However,

there is an offset to this value that can make the output (at 0V

input) either higher or lower than this 2.5V nominal value. The

output with 0V current sense input ranged from 1.8 to 2.9V for

the devices under test (at a gain of 160). The difference in

voltage between the nominal output and the actual output

becomes more apparent at higher gains. This is due to a small

input offset in the current sense amplifier that is magnified at

the output by the gain of the amplifier. This offset is mostly

cancelled by reading the voltage across the CSI pins in both

directions and subtracting the two values.

While the process of reading the voltage in both directions and

subtracting the two values cancels most of the offset on the

X3100 current sense amplifier, it does not cancel it all. The

remaining output offset is approximately 668uV at a gain of 10,

1.7mV at a gain of 25, 5.7mV at a gain of 80 and 11mV at a gain

of 160. This offset becomes significant at the lowest input

voltages, since it can be nearly the same voltage as the expected

output. In some of the charts that follow, the output has been

compensated by subtracting this offset from the recorded values.

This cancels out the error for average conditions, but leaves large

errors at small input voltages due to variations in this offset. With

the test setup used it is not possible to determine whether these

offset error variations are caused by the environment or the

device.

The charts in this document that show errors after offset

compensation use an average offset value. This value is an

average over all devices across all temperatures. To improve the

accuracy, especially at lower input voltages, a calibration

procedure could be used. This calibration procedure would apply

0V across the VCS1 and VCS2 pins. Then, at each gain setting,

the measured output (after reading in both directions and

subtracting) is saved as a common offset value.

The gain adjust is determined by dividing the output by the input

voltage. The gain is averaged over all inputs and all devices at all

temperatures. This average value is then used to divide into the

output reading to determine the compensated error. Based on the

data collected, the average gain for the 160 setting is 160.17, the

average gain for the 80 setting is 79.24, the average gain for the

25 setting is 24.53 and the average gain for the 10 setting is 9.77.

Using these average values will reduce the error when computing

the input current. However to improve the accuracy further, each

device can be calibrated for the particular gain. To do this, apply a

known current across the current sense resistor. Use a current

that is relatively high, such as 500mA. At each gain read the

output and divide this by the known input. Use this gain value

subsequent calculations.

Raw

---- I–

------------=

Avg

avgO

--------------- I–

-----------------------=

AvgOGadj

avgO avgoff–

avgG

---------------------------------------- I–

------------------------------------------------=

StdOGadj 0

avgO 0 3std0+

avgG

---------------------------------------------------------- I–

-----------------------------------------------------------------=

StdOGadj 50

avgO 503std50–

avgG

---------------------------------------------------------------- I–

------------------------------------------------------------------------=