Confidence Levels

# Confidence Levels

Last updated: December 2, 2008

## What are Confidence Levels?

Using confidence levels for your measurements can reduce your test time. Confidence levels simply end the measurement early, that is, before all the multi-measurements you specified are taken. It ends the measurement early only if it determines statistically that the rest of the measurements should pass. You specify the level of certainty that you are satisfied with, for example, you may decide that you are satisfied in passing a measurement on your mobile if the test set is 95% confident that the rest of the measurements will pass.

Note, the test set only reduces your test time if it determines that the measurement will pass. If it determines that a future measurement may fail, or if a measurement does actually fail, the test set continues as normal executing your test plan.

It is not possible to determine the amount of time by which confidence levels reduce your test time as this is determined by a number of factors, some of which cannot be controlled. The following list details some of the factors:

• The confidence level that you select. This can be set between 80% to 99.99%. Obviously, the lower the level you select, the greater the potential for a reduction in test time.
• The stability of the mobile station's results. That is, if the individual results are variable (that is, differ significantly but are still within the limits), then the measurement will take longer. However, if the results returned are very similar and well within the limits then the measurement will end sooner.
• The test limits that you select. For example, the tighter your test limits the longer the measurement will take.

To ensure that the prediction of the measurement result is accurate there is a minimum number of measurements made by the test set. This number is independent of the confidence level you set. However, this number is dependent on the multi-measurement count and is shown below.

Minimum Number of Measurements

Multi-Measurement Count

Minimum Number of Measurements

1-4

Same as multi-measurement count
5-999 4

## How do you set Confidence Levels?

You can set confidence levels for a measurement using its relevant SETup subsystem. Currently the only measurement in GSM with the confidence level feature is Phase and Frequency Error.

Confidence levels are only relevant when the measurement's multi-measurement count is on. Therefore if you turn on the confidence level feature for a measurement the test set automatically enables that measurement's multi-measurement count feature.

The following procedure and example show the steps needed when you are using confidence intervals via the GPIB.

1. Set-up the mobile station and the test set for your required measurement.
2. Use the SETup subsystem to turn the measurement's multi-measurement count on and specify the count number.
3. Use the SETup subsystem to turn on the measurement's confidence state and set its confidence level.
4. If required, use the SETup subsystem to set the measurements pass/fail limits. The test set uses the defaults pass/fail values if you do not specify any.
5. Start the measurement using the INITiate subsystem.
6. Use the FETCh? command to obtain the measurement results.

### Programming Example

The following example uses the GSM Phase and Frequency Error measurement to demonstrate the use of confidence levels.

```
10  OUTPUT 714;"SETUP:PFERROR:CONTINUOUS OFF" !Configures a PFER measurement to
20                                            !single trigger mode.
30  OUTPUT 714;"SETUP:PFERROR:TRIGGER:SOURCE AUTO"!Configure trigger source
40                                                !to auto.
50  OUTPUT 714;"SETUP:PFERROR:SYNC MIDAMBLE" !Configures a PFER measurement so
60                                           !that burst synchronization, which
70                                           !will synchronize the timing of the
80                                           !measurement algorithm relative to
90                                           !the data sample, will be set
100                                          !to midamble.
110 OUTPUT 714; "SETUP:PFERROR:COUNT 20" !Sets the multi-measurement count state on,
!and the number of measurements to 20.
120 OUTPUT 714; "SETUP:PFERROR:CONFIDENCE 90" !Sets the confidence level feature on,
!and the level to 90%.
130 OUTPUT 714;"INITIATE:PFERROR" !Starts the PFER measurement.
140 OUTPUT 714;"FETCH:PFERROR:FAIL?"
150 ENTER 714;Integrity, Rms_ph_err_fail, Peak_ph_err_fail, Freq_err_fail, Number_of_meas
160 END```

## Using Confidence Levels with the Phase and Frequency Error Measurement

Currently the only measurement in GSM with the confidence level feature is Phase and Frequency Error. The Phase and Frequency Error measurement actually consists of three measurements: RMS Phase Error, Peak Phase Error and Frequency Error. The test set will not end a Phase and Frequency Error measurement early until it determines statistically that all three individual measurements should pass, with a level of confidence that is at least as high as the confidence level you set.

For example, if you set the confidence level to 95%, the test set only ends the measurement when it is:

• at least 95% confident that the RMS Phase Error will pass, and
• at least 95% confident that the Peak Phase Error will pass, and
• at least 95% confident that the Frequency Error will pass.

The level of confidence with which the complete Phase and Frequency Error measurement passes early is equal to the product of the confidence levels with which each of the three individual measurements passed early. This means that for this example, the test set will only end the measurement early if there is at least an 85.74% (95% x 95% x 95%) level of confidence that the measurements that have still to be taken will pass.