Using Confidence Limits for Sequence Speed and Accuracy (100 Things #77)
SoundCheck’s Measurement Confidence is a powerful way to qualify factors that may influence measurements, like noise, settling time, latency, and quantization. Using the confidence limits feature in the analysis step, SoundCheck can display measurements along with any measured discrepancies. A sequence’s stimulus can be edited with tons of flexible options, including increasing both cycle count and minimum duration of a Stweep’s steps. Adjusting these stimulus parameters can fine tune your sequence, yielding high measurement confidence.
Using Confidence Limits for Sequence Speed and Accuracy
Learn more about using confidence limits in SoundCheck
Our Working From Home with SoundCheck video series features an even more in-depth demonstration of confidence limits, diving into the trade off between measurement speed and accuracy.
Video Script:
There are many factors that affect measurement accuracy. Noise, settling time, latency, quantization, sampling and other impediments, all affect the quality of your measurements. SoundCheck’s powerful ‘Measurement Confidence’ tool helps you assess how these factors impact your measurements and helps you fine-tune any trade-offs between speed and accuracy. This feature is found in the HarmonicTrak analysis editor on the Distortion tab.
Let’s focus on noise as the source of our measurement error. SoundCheck’s HarmonicTrak algorithm uses FFT analysis to measure the levels for each step of a stepped sine sweep. The shorter the duration of a step, the larger the width of the FFT filter around its center frequency which allows a greater amount of noise to collect in the filter band. Noise adds to the sinewave and if you measure it 10 times, you’ll get 10 slightly different results. The standard deviation of the 10 measurements is called the standard error which is directly proportional to the noise collected in the FFT filter and to the duration of the step. By evaluating the level of the noise present in the FFT, SoundCheck is able to give you the standard error at each frequency.
Additionally, SoundCheck can produce limits within which you can expect the true value to lie. These are called Confidence Limits and there’s even a sequence included in your Soundcheck installation which demonstrates this feature. Let’s take a look.
The sequence uses three different Stweep configurations to show how the Confidence feature works. SoundCheck’s stimulus editor gives us precise control over the duration and number of cycles per step in our Stweep. In this example sequence, the first stimulus is configured so that each step contains 3 cycles. In the second stimulus, each step contains 10 cycles and in the third stimulus we add a minimum duration value of 10 ms in addition to the 10 cycle requirement. Let’s run the sequence and take a look at the results.
In sweep 1, the confidence limits are quite wide and drop off dramatically at the low frequency and high frequency where the signal to noise decreases due to the lower levels. It’s no surprise, considering the small number of cycles per step. Increasing the minimum number of cycles to 10 dramatically improves the results. Confidence is overall very good with the exception of the very high frequencies. In the last sweep, by introducing a minimum duration requirement of 10ms, we can see a further improvement at high frequencies since we are extending the duration of the steps at the highest frequencies.
I hope that this demo gives you a taste of what you can do with the Measurement Confidence feature in SoundCheck. There’s a more extensive demo of this feature in our Working From Home with Soundcheck video series and further details in the SoundCheck user manual.
Learn more about SoundCheck features and functionality.
