Sound Card Spectrum Limit Testing

Daqarta trial and Pro-license users can apply limit curves to the spectrum data displayed in Y-log (power spectrum) mode, to allow Pass/Fail testing of speakers, microphones, or other frequency-sensitive devices or circuits.

The image above shows a test of a simple active filter. This is a 2-pole equal-component Sallen-Key design, with a 3 dB Chebyshev low-pass response characteristic. The response is flat at low frequencies, with a level of about -45 dB relative to a 1 VRMS signal. It rises +3 dB to a peak at about 4500 Hz and falls beyond that, reaching -51 dB (-3 dB from the start, or -6 dB from the peak) at about 7200 Hz. The vertical cursors are set to these frequencies, so the readouts show the values there.

At higher frequencies the response nominally falls at -6 dB per octave, until it runs into the sound card anti-aliasing filter response above about 18000 Hz.

The Limits dialog shows that limits are being applied to the Left In channel. The upper and lower limit curves are simple text files called FiltMax5.LIM and FiltMin5.LIM. These are shown as orange lines on the trace. (The line color and style can be set as desired.)

These limit files were created from a reference unit by saving its response as a LIM file with a +5 dB offset, and again with a -5 dB offset. (These offsets were chosen for clarity with this example; real production tests would probably have much tighter limits.) Since these are text files, they can be easily modified. For example, "don't care" regions of the spectrum could be set, as described in the Max and Min Files topic. The same topic shows how limit files can be created entirely with a text editor.

When Show Pass/Fail is selected, a bright green PASS appears at the upper right corner of the trace whenever the measured response falls completely between the upper and lower curves. If not, it changes to a bright red FAIL. The response in this image passes.

For production testing or adjustment of this filter, the frequency response is measured continuously using an Impulse Response test. Spectral averaging is used in Exponential mode with a time constant of 32 frames. This smooths the noise response so random noise peaks don't cause spurious failures.

This allows a continuous test approach that is especially useful for rapid production tests on devices like circuits or transducers (microphones, speakers, etc) that can be quickly inserted into a test fixture, since there is no need for the operator to interact with the computer. The device is connected, the operator reads the PASS or FAIL status, and sorts the part into the appropriate bin.

Alternatively, the same setup facilitates rapid adjustments on the unit under test. The operator can be guided by the limit curves, and can see the results of changes in real-time.

You can use spectrum limits with other frequency response approaches such as Step, White Noise, Pink Noise, or Arbitrary Spectrum responses. The noise tests typically require more averaging than Impulse or Step, but the operator can often detect a defective unit and abort the test well before there is enough smoothing for a valid pass/fail determination.

Swept or Stepped Sweep responses are usually less desireable for production tests because they take much longer to complete. Also, since the last part of the response spectrum isn't visible until the end of the test, a full sweep may be needed to detect a defective unit.

Spectrum Limits can be applied to any spectrum, not just a conventional "driven" frequency response. For example, you could test the output from an automobile horn, smoke detector alert, wind chime, musical instrument, or even a human vocal sound like a sustained vowel.