Standard Weighting Curves

Included with Daqarta are several standard weighting curves, intended particularly for use with the Sound Level Meter option. They can be loaded and selected from the Spectrum Curves dialog, which is opened from the Spectrum control dialog or from the Weighting Curve button in the Sound Level Meter.

A (A-Weight.CRV): One of the most common weighting curves and arguably one of the worst. It was an early attempt to reflect perceived loudness, but it is actually valid only for pure tones at relatively soft levels. Its response falls off steeply below 1 kHz and above 6 kHz, with a slight rise between those frequencies. This curve is usually available (along with C on even the most inexpensive sound level meters.

B (B-Weight.CRV): Flat between 400 and 4000 Hz, falling off steeply outside this range. It is no longer used but is included here for legacy work.

C (C-Weight.CRV): Flat between 100 and 4000 Hz, falling steeply above and below. This curve is used almost as much as A, sometimes in lieu of a true "Flat" response on inexpensive meters.

D (D-Weight.CRV): Special curve developed for measuring high-level aircraft noise, this curve has a large peak around 6 kHz to reflect human noise sensitivity.

ITU-R 468 (ITU-R_468.CRV): Special curve based on research by the BBC to reflect the perceived loudness of all types of noise, including bursts. It also has a large peak around 6 kHz. It is intended to be used with an exotic "quasi-peak" response designed to give reduced sensitivity to narrow noise bursts, similar to human hearing.

ITU-R ARM (ITU-R_ARM.CRV): Identical to ITU-R 468 except shifted down by 5.6 dB so it crosses 0 dB at 2 kHz instead of 1 kHz. Instead of the special quasi-peak response, it uses a simple time constant. (ARM stands for Average Response Meter.)

You can view the shape of any curve by applying it to a signal with a flat spectrum. There are two basic ways to get a flat spectrum with the Generator: White noise or narrow pulses. View the Generator output directly in Daqarta for best results.

White (or Gaussian) noise has a flat spectrum when averaged over a long-enough time. You will probably find Exponential Spectrum Averaging to be easiest to deal with, with Frames set to 32 (default). The spectrum will still have a lot of tiny wiggles, but the curve shape will be obvious.

The spectrum of a single-sample pulse is perfectly flat, and doesn't require any averaging. Use a Pulse wave with the A Width set to 1 sample and the B Width set to zero. Set the Tone Frequency lower than the sample rate divided by 1024 (46.875 if the sample rate is 48000 Hz), so that you only get one pulse per trace. Note that this method is only perfect if everything is internal to Daqarta... if you try to create a real-world pulse that is then fed back to the Input, there may be distortions introduced along the way.

The spectrum must have Y-log (dB) active for Spectrum Curves to take effect. You will probably also want X-log active to see the curve the way it is customarily shown. You can use User Units and adjust Units/Volt to set the unweighted curve to 0 dB, so that the weighted relative response differences can be easily seen and measured with the cursors.