Car Audio Measurement – Part One

How do you measure car audio? In this video, we demonstrate a complete car audio measurement setup using SoundCheck, based on the AES TC-AA recommended configuration. You’ll see how to test automotive infotainment systems inside a real vehicle using a calibrated 6-microphone array in the driver’s seat to capture spatially averaged data. We walk through measurements for frequency response (using both correlated and uncorrelated pink noise), seat-to-seat spectral uniformity, total harmonic distortion (THD), and intermodulation distortion.

Watch this Video Demo of Car Audio Measurement Techniques

Learn More about How to Perform Car Audio Measurements

Here are some additional resources:

• Standardized Automotive Audio Testing Seminar: https://www.listeninc.com/standardized-automotive-audio-testing-seminar/
• Automotive Audio Measurements Seminar (Frequency Response, Max SPL, Buzz, Squeak & Rattle): https://www.listeninc.com/automotive-audio-testing-seminar-2/
• Standardized Automotive Audio Measurements (presented at ISEAT conference): https://www.listeninc.com/standardized-automotive-audio-measurements/

Video Transcript: How to Measure Car Audio – Part One

Today I’m going to demonstrate in-car audio measurements, specifically frequency response, spectral uniformity and THD, using the test configuration recommended in the AES TC-AA white paper on automotive measurements.  

We’re using a precisely positioned 6-mic array in the driver’s seat to represent our listener. This is connected to  an AmpConnect 621 Audio Interface. Our laptop is also connected to a measurement-grade Bluetooth interface  which we’re using to send the stimulus signal into our car’s head unit. If you want to know more about the test configuration, you can download the AES white paper, and we also have full details on our website.

I’ve already calibrated the inputs and outputs of the test system by measuring the sensitivity of each individual microphone in the array and storing these readings in the software calibration menu. I also calibrated the audio interface to ensure proper response frequencies and levels, both in the analog and digital domains, so we’re ready to measure.

First we’ll measure frequency response using 80 dBA monophonic, correlated, pink noise on both channels. I’ve already calibrated the level, and we’ll play a short burst. Ideally the stimulus should be longer but I’ll still get a pretty accurate result. Let’s start the sequence. Enter the information it’s asking. OK, the overall level is 79.5 dB, quite close to 80, with a delay of 1.45 seconds which SoundCheck has measured and compensated for. What I’ve actually done here is a 1/3 octave analysis of each individual microphone, from 1 through 6, and you can see the difference due to location, reflections, and standing waves in the car. The black curve is the overall power average or spatial average of all six mics and represents what a human would hear.

Now let’s  measure the frequency response using uncorrelated pink noise. This randomizes the noise between the two channels rather than playing identical signals out of both. We expect to see some low frequency cancellation due to phase differences.

So let’s run the sequence… and we’ve got our results.

We can compare the results with uncorrelated noise to the results when I used correlated pink noise. Let’s bring them both onto the graph… The orange line is the correlated spatial average and the black line is the uncorrelated spatial average. We can see the uncorrelated pink noise has  about 3db less output than the correlated pink noise below 100 Hz and very little above 1KHz.

We can also look at the seat to seat spectral uniformity. Ideally each passenger in the car experiences the same high quality sound, but it’s hard to independently equalize the loudspeaker playback system for each seat in the car. Typically the driver’s seat is favored, the frequency response in the other seats is not quite as smooth.

I’m not going to do this now as it takes a while, but I’ll show you some measurements I made earlier. I repeated the first frequency response measurement with correlated pink noise in each of the car’s four seats and here you can see that the base performance in the rear passenger seats, the orange and green lines are not as extended or smooth as the front passenger seats.

To calculate the seat to seat spectral uniformity, We compare the individual seat response to the maximum and minimum responses to find the seat with the biggest difference. In this example both the left and right rear passenger seats have over 10 dB difference from the front seats at some

Frequencies. The lower the number the more uniform the frequency response is from seat to seat.

Now let’s make some distortion measurements. We’ll start with THD.

THD indicates the overall nonlinear performance of the device, and if it’s outside the limits it might indicate a poorly centered voice coil or mechanical defects. We measure total harmonic Distortion at 80 dBA,  this time using a stepped sign sweep from high to low with 12th octave steps.

Let’s run it…  As before, we power sum the spatial average of all six microphones to produce the THD power average curve, this black line here, which smooths out the standing waves and reflections picked up by the individual microphones. That’s not too bad –  if we look at the power average we’re getting roughly 2.2% from 50 to 10 kHz.

Next we’ll look at intermodulation Distortion. This is a good test for multi-way speaker systems with a crossover- if a small driver is trying to play back low and high frequencies simultaneously it’ll struggle to play both frequencies and create IM distortion. Here we use a two-tone stimulus with a fixed tone set to 50Hz, and a sweep tone that sweeps high to low from 20 kHz down to 150 Hz. I’ve already run a couple of sweeps to set the auto delay and to Auto Range the input gain for the best signal to noise ratio – this is important to ensure that the harmonics are above the noise floor of the measurement. Now let’s start the measurement. 

OK… The graph shows the fundamental and total Intermodulation Distortion of the second and third order intermodulation products, power averaged from the six microphone array. As you might expect, there’s more IM distortion at the lower frequencies, but you can see an increase around 700 Hz – that’s probably a crossover frequency in this car. If we take the spatial average of all six mics we get an overall IM distortion level of 1.4%.

That’s all we have time for here – check out part 2 for Max SPL and more distortion measurements!