Loudspeaker measurement technique

Note: this historical article refers to the ADC-216 high-resolution oscilloscope, which has been replaced by the PicoScope 4262.

Safety tips

Though the energy in the impulses used is very small, the peak level may be enough to damage loudspeakers. Make sure that the level is set within the working range of the loudspeaker you are measuring.

A list of the equipment needed

Initial setup

To make an on-axis response for the loudspeaker you will need to set your speaker up on a stand as far from all reflecting surfaces as possible. You should also try to find the quietest place to do the measurements as noise added to the response will degrade the results. The measurement microphone should be placed about 1 m in front of the speaker on your chosen measurement axis. When you have set up the mic and speaker measure the distances to the nearest reflecting surface with a tape measure. Normally this will be either the floor or the ceiling.

A little simple geometry (Diagram 1) will tell you what sort of low frequency cutoff to use.

Low frequency cutoff = 343 /( (2 * (x2 + h2)0.5)

Where 343 m.s-1 is the speed of sound. For a room with a ceiling height of 2.4 m and a measuring distance of 1 m this gives:

Low frequency cutoff = 343 /( (2 * ((0.52 + 1.22)0.5) = 132 Hz

So, for this size room you know that any data shown below 132 Hz is rubbish! Bear in mind that you should put an extra margin on this because there are very few data points collected in the lower octaves of the measurement.

The next stage is to connect up the equipment (Diagram 2). Notice that the electrical pulse is fed into the ADC-216 as well as the pulse received from the microphone. This means that you can check the bandwidth of the electrical pulse before you begin the acoustic measurement. It is a good idea to also check the output from the power amplifier to make sure that it is not overloaded by the pulse input. This sort of overload will be hard to detect from the microphone output so check that the output of the power amp is within its rating.

Loudspeaker measurement technique

Diagram 1

Loudspeaker measurement technique

Diagram 2

Frequency response measurement

Note: If you want to make a set of measurements for comparison you must use identical settings for each. Otherwise the relative levels of the measurements will be different.

For this note it is assumed that the output of the pulse gen is in channel A and the return signal from the microphone in channel B.

Start off by looking at the pulse. The screen below shows the pulse and its spectrum (Figure 1).

Note the settings for trigger and spectrum view (see the PicoScope help files for more details).

The Scope View Settings > Options were set to:

  • ‘data to display’ = current.

The Spectrum View Settings > Options were set to:

  • ‘X scale’ = Log
  • ‘Y scale’ = dB
  • ‘Window’ = rectangle
  • ‘No of spectrum bands’ = 512
  • ‘display mode’ = normal

In Spectrum View Settings > Measurements > Measurement List > Add:

  • ‘Measurement’ = Scan time
  • ‘Channel’ = ch A

Notice that the pulse width is 25 µs and that this has a 3 dB rolloff at around 20 kHz. If the pulse is made longer, then the highest frequency of measurement drops. There is a tradeoff between: getting enough power into the pulse to overcome background noise reaching the mic, and bandwidth. There is an upper limit on the pulse height set by the power amp and speaker. If you do not want to measure above 10 kHz increasing the pulse length will improve the noise performance. Notice that the Spectrum View window is rectangular. Do not confuse this with the measurement time window. To get good results for the loudspeaker measurements you can try changing the Spectrum View window. The Blackman window is probably best.

Having set the pulse length you can now start looking at the acoustic output, you will need to adjust the repetition rate of the pulses. Set the impulses going through the speaker and observe the results on PicoScope.

Note: set the trigger to operate on the microphone signal.

The screen below shows the results. Notice that the red pulse from the microphone is delayed just over 2.8 ms relative to the electrical (blue pulse). The measurement distance in this case was 95 cm which gives us a measured speed of sound of 333 m.s-1 (a good check!) (Figure 2).

Figure 1

loudspeaker measurement

Figure 2

loudspeaker measurement

Figure 3

When setting the repetition rate the reverberation time of the room must be considered. As stated earlier the first reflection from the room determines the measurement window we can use. However, the energy from the room reflections continues to arrive at the microphone for some time after this. In an average lab the sound from the pulse would probably take from 0.5 s to 1 s to decay away. If your room is very reverberant (large with little absorption) the decay may take longer, up to 5 s or 10 s. The importance of this is that the tail of the reverberation from one pulse must be allowed to decay away before the next one arrives. If this is not allowed to happen the reverberant energy is added on to the measurements. It is best to set your repetition rate for between 0.5 s and 1 s. Check what the Scope View is showing immediately before the mic impulse and reduce the repetition rate to see whether the noise floor reduces.

Once you have set the pulse repetition rate you can start to make measurements on your loudspeaker. The screen to the right shows what happens if you put a hard reflecting object (an A4 size notebook) about 10 cm behind the measurement microphone (Figure 3).

The reflection is quite clear about 0.5 ms after the main pulse. The reflection produces very noticeable peaks and dips in the measured response. This demonstrates the importance of removing reflections from the measurements. Large reflections from hard surfaces will make a big difference to your results but even the smaller reflections will introduce errors that may mask real attributes of the loudspeaker.

loudspeaker measurement

Figure 4

Tuning out the reflections

To make the best use of the features of the ADC-216 and PicoScope and produce good measurements you will need to adjust the triggering, bandwidth and number of data points collected.

For the fastest measurement time set the number of data points in Spectrum > Options to 128. You can increase this number for greater frequency resolution. If you do increase the number of data points the size of the measurement window also increases. This will be shown by the scan time measurement. Decreasing the bandwidth of the capture also increases the scan time.

To effectively remove reflections (echoes) from the measurement you should change the trigger delay. By increasing the delay (setting the percentage more negative) you throw away the data at the end of the trace. The best way to set this up is to put one cursor in Scope View at the start of the pulse and the other at the value shown for scan time in Spectrum View. The cursors will track as you change the trigger point. Once your end cursor moves off the Scope View you know that you are effectively truncating the captured data. The window length cannot be any longer than the scan time, moving the trigger point reduces the value further.

Remember that the window needs to be just short enough to remove the first reflection from the measurement. Making it much shorter than this reduces the measured data and raises the low frequency cutoff of the measurement. It is worth taking time to adjust the bandwidth, number of samples and trigger point to see what effects these have.

The screen to the right (Figure 4) shows the results for a 20.8 kHz bandwidth measurement taken with 256 data points and the trigger delay set to –55 % giving a window of about 4.5 ms. This measurement was taken using the Blackman window for the spectrum calculations. The time to the nearest reflecting surface works out to be about 6 ms. The effective window time means that the measurement is good only above about 220 Hz. With 256 points the frequency spacing is 81.5 Hz so there are only 3 data points below the cutoff frequency.

Dealing with noise

Where possible you should try to use a quiet room for your measurements, maybe using the room outside normal working hours. However, you may find that the background noise in your test environment still makes it difficult to get good results. If this happens there are three things that may help. First, use a steep high pass filter on the microphone to remove low frequency noise. You can tailor this filter to the practical low frequency limit of your measuring room. Second, use the averaging built in to PicoScope. Using averaging is a very effective way of removing random noise from your measurements. You can turn on the averaging for both Scope and Spectrum Views in their respective Options.

Note: the averaging applied to the Scope view is not transferred to the spectrum calculation.

When you use averaging you must be prepared to wait for the data to reach an equilibrium. For each 3 dB of noise improvement you must double the length of time over which the averaging takes place. This means that if your data blocks are being collected each second you will have to wait at least 26 s (just over a minute) for an effective improvement of 18 dB.

Your third option is to increase the power in the impulse. This means either increasing its height or width (or both). Take care if you opt to increase the height of the pulse as it may overload the power amplifier or result in damage to the loudspeaker. If you increase the width of the pulse you reduce the upper limit of the frequency measurement. You should check the response of your electrical pulse (described above).