Whitepaper: Driver Time Alignment

Transcription

1 Whitepaper: Driver Time Alignment definiteaudio GmbH Peter-Vischer-Str.2 D Erlangen Tel: Fax: Web: Umsatzsteueridentnummer: DE HRB Fürth Seite1

2 Index Index...2 Preamble...3 Effects of incorrect Time Alignment...3 Measuring Time Alignment...4 Seite2

3 Preamble Implementing the The Perfect XO for High End Stereo Systems described in (1) special care has to be taken of the speaker s time alignment. Effects of incorrect Time Alignment To show the importance of time alignment let s first revisit the 40 th order Bessel filter described in (1). Figure 1: Bessel LP 40 th order with derived HP frequency response Figure 2: Bessel LP 40 th order with derived HP step response The frequency response in figure 1 shows the perfect addition (blue curve at 0db) of the LP (red) and its derived HP (green) frequency responses. In the time domain (figure 2) the responses of both filters add to a perfect step (blue). This perfect summation of the LP and HP only could take place if the drivers of the speaker are time aligned, meaning the wavefront of both drivers reaches the ear of the listener at the same time. To show the effect of misalignment let s assume the bass driver is about 10cm further away from the listener than the mid driver. This is not very much but quite realistic for normal HiFi systems. Assuming the speed of sound at about 343m/s the sound needs 0.3ms to propagate the 10cm misalignment. This means the wavefront of the bass driver arrives at the listener s position 0.3ms later than the wavefront of the mid. Adding this delayed step response of the LP and the step response of the HP both signals did no more sum up to a perfect step (figure 3 blue curve). Figure 3: Addition of delayed LP and HP response Seite3

4 Looking at the frequency response the whole dimension of this small misalignment gets clear. Figure 4: Effect of misalignment in the frequency domain The 0.3ms time shift of the LP leads to a deep notch of -15db at the crossover point. This very unpleasant notch only could be avoided by aligning the driver the way that the wavefronts of both drivers reach the listener s ear at the same time. This necessary alignment could be achieved either by physically shifting the drivers of the speaker against each other (in this example the bass has to be shifted 10cm towards the listener) or by digitally delaying the signal transmitted by the mid driver. Using a convolution engine like BruteFir delay either could be specified in the configuration file of BruteFir or the coefficients of the filter could be shifted. Measuring Time Alignment Having understood the effect of time misalignment a method is needed to measure time delay between the drivers of the speaker. For this measurement some material is needed: A PC with a bi-directional soundcard (playback and record at the same time) A program which supports playback and recording at the same time A microphone with preamp connected to the soundcard A suitable test signal. A PC with a bi-directional soundcard should be no problem. Any 44.1kHz soundcard should be sufficient. For playback and parallel recording Audacity ( is a great and free of charge tool. For recording the test signal I use a Behringer ECM8000 microphone ( which also could be used as measuring microphone for frequency responses. As microphone amplifier I use a Behringer Ultravoice Digital VX2496 ( with all effects off. This is a quite cheap (100Euro) unit for microphone amplification with phantom voltage generator and a digital S/PDIF output up to 96kHz! To measure time delay a simple pulse should be the best suited. A tweeter would transmit such a (dirac-) pulse quite well but with a bass driver such a low amplitude response would be created that it could no more be measured exactly. Therefore a test signal with a broader pulse seems to be better suited for this measurement. For adjusting the bass/mid XO I use a signal consisting of a 1ms pulse. For higher XO points a shorter signal is recommended. Seite4

5 To measure the system each driver has to be measured separately. For measurement BruteFir has to be in its place in front of the power amps and it has to be equipped with the XO filters. Playing back the test signal BruteFir will convolve it with the transfer function of the filter assigned to the driver you measure. Figure 5 gives an impression of the signals reaching the driver after the test signal is convolved with the filters. Figures 5 and 6 show the response of the XO I use to drive my bass and my full range driver. The XO is a 40 th order Bessel XO described in (1) with a take over frequency of 180 Hz. Figure 5: Response to the test signal The light blue curve is the response of the HP, the dark blue curve the one of the LP. As you see both responses are perfectly matched to each other. If this perfect match also would be achieved at the listening position, time alignment of the speakers would be perfect. Running a two-way system like I do, four measurements two for the right speaker and two for the left have to be taken. Running a three-way system six measurements will be necessary. During the measurement only the driver being measured should be connected to the amplifier. All other drivers should be disconnected to avoid accidental measuring errors. Now the microphone has to be placed exactly in the middle of the stereo triangle (where the head of the listener will be). The microphone position has to be adjusted the way that the distance from the front of the microphone to the center of the tweeters inside the speakers is the same for the right and the left speaker. Now Audacity should be started and the test signal loaded. The option Multiplay has to be enabled to allow recording during playback. This could be done with the File/Settings dialog. On the card deck Audio I/O the box Multiplay has to be checked. Now the test signal can be played while the signal of the microphone will be recorded. To perform level adjustment the measurement should be started with the highest frequency driver first. Figure 6 shows two measurements (bass and full range superimposed and zoomed) in one figure. Seite5

6 Figure 6: Response of the pulse measurement This looks somehow different than it should do! Despite this the information needed for time alignment can be extracted. Let s first take a look at the bass response. The main peak is clearly visible and was marked with the red vertical line. The center of the full range s impulse response can be found in the middle of the two steep edges of the response. This is marked by the blue vertical line. Time difference between the red- and the blue line is the delay that has to be added to the full range driver. Figure 7 shows the time aligned response of the full range driver (red curve). Figure 7: Time aligned impulse response The correction could be done easily in the BruteFir setup file by specifying a delay for the full range driver. The delay measurements have to be done separately for the left and the right speakers. Despite of time alignment the measured waveform during playback of the impulse is not as clean as it should be. One reason is the room which is included in this measurement the other is the impulse response of the driver which in general is not ideal. So there is much room for further improving the impulse response. References 1. C.W. Oehlrich, The Perfect XO for High End Stereo Systems?!, Feb Seite6