Date: Tue, 16 Jun 1998 01:25:47 -0700
To: bass@mcfeeley.cc.utexas.edu
Last post I promised a look at the Virtual Diffraction Line-Source (VDLS) signals that seem to influence the response of RD75 dipole baffles so much. This 'look' is fraught with all the problems that beset the analysis of impulse responses. Acquiring data in a non-anechoic environment seems to always provide for intrusive reflections which contaminate the signal.
The data acquired in the Auditorium is so contaminated. In particular the 1.25" diameter microphone stand created to mount the microphone 8' or more above floor level provided a wonderful 'tutorial grade' reflection in the data. Much of the business of objective measuring is subjective in nature. One must choose the data to analyze, and that choice always has consequences. This is not a perfect world.
Generally, using gated-time instrumentation for the acquistion of impulse responses for subsequent Fast Fourier Transform analysis requires one to select a portion of a data record containing the impulse response. For a semi-anechoic measurement that time record is choosen to exclude obvious reflections in the data. But how are reflections identified? Through experience and theoretical considerations.
Let's look at the impulse response of the bare RD75 suspended in free air at 1 meter distance:
RD75 'Free Air' 1m impulse response. 63.7mS/Gain of 1.
That record contains 63.7mS of data, and is displayed at the IMP/MLS gain setting of 1. Note the impulse complex at the extreme left and intruding reflections starting just past the mid-point of the plot. This isn't the whole story though, what happens when we increase the 'gain' of the plot? Here's the same plot at a gain of 5:
RD75 'Free Air' 1m impulse response. 63.7mS/Gain of 5.
Much more detail starts to appear, and the main impulse is 'clipped' in the plot display. The frequency response data presented earlier in this series for this condition was based upon a 30.2mS data domain selected from the above data. Here's both a gain of 1 and a gain of 5 look at that data sequence:
RD75 'Free Air' 1m impulse response. 30.2mS/Gain of 1.
RD75 'Free Air' 1m impulse response. 30.2mS/Gain of 5.
Clearly, the main reflection 'problem' has been eliminated by truncating the data to 30mS; but, the gain of 5 plot disabuses us of any smuggness over how 'clean' the signal is in the gain of 1 plot. Rarely is the data domain selected at any greater resolution than in the gain of 1 condition due to pragmatic considerations.
Let's keep the gain at 5 and reduce the time scale further. There's a 'glitch' seen in the gain of 5 data at about 8.5mS or so. Here's a look:
RD75 'Free Air' 1m impulse response. 8.9mS/Gain of 5.
Things look interesting now. The 'glitch' at about 8.5mS is identified as a reflection due to the sharp spikes of the pattern. Big as life, right at 2.6mS is a full-blown reflection from the microphone stand. The initial spike is followed by two smaller spikes that match the shape of the main impulse (which is obscured in this plot due to the gain increase). The duration for the reflection is 0.212mS, measured from the start of the first peak to the top of the second peak. That is the same duration as the main pulse.
The above mentioned 2.6mS value from the start of the main pulse to the start of the reflection computes to a distance of 896mm or 35.28". The tip of the Mitey Mic wand was ~18" in front of the microphone stand, hence there is good confirmation of the source of this reflection (for a bounce and return path length from the microphone stand).
There is also a sharp inverted spike at 1.1mS after the start of the main pulse. It will have to remain a mystery for now, since I can find no physical relationship to explain it's presence.
Right at about 4.2mS there is a rounded, inverted dip. Let's decrease the time scale once more:
RD75 'Free Air' 1m impulse response. 4.5mS/Gain of 5.
There are at least two structures in this section that might be identified as Virtual Diffraction Line Source signals. What looks like a single sine-wave cycle at about 2.4mS and the 'half-cycle' at about 4.2mS. Their rounded shapes imply an attenuated high frequency content.
At this point I might expect many to mutter "harumph, those signals are pathetic, and how do we distinguish them from noise, reflections, or other crud?" I offer a final look at the realm of the VDLS signals from earlier measurements taken in the very large gymnasium. This data is much cleaner; and, these measurements are unique in that they were taken from two different floor-mounted baffles in the plane of the baffle, not 'on-axis' as in the data above. These are baffle edge measurements. As such the 'main impulse' signal is much attenutated due to dipole cancellation and appears in the plot at full amplitude. The time duration for the plot overlay is 5.0mS, and the gain is 5.
Comparison: Baffle edge VDLS signals.
Note that there are time-locked paired VDLS signals seen in each plot, and that they move forward in time as the microphone distance is increased from the baffle's edge. These VDLS signals have been 'generated' from a dipole baffle and not just from the edges of the bare driver as we've seen above. Hence they are more fully developed, and their amplitudes compare directly with that of the attenuated main impulse.
With some rigor and attention to detail one can compute the distances represented by the time displacement of the signals and detail the effects of baffle width (can you explain why the signals move forward in time as the microphone is moved away from the baffle's edge?).
The amplitude/shape relationship of the paired VDLS signals in the 1m to 3m plots for the 1S6T baffle is highly unusual, and not yet fully explained. They cannot be microphone stand reflections given the 3m measurement time and distance relationship. Note that nothing is seen in this plot which bears any relationship to the microphone stand reflection seen in the data from the Auditorium.
The VDLS signals apparently directly relate to the issue of "John's Bump" that we saw in the frequency response of the two-sided baffles. They may have direct relevance to the creation and/or enchancement of the dipole ambience phenomena. They are clearly something to be reckoned with in large line-source dipole speakers. It is not clear whether or not they are part of the mechanism that generates the 'moving dips' seen in the frequency responses of the dipole baffles at different distances and heights.
Next time, the swan song of the RD75 Dipole Baffle Study (or as least notice of official hiatus status) in the guise of presentation of the new and best 'Compound Shape' baffle.
John Whittaker
Dipole Baffle Study Report#21.
Dipole Baffle Study Report#23.
The RD75 Dipole Baffle Study - Table
of Contents
Acoustic Line Source Research -
Table of Contents.
