Velodyne DF-661 loudspeaker Measurements
The Velodyne's sensitivity, calculated using B-weighted noise, was basically to specification at 87.5dB/W/m. This is quite high, implying that the DF-661 will go loud with relatively moderately powered amplifiers, provided they can deliver high current. The speaker's plot of impedance magnitude and phase (fig.1), measured using the magazine's Audio Precision System One, indicates a relatively demanding load over most of the audio band, with a minimum value of 3.2 ohms at 235Hz. Note also the slight wrinkles in the traces, between 400Hz and 500Hz, and at 26kHz. These indicate the presence of resonances of some kind. The latter is due to the tweeter dome's "oil-can" resonance and is too high in frequency to have any effect on sound quality. The former, however, is probably due to a cabinet resonance.
Part of the reason for the DF-661's disappointing sound quality can be seen in this graph. The Velodyne does have the double-humped characteristic in the bass typical of a conventional ported design, but it is overlaid with a rising trend with decreasing frequency, associated with increasing negative phase angle. This is due to the capacitor being placed in series with the woofer to change its high-pass rollout from fourth- to fifth-ordera steep 30dB/octave slope. I asked David Hall why he had done this; his reply was that, without the capacitor, the woofer's cone excursion was too high at low frequencies for the speaker to feature as low distortion as possible. However, coupled with the fact that the port is tuned to a high 70Hz, this results in the speaker having very limited low-frequency extension. Which is what I heard.
This is confirmed by fig.2, which shows the individual behavior of the DF-661's drive-units and port. I used the DRA Labs MLSSA system, coupled with a B&K/DPA 4006 microphone and EAR preamplifier, to take five separate measurements to derive this graph. To the right is shown a composite of the midrange and HF units' quasi-anechoic responses on the tweeter axis at 45", spliced at 976Hz to the midrange unit's nearfield response. In the center is the woofer's quasi-anechoic response at 45" on the tweeter axis, spliced at 510Hz to its nearfield response, and on the left is the nearfield response of the rear-facing reflex port, adjusted in level in the ratio of its diameter to that of the woofer. This peaks sharply just below 80Hzthe port can be heard to have considerable output in this regionwhich is a little higher in frequency than the woofer's minimum-motion point.
The Velodyne's woofer output rises between 100Hz and 300Hz, with then a considerable overlap between it and the midrange unit. The latter peaks sharply at 500Hz, with a smooth but slightly down-tilted response all the way to above 20kHz, where the tweeter resonance results in a very high peak at 26kHz.
Fig.3 splices the complex summagnitude and phaseof the nearfield woofer and port responses, to the overall speaker's quasi-anechoic response on the optimum tweeter axis at 45", averaged over a 30° horizontal window. Two things can be seen immediately: First, the rollout below the port tuning frequency is very steep, due both to the woofer and port outputs being out of phase and the presence of the series capacitor in the woofer feed. The response is down almost 20dB by 42Hz (the frequency of the open E-string of the double bass and Fender bass). In addition, because the rollout is so fast, there will be virtually no reinforcement of the Velodyne's bass by room boundary effects. Second, the entire woofer bandpass is approximately 5dB too low in level compared with the midrange unit's average level. This is why the speaker sounds too lean: it just doesn't have enough output in the lower midrange (see Sidebar).
Higher in frequency in fig.3, the tweeter seems shelved down a little, though a slight peak at 10kHz can be seen. Fig.4 reveals how the speaker's response changes with changes in listening height. (Only the changes are shown, which is why the reference response on the tweeter axis appears as a straight line. This does not mean that the speaker's response is flat.) If you sit above the midrange axis, a deep suckout appears in the region where the woofer and midrange unit overlap. If you sit much below the tweeter axis, a suckout appears in the region where the midrange unit and tweeter overlap. With the DF-661 on 24" stands, you get the best blend between the drive-units by sitting with your ears between 33" and 38" from the ground.
Laterally (fig.5), using a drive-unit nearly 5" in radiating diameter up to 5.5kHz results in significant beaming in the midrange unit's top octave. This is revealed by the deep off-axis suckout at the 4.3kHz cursor position. (Ignore the LF wrinkles in this graph on the rear side of the reference response; this was due to my inadvertently including a room reflection in the impulse responses used to derive this half of the image.) While this beaming will not in itself be problematic if you sit on-axis, the discontinuity between the midrange-unit's top-octave dispersion and that of the tweeter in its bottom octave most definitely is a problem. See how the loudspeaker's output above 5kHz is maintained to extreme off-axis angles, due to the tweeter's radiating diameter being smaller than the wavelengths of the sound it is reproducing. The DF-661's distortion may be low, but this abrupt change from narrow beam to wide dispersion will add a mid-treble glare to the sound. (Surprisingly, during his Santa Fe visit David Hall mentioned to Larry Archibald and me that he hadn't looked at the DF-661's off-axis behavior.)
Carrying out a spatially averaged 1/3-octave spectrum analysis for left and right speakers individually in my listening room gave the curve shown in fig.6. (I average six measurements at each of 10 separate microphone positions for left and right speakers individually, giving a total of 120 original spectra, which are then averaged to give a curve which, for my room at least, has proved to give a nice correlation with a loudspeaker's perceived balance. I use an Audio Control Industrial SA-3050A spectrum analyzer with its own microphone, which acts as a check on the MLSSA measurements made with the B&K mike. I also used the Goldline DSP-30 automated spectrum analyzer, which is currently under review.) Again, the prominent upper midrange and the shelved-down lower midrange and treble can be seen, as can the suppressed low bass.
Turning to the time domain, the DF-661's impulse response on the tweeter axis is shown in fig.7. The ultrasonic ringing from the tweeter is clearly visible, as is a mild ringing much lower in frequency, with a peak just after the 5ms mark. This has a period of just under 2ms, implying that it is associated with the peak at 600Hz in the midrange unit's output noted in figs.2 and 3. It can be seen more clearly in the step response (fig.8), which also shows that the positive-polarity tweeter leads the negative-polarity midrange unit in time. (The woofer is connected in positive polarity.)
The cumulative spectral-decay, or waterfall, plot calculated from the impulse response is shown in fig.9. A residual resonant mode can be seen at the cursor position, 3.5kHz, which, in conjunction with the step in frequency response at that frequency, might tie in with the slight nasality noted by JGH. The broad ridge parallel to the time axis around 600Hz is again due to the midrange unit misbehaving at the bottom of its passband, but, other than that and the tweeter resonance at 26kHz, the DF-661's decay is quite clean.
Finally, fig.10 shows the waterfall plot of the output of a simple PVDF accelerometer fastened to the center of the enclosure top panel while the loudspeaker was being driven by high-level MLS noise with a bandwidth of 2kHz. (This is a more consistent, analytical version of the classic knuckle-rap test.) As implied by the impedance plot, a resonant mode is present at the musically significant frequency of 440Hz (the frequency of A above middle C). I didn't hear any problems in this region, however. Repeating the test with the accelerometer on the center of the side panel revealed a mode lower in level at the port tuning frequency of 78Hz, but I doubt that this will be of subjective significance.John Atkinson