Vandersteen 2C loudspeaker Measurements

Sidebar 1: Measurements

Each speaker was submitted to the following measurement schedule: The voltage sensitivity was assessed with 1/3-octave pink noise centered on 1kHz, while the change of impedance with frequency was measured using spot tones (footnote 1). The nearfield low-frequency response of each speaker was assessed with a sinewave sweep to get an idea of the true bass extension relative to the level at 100Hz.

The frequency response of each speaker in the listening area was measured using pink noise and an Audio Control Industrial SA-3050A 1/3-octave spectrum analyzer. Nine sets of six averaged measurements were taken independently for left and right loudspeakers at a distance of just over 2m in a window 72" wide and varying from 27" to 45" high. The response shown in each review is the average of these measurements, weighted slightly toward the sound heard at the listening position. This spatial averaging is intended to minimize the effect of low-frequency room standing-wave problems (below 500Hz or so) on the measurement, and gives a response curve that has proved to predict reasonably well what is heard; it also gives an idea of the off-axis behavior of the speaker under test.

In addition, for these tests, I captured the impulse response of the speakers using a PC-compatible, 8-bit Heath/Zenith digital storage oscilloscope. This enables up to 250 separate "snapshots" to be averaged, which (as long as the time relationship between the trigger pulse and the captured pulse is constant, within one sampling period) usefully increases the S/N ratio. The 'scope communicates with the computer via its RS232 port, the computer then controlling the 'scope's settings via "soft" keys. The waveform can be stored on disk as a 512-point ASCII file, which thus enables the user to carry out various mathematical operations on the data. Accordingly, I wrote an FFT program in Microsoft's Quickbasic 4.5 language (footnote 2), which takes about two minutes to calculate the equivalent anechoic response of the speaker from the impulse data file (footnote 3).

The measurement window was 10ms, it being assumed that the impulse has died away to zero by this time (footnote 4), and I arranged the position of the speakers so that any reflections of the pulse from room boundaries would arrive after this period. The trigger to the 'scope was a delayed twin of the analytical pulse, the delay time approximately arranged to coincide with the transit time of the sound from speaker to measuring microphone. In addition, use of a 10ms window results in a sampling frequency of just under 51.3kHz, minimizing audio-band aliasing.

I have included these responses in the measurement sections of the reviews, but note that I have only plotted the range—from 200Hz to 10kHz—where the calculated responses have proved to be repeatable within 0.25dB or so. (To cover the entire audio spectrum, a unidirectional square pulse needs to be of no more than approximately 20µs in length; the pulse I was using, shown in fig.1, was unfortunately 55µs long, curtailing the accuracy of the measurement in the top audio octave. Note also that the pulse used has inverted polarity: this is due to the inverting line stage of the Conrad-Johnson PV9 preamplifier used for these tests. A 20µs pulse generator is on the way, as is a math coprocessor chip (footnote 5).

Fig.1 Vandersteen 2ci, 55µs test pulse used for the measurements.

Note also that a 10ms window means that you are stuck with 100Hz resolution: midrange peaks and dips due to resonances and interference that are narrower than this will be lost. On the other hand, going to a 20ms window, say, if the room were large enough and the speaker far enough above the floor, would halve the measurement bandwidth with only 512 data points. A case of swings and roundabouts, I guess.)

Fig.2 shows the 2Ci's modulus of impedance with the contour controls set flat [see also the impedance graph in our 1993 of the slightly different 2ce, which also shows electrical phase.—JA]. The fundamental box resonance can be seen to lie at 88Hz and is well-damped, while the Acoustic Coupler tuning is an octave lower at 45Hz. Above the bass region, the impedance averages 8 ohms, though it does drop lower in the top audio octave and above. Coupled with its measured sensitivity of 90dB/W/m at 1kHz, the 2Ci should be pretty easy to drive, though tube amplifiers would probably work best from their 4 ohm output taps.

Fig.2 Vandersteen 2ci, electrical impedance (2 ohms/vertical div.).

The spatially averaged in-room response, again taken with the tone controls flat, can be seen in fig.3. Relatively smooth, it shows a gradually sloping-down trend from the bass to the treble, relieved by a touch of looseness in the bass coupler region, some liveliness in the upper midrange, and a slight peakiness in the tweeter output around 13kHz or so. This latter characteristic is much more noticeable on the speaker axis; with the speakers firing straight ahead so that the listener sits significantly off-axis, it is hardly audible. Bass extension in-room is excellent for the size of the speaker, there being useful output down to the 32Hz 1/3-octave band. Vandersteen claims a wide horizontal dispersion for the 2Ci, and this was in general confirmed up to about 5kHz (for up to ±30 degrees off-axis) by the individual responses taken to derive the averaged result shown in fig.3. Above that frequency, the tweeter become increasingly directional. Listening off-axis can therefore be used to best optimize the tweeter balance without isolating the top HF octave, which is presumably why Vandersteen recommends firing the speakers straight ahead.

Fig.3 Vandersteen 2ci, spatially averaged, 1/3-octave response in JA's listening room.

The effect of the two contour controls can be seen in fig.4. The top trace is the response at 1m taken with both controls at their +2dB maximum; the bottom with both set at minimum (approximately -3dB). The midrange control offers a 6dB swing centered on the 1-2kHz octave, while the treble control swings the top audio octave by, again, 6dB. There is broad overlap in the 2.5-8kHz region, allowing the user to get a wide range of tonal variation to best optimize the speaker's treble balance to the listening room.

Fig.4 Vandersteen 2ci, effect of tone controls: mid & HF at maximum (top), mid & HF at minimum (bottom).

I found a touch of lift in the high treble (tweeter set to +1dB) to be necessary to counteract the speaker's slight lack of off-axis air; this has to be balanced, however, against exaggerating the excess of HF energy above 10kHz. Though this might then be thought to leave the midrange a little depressed, I nevertheless found -1dB on the midrange control to give the optimal balance between the need for sufficient high-midrange energy and not to unduly emphasize the speaker's uneven response in the same region. This will be different for every listening environment and system, however.

Fig.5 is a composite response, with the individual nearfield responses of the woofer and Acoustic Coupler superimposed on the response measured on the tweeter axis at 1m of the tweeter and midrange unit with the controls set as described above. The slightly over-damped woofer response will marry with the sub-bass radiator response to give an almost perfectly aligned bass alignment extending down to 28Hz, -6dB. This is exceptional extension for what is a relatively affordable loudspeaker. Despite the woofer/midrange crossover frequency being specified as 450Hz, it can be seen that there is a broad overlap for an entire octave above that frequency due to the slow roll-off rates chosen. In addition, the woofer can be seen to get peaky on-axis above 1kHz; although the drive-unit will be very directional in this region, minimizing its effect on the listening-axis response, it could well have contributed to the low-treble coloration noted in the listening tests. The on-axis bump in the tweeter's output can again be seen, as can a touch of unevenness in the midrange unit's output centered on 3kHz. This could be heard on pink noise as a mild "sshh" emphasis.

Fig.5 Vandersteen 2ci, active bass coupler, woofer, & midrange/tweeter responses.

The design decision to use slow-rollout, first-order, 6dB/octave crossover slopes and to accept the broad overlap between adjacent drive-units is, of course, due to the fact that, when the acoustic centers of the drivers are physically aligned (as they supposedly are in the 2Ci), and when those drivers are connected in-phase, the result should be a time-coherent wavefront re-creation, with overtones arriving at the listener's ear at the same time as the fundamental.

To investigate whether this was the case, I drove the speakers with a 500Hz squarewave and investigated the shape of the waveform on the listening axis; ie, about 15 degrees off the lateral axis and with the mike approximately at tweeter height. Now with a conventional high-order crossover design with non-time-coincident drive-units, such as the Celestion SL600Si, it is impossible to get a squarewave reproduced as anything like a squarewave, due to the fact that the harmonics of the tone arrive at the microphone position at different times from the fundamental depending on their frequency.

The best result I could find for the Vandersteen 2Ci at 1m (which is a little close) is shown in fig.6. As expected from the broad overlap between adjacent drive-units, this was extremely sensitive to vertical changes in position. But, it must be noted, it is quite a good squarewave, with relatively sharp leading edges. The overshoot, which is presumably due to a damped resonance in the tweeter, can be seen have a period just over one small division, corresponding to a frequency of around 7.5kHz. The fact that the individual outputs from the drive-units are still not quite coincident in time at the 1m microphone position is shown by the fact that the output of the fundamental from the woofer, with its slow risetime, doesn't arrive until some 350µs after the tweeter.

Fig.6 Vandersteen 2ci, 500Hz squarewave response (5ms time window).

Figs.7 and 8 show the speaker's impulse response, directly in front of the tweeter and on the 15 degrees listening-position axis, respectively. (The time window shown is 5ms, the same as in fig.1, which shows the 55µs test pulse.) I find interpretation of these raw impulse responses to be fraught with difficulty. Therefore, apart from noting that: the initial polarity of the pulse is correct (remember that the stimulus was inverted by the PV9 preamp); the generally time-coherent nature of the speaker is shown by the fact that a large proportion of the energy arrives within the first 200µs; and that there is some generally lively resonant behavior in the low treble, as shown by the ringing), I will quickly move on. Figs.9 and 10 are the anechoic frequency responses derived from FFT analysis of the impulse responses taken this time in a 10ms time window. (Remember that I am not confident of the response's repeatability below 200Hz and above 10kHz, hence the limited range shown.)

Fig.7 Vandersteen 2ci, impulse response on tweeter axis (5ms time window).

Fig.8 Vandersteen 2ci, impulse response on listening axis (5ms time window).

Fig.9, the on-axis response, shows a shape in the upper-midrange and treble similar to that in fig.5. Some peakiness around 1kHz, with a slight excess energy in a broad region centered on 3kHz, is coupled with a lack of energy in the crossover region to the tweeter. The sharp dips in the response centered on 1400Hz, 5250Hz, and 6400Hz are due to destructive interference, either between two drivers carrying the same signal or between the direct sound and its reflection from some nearby part of the speaker enclosure. I very much doubt that they will be audible, and in any case their exact position and depth is highly dependent on the microphone and thus the listening height. The two sharp peaks at 7200Hz and 7800Hz may correlate with the overshoot and slight ring seen on the 500Hz squarewave.

Fig.9 Vandersteen 2ci, anechoic response on tweeter axis.

Fig.10 was derived from the impulse response taken at a distance of 1m 15 degrees off the horizontal axis—ie, on the listening axis—with the tone controls set to -1dB (midrange) and +1dB (tweeter), which is how I did most of my listening. The response can be seen to be much smoother through the treble, though there is now a lack of energy in the lower crossover region, accentuated by two deep, narrow suckouts which, again, I assume to be interference phenomena.

Fig.10 Vandersteen 2ci, anechoic response on optimum listening axis.

If you feel that these plots show severe departures from a flat response, I must point out that that is not really the case. Such phenomena are very dependent upon the listening axis and are to be expected when a speaker has first-order crossover slopes. They represent the inevitable tradeoff the designer has to endure to arrange for time coherence (see the results for the Celestion SL600Si for a contrasting design choice). Their audibility is minimal; it's the broad trend that's important. [With 14 years of hindsight, these two frequency responses are worse than they should be, but for reasons that are not immediately apparent from the archived measurement data files. For a more thorough examination of the Vandersteen's quasi-anechoic response and off-axis behavior, see my 1993 measurements.—JA] In fact, I went into so much depth in measuring the Vandersteen 2Ci because it sounded so good that I wanted to see if there were any obvious indicators why.—John Atkinson

Footnote 1: It has been worrying me for some time that while my impedance plots are the right shape, the absolute values differ slightly from the manufacturer's specification and from those published in other magazines. In particular, speakers reach a slightly higher impedance value at their LF resonance, though this itself is not changed in frequency. After considerable head-scratching, the acquisition of some even-higher-precision calibration resistors than I used before, and the purchase of a new Fluke 87 true-RMS multimeter, I can only conclude that it must be the effect of the reduced air pressure here in 7000'-high Santa Fe.—John Atkinson

Footnote 2: To those laymen like myself who find themselves interested in FFT techniques, I found Ronald Bracewell's The Fast Fourier Transform and Its Applications (McGraw-Hill) to give a usefully thorough exposition of the theory. The appendix to Williams's and Taylor's Electronic Filter Design Handbook (Second Edition, McGraw-Hill) was also of help.—John Atkinson

Footnote 3: So much toil to end up with the DIY equivalent of Julian Hirsch's IQS FFT analyzer! I have always believed, however, that the heuristic approach is the best way to gain an understanding of how test equipment works. You can then appreciate its limitations when faced with the real-world task of coping with the foibles of music reproduction systems.—John Atkinson

Footnote 4: This may be an unjustified assumption. I will be trying various shaped windows, such as the Hamming, which imposes a raised cosine amplitude function on the windowed data, in future tests.—John Atkinson

Footnote 5: In the fall of 1989, the magazine purchased the then-new DRA Labs MLSSA system for loudspeaker measurements, which I use to this day, with calibrated B&K, Mitey Mike, and Earthworks microphones.—John Atkinson

116 West Fourth Street
Hanford, CA 93230
(559) 582-0324