Celestion SL600si loudspeaker & DLP600 digital equalizer 1992 Measurements

Sidebar 5: 1992 Measurements

Before I examined the measured effect of the DLP600, I ran the SL600Si through Stereophile's almost standardized test regime. Fig.1 shows the manner in which the speaker's electrical impedance magnitude and phase vary with frequency. The single peak of 30 ohms at 63Hz indicates the sealed-box tuning frequency. The speaker is an easy load, apart from the drop below 6 ohms above 4500Hz. Though this dips to 2.7 ohms at 34kHz, there is insufficient musical energy in this range to lead to any problems. Note the sharp impedance peak at 21kHz; this is due to the individually tuned notch filter Celestion uses to kill the tweeter's primary resonance, which is a little close to the audio band for comfort compared with the lighter aluminum, magnesium, or titanium domes used in more recent metal-dome drive-units.

Fig.1 Celestion SL600si, electrical impedance (solid) and phase (dashed). (2 ohms/vertical div.)

Fig.2 shows the individual responses of the woofer and the tweeter measured with DRA Labs' MLSSA system on the tweeter's axis at a distance of 45". (The response of the B&K measuring microphone has been subtracted from all the curves shown in this report other than the cumulative spectral-decay plots.) The crossover frequency appears to be in the region of 2.5kHz. Though both drive-units roll out in a generally well-behaved manner, the woofer features a little peakiness at the top of its passband, perhaps due to a cone-termination problem. The tweeter's balance trend is respectably flat over its first two octaves, but rolls off somewhat in the top octave. The response can be seen to start to rise at 20kHz, but then drops due to the action of the notch filter. Overall, the tweeter seems rather shelved down in comparison to the woofer, this due, I assume, to the relatively high mass of the copper dome slugging the unit's sensitivity.

Fig.2 Celestion SL600si, acoustic crossover on tweeter axis, corrected for microphone response.

How the two drive-units' outputs integrate acoustically can be seen on the right of fig.3, which shows the SL600Si's response at 45" on the tweeter axis, averaged across a 30 degrees horizontal window to minimize the effect of position-dependent interference effects. Again the tweeter's slight shelving-down can be seen, as well as a lack of energy in the crossover region. This suggests that this axis is not quite the optimal one on which to listen to and measure the speaker, something which will be confirmed later.

Fig.3 Celestion SL600si, anechoic response on tweeter axis at 50", averaged across 30 degrees horizontal window and corrected for microphone response, with the nearfield woofer response plotted below 300Hz.

To the left of fig.3 is plotted the woofer's bass output, taken with the microphone almost touching where the dustcap would have been had the latter not been inverted. The measured -6dB point, referenced to the maximum level, coincided almost exactly with the specification at 59Hz. Though this may sound not particularly low, the "infinite baffle" alignment gives a relatively slow rate of rolloff, which, combined with the typical boost in the low bass due to the room, will give respectable extension down to around the lower notes of the 4-string double bass or bass guitar.

The manner in which the speaker's balance changes as the listener moves to its side is shown in fig.4. (As this only shows the differences, the on-axis response is depicted as a straight line.) The top octave smoothly depresses with increasing off-axis angle, as does the sound at the top of the woofer's passband. This does leave the mid-treble a little boosted in comparison, however, which might mean that sidewalls that are too close, too reflective, or both, could make the perceived balance a tad bright. For a listener to get the full measure of the extreme highs, however, fig.4 does suggest that the speaker should be toed-in to the listening position.

Fig.4 Celestion SL600si, lateral response family at 50", normalized to response on tweeter axis, from back to front: differences in response 90 degrees-5 degrees off-axis, reference response, differences in response 5 degrees-90 degrees off-axis.

Looking at the way the speaker's balance changes in the vertical plane (fig.5) confirms that the tweeter axis is a little too low to get the smoothest transition between the drive-units. Again, only the changes in measured response are shown, which is why the tweeter-axis response appears to be a straight line with frequency. Fig.5 appears to indicate that the flattest treble will be obtained with the listener level with or just above the cabinet top, suggesting that the 18" Celestion Si stands will be optimal. Sit so you are on or below the tweeter, and the suckout at crossover becomes progressively deeper.

Fig.5 Celestion SL600si, vertical response family at 50", normalized to response on tweeter axis, from back to front: differences in response 10 degrees-5 degrees above axis, reference response, differences in response 5 degrees-15 degrees below axis.

So what does the DLP600 do to the measured performance? As a digital equalizer works by synthesizing the time-domain behavior of the desired tone-shaping network, I first looked at the DLP600's impulse response. I did this in two ways. The first was to take the MLSSA's analog MLS stimulus, convert it to 16-bit digital with the excellent Manley Reference A/D converter, and feed the Manley's S/PDIF output to the DLP600, this in turn feeding the data input of an Audio Alchemy Digital Decoding Engine. The DDE's analog output fed the analog input of the MLSSA system.

The second procedure was conceptually more simple. I loaded a single positive impulse waveform into the Audio Precision System One Dual Domain's waveform store, fed this in digital form into the DLP600, and fed the DLP600's data output back into the Audio Precision. The result of the first technique is shown in fig.6; the impulse response measured the second way was identical. The reason I ultimately used the MLSSA technique was that I wanted to have the DLP's impulse response and its frequency-domain equivalent available to the MLSSA system so that I could double-check my findings. Note both the lower-frequency pre-ringing in fig.6, which is to compensate for the woofer being further back in time than the tweeter, and the overall time delay introduced by the whole ADC/DSP/DAC process, about 1.4ms to the beginning of the low-frequency oscillation—equivalent to moving the speaker 19" farther away.

Fig.6 Celestion DLP600, impulse response via Manley ADC and Audio Alchemy DDE DAC (5ms time window, 30kHz bandwidth).

To examine what the DLP does in the frequency domain can be achieved by transforming the impulse response in fig.6, but this process will include the contribution of the Manley ADC and Audio Alchemy DAC. However, the MLSSA system allows you to subtract one response from another, so I hooked up the ADC and DAC back to back, measured their impulse response, transformed that to the frequency domain, then subtracted it from the DLP's transformed frequency response. (This is what the word "equalized" means on the description line of the MLSSA graph: that a reference response has been subtracted from the indicated curves.) The result can be seen in fig.7, which shows both the amplitude response changes introduced by the DLP and the change in phase response. The former may seem severe, but note the 0.5dB/division vertical scale. There is actually a mild boost, 3dB, in the crossover region, centered on 2670Hz, with then some minor peaks and dips in the next two octaves, followed by a larger 4.7dB peak centered on 15.1kHz.

Fig.7 Celestion DLP600, frequency response (top, 1dB/vertical div.) and phase response (bottom, 90 degrees/vertical div.)

How do these changes in frequency and phase response affect the SL600Si's time- and frequency-domain behavior? Fig.8 shows the uncorrected impulse response of the loudspeaker, taken on the tweeter axis at 45" with a 30kHz-bandwidth drive signal. Note the sharp spike from the tweeter, under which lies the lazier, opposite polarity and time-delayed output of the woofer. There is also some residual ultrasonic ringing from the tweeter, not all of which has been removed by the notch filter. To get an idea of what the impulse response of a perfect linear-phase loudspeaker should look like, fig.9 shows the impulse response of a steep, audio-bandwidth low-pass filter. (This is flat across the band, then rolls off abruptly above 20kHz.)

Fig.8 Celestion SL600si, impulse response on tweeter axis at 45" (5ms time window, 30kHz bandwidth).

Fig.9 Digital low-pass FIR filter, impulse response (5ms time window, 30kHz bandwidth).