Stand Loudspeaker Reviews

Sidebar 3: Measurements

I used DRA Labs' MLSSA system and a calibrated DPA 4006 microphone to measure the Dragonfire Mini Dragon Satellite's frequency response in the farfield, and an Earthworks QTC-40 mike for the nearfield responses of that and the SB-8P powered subwoofer. I measured the performance of the Dragonfire MD-4 power amplifier with my Audio Precision SYS2722 system (see the January 2008 "As We See It"). Because the MD-4 uses a class-D output stage, around 320mV of ultrasonic switching noise was present on its outputs. As this noise might overload the SYS2722's input circuitry, I used the Audio Precision auxiliary AUX-0025 passive high-power filter, which eliminates noise above 200kHz, for the amplifier testing.

Although Dragonfire recommends that purchasers of the Mini Dragon Satellites use them with their MD-4 amplifier, which is how they were auditioned by Jason Victor Serinus, the manual states "As an ultra-efficient purely resistive load, the Mini Dragons require only a few watts of power to reach normal listening levels and may be driven by a wide variety of high-end amplifiers." I therefore performed the first round of testing using one of the Mini Dragon speakers driven by the Krell KSA-50 amplifier I routinely use for Stereophile's speaker testing.

The Mini Dragon speaker is indeed close to being a "purely resistive load," though at 8 ohms (fig.1, solid trace), our sample's impedance was nowhere close to the specified 20 ohms. On the other hand, my estimate of the Mini Dragon's basic sensitivity of 103.5dB(B)/2.83V/m was even higher than the specification (in the upper midrange) of 100dB/w/m. However, as the Mini Dragon speaker's frequency response when driven by the Krell reveals, this extraordinarily high sensitivity is achieved by the loudspeaker having a rise in its response of more than 15dB in the top octaves (fig.2). This behavior might be worse than it looks, due to the Mini Dragon speaker becoming extremely directional in the same region (see later). If you use this loudspeaker with a regular amplifier, you can drastically reduce the amount of high treble by experimenting with the toe-in.

Nevertheless, I feel fig.2 implies that it is mandatory for the Mini Dragon speakers to be used with Dragonfire's MD-4 amplifier, which uses digital signal processing both to correct the speaker's frequency response and to provide the crossover for it to be used with a subwoofer, like Dragonfire's own SB-8P. The MD-4 has two sets of loudspeaker outputs, each on a Neutrik Speakon jack. (The CH1–2 jack provides a pair of low-pass filtered signals to feed a subwoofer; the CH3–4 jack is used to drive the satellites.) There are three sets of inputs: balanced analog, coaxial S/PDIF, and USB. Apple's USB Prober utility identified the amplifier as the "xCORE USB Audio 2.0" from "XMOS" and confirmed that the USB input operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI app indicated that the MD-4 operated at all PCM sample rates up to 384kHz via USB, with integer bit depths of 16 and 24.

The balanced analog XLR jacks had an appropriately high input impedance of 10k ohms from 20Hz to 20kHz but a low voltage gain into 8 ohms of 6.34dB at 1kHz (see later). The output impedance was very low at 200Hz and 1kHz, at 0.09 ohms, but rose to 0.3 ohms at 20kHz. With the Audio Precision filter, the unweighted, wideband S/N ratio, ref. 2.83V into 8 ohms and measured with the analog input shorted to ground, was a modest 72.7dB. An A-weighting filter improved the ratio to 81.8dB. A 1kHz digital signal at 0dBFS resulted in an output level of 6.01V into 8 ohms, at the CH3–4 jack. This level is equivalent to 4.5W into 8 ohms, which is therefore the maximum power at 1kHz that the amplifier will be able to deliver to the Mini Dragon speakers (see later).

The MD-4's impulse response with 44.1kHz digital data (fig.3), measured at the loudspeaker output, indicates that the amplifier inverts absolute polarity. (I had connected the supplied loudspeaker cables as recommended in the manual, with the white lead positive and the black lead negative. The analog inputs also invert polarity using the dedicated Dragonfire Speakon–mini-XLR cables.) Fig.3 shows that the MD-4's reconstruction filter is a conventional linear-phase type, with time-symmetrical ringing to either side of the single sample at 0dBFS. However, the complex shape of the impulse is due to the amplifier's digital signal processing. The effect of the DSP can be seen with 44.1kHz-sampled white noise (fig.4 red and magenta traces). While the MD-4's response rolls off extremely sharply above 20kHz, reaching full stop-band suppression at 25kHz, a little above half the sample rate (vertical green line), a considerable amount of frequency shaping can be seen in the audioband. The aliased image at 25kHz of a full-scale tone at 19.1kHz (blue and cyan traces) is completely suppressed, and the harmonics of this tone are all very low in level.

With the MD-4's aggressive use of DSP and a class-D output stage, I wasn't surprised that when I changed the bit depth of the incoming data from 16 to 24 with a dithered tone at –90dBFS, the noise floor didn't drop in level (fig.5), even though the Audio Precision low-pass filter was in-circuit. With an undithered tone at –90.31dBFS (fig.6), the three DC voltage levels were obscured by high-frequency random noise. When I tested the MD-4's rejection of word-clock jitter with 16-bit data, the odd-order harmonics of the LSB-level, low-frequency squarewave, which lie between –124dBFS and –128dBFS, were all buried beneath the analog noise floor (fig.7). These results imply that the amplifier's practical resolution is limited to around 14 bits.

The blue and red traces in fig.8 show the frequency response of the MD-4's satellite outputs and show in more detail the response shaping seen in fig.4. The peak at 16.375kHz seen in fig.2 is notched out and the upward slope with frequency of the trace in fig.2 is compensated for. There is a rise in response in the lower midrange to compensate for the small panel's dipole cancellation, and the output then drops rapidly below 200Hz. The green and gray traces in this graph show the output of the MD-4's subwoofer output. It is aggressively rolled off above 200Hz, which is why Dragonfire instructs users of the matching SB-8P subwoofer to set its rear-panel switch to "LFE." An LFE signal is already high-pass–filtered, so setting the SB-8P to "Sub," which might be thought intuitive, is incorrect as this will apply two filters. Note that several small peaks and dips overlay the traces in fig.8. I assume these are due to Dragonfire's adjustments of the MD-4's parametric equalizers for JVS's placement of the satellites and subwoofer.

Fig.9 shows the MD-4's reproduction of a 1kHz squarewave. It is overlaid with the linear-phase ringing of the amplifier's antialiasing filter. Dragonfire rates the MD-4's maximum power as 245Wpc into 8 ohms (23.9dBW). Remember, however, that I calculated the amplifier's maximum power with a full-scale digital signal at 1kHz as just 4.5W (6.5dBW). With both analog input channels driven at 1kHz and with clipping defined as when the THD+noise in the output reaches 1%, I measured the apparent clipping power into 8 ohms as 4.8Wpc (6.8dBW, fig.10). However, repeating the measurements with a 300Hz signal gave an apparent clipping power of 35W into 8 ohms (15.45dBW, fig.11).

A moment's reflection made it obvious that these two graphs are not showing the powers at these two frequencies at which the MD-4 clips. Instead, if you look at the response shaping in fig.8, it becomes clear that the 1% THD+N point in both graphs is reached when the input signal is 3V. This is therefore the input voltage at which the MD-4's analog/digital converter overloads, not when the amplifier runs out of power. Some D/A processors and disc players have balanced outputs that offer maximum levels greater than 3V, so it will probably be best to use the MD-4's digital inputs to avoid distortion with those sources. Overall, the MD-4 offers commendably low levels of distortion, which is primarily third-harmonic in nature (fig.12).

Returning to the Mini Dragon Satellite, with the speaker now driven by the MD-4, the equalization has reduced the sensitivity to around 83dB(B). Even with the limited amount of power on tap, the Mini Dragon should provide sufficiently satisfying listening levels on a desktop, though, as JVS noted, "it is not a rock-the-joint party animal whose volume can be cranked up anywhere near floor-shaking levels."

The Dragonfire's farfield response driven by the MD-4, averaged across a 30° horizontal window centered on the central axis, is shown as the black trace above 300Hz in fig.13. It is flat on average up to 10kHz, with various small peaks and dips evident. The nearfield response is plotted below 400Hz; it rolls off rapidly below 200Hz. The nearfield output of the Dragonfire SB-8P subwoofer's drive-unit is shown as the green trace in fig.13, that of the passive radiator as the blue trace, and their complex sum as the red trace. The subwoofer was set to "Sub" and the crossover to its highest frequency, "160Hz," for these measurements, as I wanted to examine the SB-8P's intrinsic behavior without the MD-4's equalization being applied. The subwoofer usefully extends the Mini Dragon's response into the midbass region, below which it quickly rolls out.

When I plotted the Mini Dragon speaker's horizontal dispersion (fig.14), it became apparent that the response shown in fig.11 does not accurately indicate the Dragonfire satellite's behavior above 10kHz. I have been averaging a loudspeaker's farfield output across a 30° horizontal window for a quarter-century, because the resultant response correlates well with the direct sound balance perceived by the listener. However, the Mini Dragon becomes increasingly directional above 3kHz and the output 15° off-axis above 10kHz is well down in level, which affected the trace shown in fig.13. This graph also reveals the panel's typical figure-eight behavior, with the output 90° to the side down by 30dB compared to the on-axis level. In the vertical plane (fig.15), as the Mini Dragon Satellite is taller than it is wide, the dispersion is even more limited. When it is driven by the MD-4, the Mini Dragon speaker really does need to be listened to on-axis if the balance is not to become too mellow.

In the time domain, the Mini Dragon's step response (fig.16) indicates that the loudspeaker driven by the MD-4 inverts acoustic polarity. (With it driven by the Krell amplifier, the speaker preserved polarity—ie, was noninverting. In addition, the step arrived at the microphone 0.5ms earlier with the Krell, suggesting that this is the latency of the MD-4's DSP.) The decay of the loudspeaker's step is overlaid with some undulations with a period around 1ms. These undulations correlate with a ridge of delayed energy at 1kHz in the Mini Dragon's cumulative spectral-decay plot (fig.17), and there is a stronger ridge present at 1687Hz. Both these ridges coincide with small peaks in the loudspeaker's on-axis response. Though JVS didn't comment on noticeable coloration—he found the speaker's midrange to sound very natural—I would have expected this behavior to add a slight nasality to the Mini Dragon's sound. The treble region in this graph is superbly clean, however.

Finally, I didn't measure the performance of the complete system—Mini Dragons driven by the MD-4 with room correction provided by Dirac and the miniDSP SHD Studio preamplifier—as the miniDSP's corrections will only be relevant to JVS's listening position. However, Dirac's target response for that position is shown in fig.18.

Overall and within the necessary compromises associated with its unique design, the Dragonfire system offers excellent measured performance. However, it will only give of its best when listened to on a desktop.—John Atkinson