Illusion Audio Carbon C12

Overall summary

In this sample and under these test conditions, the Illusion Audio Carbon C12 shows a fairly unusual split between the 1 V baseline and the 11 V high-level sweep. At 1 V, the distortion curve is jagged from 20 to 120 Hz, which may indicate some low-level noise or non-linear behavior that is more exposed at very low drive levels and becomes increasingly masked by the fundamental output as playback level rises. While that cannot be confirmed from these measurements alone, it is one possible explanation for why the relative distortion behavior appears less problematic during the higher-level sweep, with THD exceeding the visible 10 percent window at the very bottom and multiple peaks through the lower and mid-bass range. H3 distortion is strongest at the lowest frequencies around the low 20 Hz range, while H2 distortion becomes the main contributor through much of the 25 to 120 Hz range. The response curve itself is usable through the subwoofer band, but the distortion behavior is not especially quiet at the 1 V baseline.

At 11 V, the relative distortion percentage is lower through much of the same range, which means the high-level sweep does not simply look like the 1 V curve scaled upward. One possible explanation is that very low-level mechanical or airflow-related noise becomes more visible in the baseline measurement, although the data cannot confirm that with certainty. Regardless of the cause, the cleaner-looking high-voltage sweep is not unusual in relative terms and is generally a more favorable outcome than the reverse scenario, where distortion rises sharply as output increases. THD is highest at ~8 percent near 20 Hz, then settles lower at around 2-3 percent through most of 30 to 120 Hz range, with a broad rise up to 5 percent around roughly 80 Hz. H2 distortion is the dominant harmonic through most of the high-level sweep, while H3 distortion is present but not the main feature in the primary subwoofer range. That makes the high-level result look more stable than the 1 V baseline would suggest.

The LSI data explains part of this behavior. BL 70 percent lands at 12.82 mm one way, CMS 50 percent lands at 13.11 mm one way, and the Le 17 percent criterion lands slightly earlier at 12.03 mm one way. The BL curve is broad and usable for the stated shallow-driver format, but it is not perfectly flat, and the stiffness plot shows clear suspension asymmetry toward the end of stroke. Le(x) also varies enough with position to set the earliest standardized limit, although Le(i) is stable with current.

For sealed use, the manufacturer’s 0.75 ft³ recommendation calculates to a somewhat high Qtc of 0.903 on this sample. A 0.707 Qtc requires 1.54 ft³ based on the large signal cold parameters. This is not a strong infinite baffle candidate because it reaches its linearity limits near the published xmax range, and removing the sealed air spring would reduce mechanical safety margin.

Manufacturer's suggested use case

Illusion Audio positions the Carbon C12 as a shallow 12 inch subwoofer using a woven carbon fiber cone, rubber surround, cast aluminum basket, and 51 mm voice coil that is good for moderately small sealed enclosures. The manual lists a nominal 4 ohm impedance, 400 watt nominal power rating, 800 watt maximum power rating, 87.8 dB sensitivity at 2.83 V/1 m, 12 mm xmax, and 25 mm Xsus. Published mounting depth is 83 mm. The manufacturer recommends a sealed enclosure of 0.75 ft³, with a published -3 dB point of 42 Hz.

Our suggested use case

Based on the data, this sample is best kept in sealed enclosures where excursion is mechanically controlled. The manufacturer’s recommended 0.75 ft³ enclosure results in a higher-Q 0.903 Qtc on this sample, while a larger 1.54 ft³ enclosure is required for 0.707 Qtc. The practical use case is shallow-mount sealed subwoofer duty where mounting depth matters and output expectations stay within the measured linearity range. Infinite baffle is not recommended here because the driver would lose the reduced compliance of a sealed airspace, allowing it to more easily lose control of the moving assembly.

Testing and linearity limits vs. what is advertised

What it took to reach our high-level sweep limit, and how that compares to the published specs.

High-level sweep rule: Set just under the BL 70 percent point from LSI

High-level sweep limit for this sample: 11 V volts

Approximate electrical power at that limit at 20Hz: ~30 watts. Real power varies with frequency and impedance. volts

Rated power (published): 400 watts

Power used to hit the standardized limits in free air, relative to their xmax rating free air: The 11 V high-level TRF sweep used ~30 watts, about 7.5 percent of the 400 watt published rating. Based on the measured 8 mm one way at 20 Hz during the 11 V sweep, reaching 12.8 mm one way in free air at 20 Hz would require approximately 80 watts, about 20 percent of rated power. Real power varies with frequency and impedance.

Claimed Xmax vs. measured at BL 70%: 12.82 mm one way, 106.8 percent of the manufacturer’s 12 mm claim.

Xmax @ 50% Cms: 13.11 mm one way, 109.3 percent of the manufacturer’s 12 mm claim.

Xmax @ 17% Le: 12.03 mm one way, 100.3 percent of the manufacturer’s 12 mm claim.

Manufacturer suggested sealed enclosure size (and its resulting QTC): 0.75 ft³ with fill nets a Qtc of 0.903 on this sample.

Required sealed enclosure for 0.707 QTC: 1.54 ft³ nets a 0.707 Qtc on this sample.

Xmax @ 50% Cms: 13.11 mm one way, 109.3 percent of the manufacturer’s 12 mm claim.

Overall performance snapshot

This is our subjective interpretation of the objective data. How we derive these scores can be found on the home page of the testing section.

Distortion & frequency response - TRF measurements

Method recap: Nearfield mic positioned at 1/10th the cone diameter plus 2 inches, on-axis. Response measured to 1 kHz and THD to 500 Hz. 1/6-octave smoothing. Two drive levels, 1 V baseline and a high level set at 11 V per the under BL 70 rule derived from LSI for this unit. This sample in this test at this voltage level hit approximately 8 mm one way excursion at 20 Hz in free air, below the accepted standard of 70 percent of BL. Distortion is reported both as percent and by harmonic.

At 1 volt - baseline

The 1 V distortion curve is jagged, with multiple peaks between 20 and 120 Hz. THD exceeds the visible 10 percent plot window at the lowest frequency shown, then shows notable peaks around the upper 20 Hz range, near 58 Hz, and again around the 95 to 100 Hz area. H3 distortion is the dominant harmonic right at the lowest part of the range, while H2 distortion becomes the stronger contributor through much of the 25 to 120 Hz band. H3 distortion also remains close enough in parts of the mid-bass that the harmonic balance is mixed rather than cleanly dominated by one component. One observation worth noting is that the measured 1 V THD level of this driver is higher than would be typically expected from the summed harmonic content shown on the plot. My personal suspicion is that some of the elevated 1 V result may be operational noise from the driver itself rather than purely harmonic distortion. Because this design places the motor structure on the front side of the cone, it is possible that air movement and turbulence within or around the motor assembly could be contributing additional noise that is being captured in the THD calculation. This is only a hypothesis based on the measurements and driver construction, not a confirmed finding.

At high level voltage (11 V volts)

At 11 V, the relative distortion curve is lower and less jagged than the 1 V baseline through most of the 30 to 120 Hz range. THD is highest at the bottom of the sweep, at about 8 percent near 20 Hz, then drops quickly by the 40 to 60 Hz region. A broad rise appears around roughly 75 to 85 Hz, peaking around 4 percent THD. H2 distortion is the dominant harmonic through most of the high-level sweep, including the low bass and the broad 80 Hz rise. H3 distortion is present, but it does not become the main feature through most of the desired subwoofer passband. The high-level result does not show a new narrow spike in the 20 to 120 Hz range, which suggests the main change is the distortion ratio and harmonic balance rather than a new discrete high-level feature.

Delta - 1 volt distortion vs. high level distortion

The main change from 1 V to 11 V is that the jagged low-level curve becomes smoother and lower in relative percent through most of 30 to 120 Hz. The 1 V curve shows several sharper peaks, while the 11 V curve becomes a broader, lower distortion shape with one main rise near 80 Hz. Harmonic balance also shifts, with the 1 V sweep showing more mixed H2 distortion and H3 distortion behavior, while the 11 V sweep is more consistently H2 distortion dominant. One possible explanation is the Carbon C12's front-mounted motor structure. At lower drive levels, motor or mechanical noise may be less masked by the fundamental output, making irregularities appear more prominent in the distortion data. I have personally heard a small amount of motor noise from this subwoofer in certain situations, which is where my hypothesis is coming from. The drop in relative percentage should not be read as the driver producing less absolute distortion energy, only that distortion is lower relative to the much higher fundamental output.

What this means in practice

At the 11 V high-level sweep, this sample’s measured THD is highest near 20 Hz, then settles lower through most of the 30 to 120 Hz range with a broad rise centered around roughly 80 Hz. H2 distortion is the dominant harmonic through most of the high-level sweep, while H3 distortion remains present but secondary through the primary subwoofer band. Compared to the jagged 1 V baseline, the high-level distortion shape is smoother and lower in relative percentage through much of the operating range, though the practical clean stroke ceiling still lands close to the published 12 mm xmax range.

Motor & suspension linearity - LSI measurements

Method recap: Klippel LSI large-signal identification for this unit, cold and used for enclosure computations. Standard thresholds in this project are BL 70 percent, CMS 50 percent, and a 17 percent inductance variance criterion. Commentary below ties the large-signal behavior to the acoustic results.

Bl(x)

Bl(x) shows how much motor force a speaker produces as the voice coil moves, B is magnetic field strength and L is the wire length in that field. A high, wide, symmetrical BL curve means linear control and low distortion, a steep or uneven drop means earlier output limits and rising distortion, which is why BL(x) is often the most telling single Klippel LSI indicator of real performance.

Bl(x) window and shape

BL 70 percent occurs at 12.82 mm one way. The BL curve is not ideal with a somewhat offset, peaky shape rather than a flat plateau, (though not as extreme as some of the other shallow subwoofers we have tested such as the Wavtech ThinPro12), with force factor tapering gradually on coil in, but rather abruptly on coil out as displacement approaches the ends of the test window. For this sample, the BL window supports the published 12 mm xmax claim under the project’s 70 percent BL standard.

Bl(x) symmetry

The LSI note calls out slight BL asymmetry toward the outer stroke. The table reports the BL symmetry point at 0.17 mm at xprot, with BL asymmetry of -2.68 percent, so near-limit BL symmetry is not the dominant issue. The symmetry range plot does show more offset at low excursion, which is consistent with some H2 distortion contribution, but the high-stroke BL asymmetry is relatively small in this sample.

Cms(x)

Cms(x) is suspension compliance versus displacement, the inverse of stiffness. When the curve is broad and symmetrical, motion is linear and distortion stays low. Early roll off or offset indicates progressive stiffening or mis-centering, which adds mechanical distortion and caps clean excursion.

Cms(x) window and shape

CMS 50 percent occurs at 13.11 mm one way, slightly beyond the published 12 mm xmax claim. The compliance curve stays usable through the rated excursion range, but it falls more aggressively toward the ends of travel, especially on the coil-in side.

Cms(x) symmetry

The suspension symmetry plot shows the suspension behavior offset toward the outer stroke, matching the LSI note. The table reports stiffness asymmetry of 26.99 percent at xprot, so the suspension is a larger asymmetry contributor than BL in this sample. That type of compliance asymmetry is consistent with added H2 distortion, especially where the TRF plots show H2 distortion becoming the stronger component.

Inductance - Le(x) and Le(i)

Le(x) and Le(i) measure how a subwoofer’s voice coil inductance changes with position and current. These curves show how stable the motor’s magnetic field is under real movement and drive conditions. When inductance varies heavily, it causes distortion, uneven response, and a loss of upper-band clarity, which is why Le(x) and Le(i) are critical for evaluating how clean and consistent a motor’s behavior really is.

Level and shape

Le at rest is 0.59 mH cold. The Le(x) curve changes pretty drastically with position, dropping much lower on coil-in travel and rising through the outward stroke before tapering near the end of the window. The 17 percent Le variance criterion occurs at 12.03 mm one way, making it the earliest of the three standardized limits on this sample. Even though it still meets the project’s xmax standard, this surprised me personally considering how small the coil is. Based on the inductance behavior, it is probably safe to say there are no obvious inductance management measures used in this motor, such as shorting rings.

Current dependence

Le(i) is very stable over the plotted current range, with only a slight downward tilt as current increases. Frankly, I think this is the first time I have seen a driver with a non-flat Le(x) but with a very flat Le(i). I do not know what causes or allows for that.

Qts(x)

Qts(x) is the driver’s total damping versus excursion, combining electrical and mechanical losses. Stable, symmetrical Qts(x) means consistent control, while large variation or asymmetry signals uneven damping that can shift response, raise distortion, and cause compression.

Qts stability

Qts is 0.38 cold near center and 0.39 warm near center. The Qts(x) curve is lowest near the center of travel, then rises as stroke increases in either direction. The rise is stronger on the coil-in side than the coil-out side, so damping is not symmetrical at high excursion. This can contribute to response and compression changes, as well as added distortion as the driver approaches the ends of its usable stroke.

LSI takeaway

The earliest standardized limit is the Le 17 percent criterion at 12.03 mm one way. BL is broad enough to support the published 12 mm xmax claim, though with some asymmetry involved. The CMS plot shows the suspension is more asymmetric than the BL curve, which is consistent with H2 distortion showing up strongly in the TRF data. Le(x) varies enough with position to set the practical clean xmax ceiling, while Le(i) is stable with current. Qts rises noticeably with stroke, especially inward, so damping changes as the driver is pushed toward the ends of travel, and is asymmetrical.

Enclosure alignment calculations

Manufacturer sealed enclosured recommendations and the resulting QTC: 0.75 ft³ with 50 percent fill nets a Qtc of 0.903 on this sample.

Sealed volume required for 0.707 QTC on this sample: 1.54 ft³ nets a 0.707 Qtc on this sample.

Applicable for infinite baffle? Not recommended for this driver. It is designed for moderate sized sealed enclosures, and does not have much xmax, and has a low enough QTS to support infinite baffle use.

T/S parameters