Speaker Measurements: Woofers

About the Results:

Drivers are tested on a 1.2x0.9m quasi-infinite baffle, with a measurement mic placed close range. The baffle accepts flush 225x225mm inserts which are constructed specifically for each driver, allowing the baffle to be reused for drivers having vastly different mounting hole requirements. Drivers with thick baskets/frames are flush-mounted in their inserts while drivers with thin stamped steel frames are surface mounted. The rear opening is always chamfered to aid air airflow, masking tape is applied to cover any gaps such as between the insert and baffle and around imperfect driver cutouts. Technical details about the setup can be found in this article.

For each driver, at least three plots are produced: A Frequency response plot and two types of Harmonic distortion plots.

Frequency Response:

4 far-field (30cm) measurements (0°, 15°, 30°, 45° off-axis) are combined with a single near field (10cm) measurement to produce the frequency response plot. The near-field measurement is crossed over to the far-field at 1khz. All measurements are windowed (frequency-dependant windowing) to reject baffle and room effects, however unavoidable room effects are present in all plots below 500Hz. A recession due to destructive floor/wall reflection is observed in all measurements between 300 and 500Hz.

The amplifier output is set to 2.83Vrms into 8ohms (1Watt into 8ohm), regardless of the nominal impedance of the driver under test. Therefore in order to compare the sensitivity at 1Watt, subtract 3dB from 4ohm drivers and add 3dB to 16ohm drivers. Subtract about 10.5dB from the curves shown to get the SPL at 1 metre instead of 0.3 metres.

Harmonic Distortion:

Harmonic distortion plots are produced from a single measurement at 30cm (0°), with the amplifier adjusted to achieve an average of 96dB SPL @ 30cm throughout the usable frequency range of the driver. This is equivalent to around 86dB SPL at 1metre. 96dB SPL at the mic is normalised to 0dB on the plots.

The 'raw' and 'relative' plots are produced from the exact same data, the difference being that the fundamental is normalised flat in the relative plot. The relative plot offers a more realistic representation for drivers that have large deviations in frequency response and would be filtered to near-flat response in practice. Be aware that the amplitude of harmonic distortion does not change linearly with the amplitude of the fundamental - i.e. after notching out a 10dB peak in frequency response, the harmonics of the filtered frequencies will drop more than 10dB in amplitude. Therefore the 'relative' plot is slightly biased against drivers that have large peaks in frequency response.

Two additional HD sweeps are performed at elevated levels. Firstly the mic is moved back to 60cm and the amplifier level adjusted +6dB (to maintain approximately 96dB SPL at the mic), therefore achieving a level of 92dB SPL at 1metre or +6dB over the first measurement. The mic is then moved back to 120cm and the level increased a further 6dB, therefore achieving a level of 98dB SPL at 1metre or +12dB over the first measurement. When the mic is moved further away the curves become rougher/fuzzier as the measurement is more influenced by baffle edge diffraction and room reflections, however this is not important as we are only interested in seeing the average level of the harmonics and if there are any narrowband peaks at certain frequencies.

It's important to maintain the same level at the microphone when doing harmonic distortion measurements as the microphone itself produces some distortion. When subjected to 96dB SPL the microphone I use produces a second harmonic at around -50dB. Higher order harmonics produced by the microphone are negligible in magnitude. I choose to measure at 96dB as it allows me to see high order harmonics closer to the noise floor, at the expense of 2nd order harmonic distortion of the mic. If the measured 2nd order harmonic smoothly follows the level of the fundamental and remains ~-50dB we can be confident that the 2nd order harmonic from the driver is significantly below -50dB. If the measured 2nd order harmonic becomes very rough around -50dB we can assume that the level produced by the driver is around -50dB and the differences in phase are causing constructive and destructive interference leading to the roughness. If the measured level of 2nd harmonic is significantly above -50dB (say, -30dB) and the level of the fundamental is not significantly above 0dB, we can be confident that this is 2nd order distortion coming from the driver.

Bass Harmonic Distortion (below 200Hz):

I've chosen to add an additional HD sweep for this set of woofers to allow the harmonic distortion of the bass to be compared fairly. Large differences in frequency response occur due to different T/S parameters, which means that drivers with low Qts and/or high Fs rolloff significantly below 150Hz and therefore are subject to far lower excursion when tested on an infinite/open baffle. The reduction in movement of the coil and suspension significantly lowers distortion, therefore a fairer way to compare the bass capabilities of woofers with different T/S parameters is to equalise them such that their excursion is the same versus frequency. I leave it as an exercise for the reader to consider what the enclosure requirements are for different woofers, how this alters the frequency response (and therefore excursion) and what that will do to lower distortion from the levels measured here

The test method is as follows: I place the microphone level with the baffle only a couple of centimetres away from the cone and equalise the measured frequency response to within 1dB of a 40Hz 2nd order Butterworth highpass response. Equalisation is accomplished using the parametric EQ capability of a MiniDSP 2x4 processor. Once the target frequency response has been achieved the microphone is moved to 10cm away from the baffle and a sweep is run such that the level is 96dB at the microphone between 100Hz and 200Hz. This achieves a level of around 76.5dB SPL at 1metre or -9.5dB compared to my normal testing level. 6th, 7th and 9th harmonics are plotted as they become significant when a woofer reaches excursion limits (mechanical, wind noises).

The microphone is then moved back to 30cm and the level readjusted to achieve 96dB from 100-200Hz again, therefore achieving around 86dB SPL @ 1metre, my usual base test level. Some slight variation in the frequency response is noted due to room effects and dipole effects as my 'quasi-inifinite' baffle is actually a large open baffle. All drivers however are subject to the exact same effects so comparisons can still be made.

Finally the level is increased by 6dB without moving the microphone. This achieves 92dB SPL @ 1metre or +6dB however the microphone is subject to 102dB SPL which raises the level of 2nd order distortion from the microphone to around -40dB. I made the decision to tolerate a higher amount of distortion from the microphone as moving the microphone any further back introduces significant room/baffle effects which would compromise the repeatability of this testing should I ever need to perform future testing in a different room.

There is no need to test at a higher level since at 20Hz in the +6dB test, the excursion from a 6.5-7" driver is about +/-6.5mm which is already beyond practical limits. It is easy to see that all the drivers are operating as pistons at these frequencies so it is possible to infer the amount of distortion due to excursion at higher frequencies and levels from these plots - at twice the frequency and 12dB louder the distortion due to excursion will be similar.

Le(x):

Le(x) is the measurement of how the voicecoil inductance changes as the coil moves in and out of the motor (excursion). Instead of just presenting the value of Le at a particular frequency versus coil excursion I provide the entire impedance sweep. Le(x) can be used to gauge the amount of intermodulation distortion produced by a speaker - as one tone causes the voicecoil to move in and out, the changes in impedance amplitude modulates other tones, particularly in the treble where the impedance changes are most severe. Good Le(x) behaviour is particularly important where you would like to use a driver well into the inductive frequency range (above 500Hz or so for these drivers), such as in most 2-way applications. For applications where the driver is only used in its resistive or mechanical resonance regions (e.g. a subwoofer) it is not important.

The measurement is performed by holding the cone at a fixed amount of excursion from its resting position and performing an impedance measurement using a Dayton Audio DATS V2. Excursion is measured with a dial indicator. Significant artifacts occur below about 500Hz due to holding the cone in position which messes up the mechanical resonance of the driver. The smoothness of the curves is not important here - the spread between the curves is what matters. Excellent Le(x) performance would result in a group of curves which lie exactly on top of each other from 500Hz to at least 3kHz.

I don't recommend using these impedance sweeps to gauge anything other than Le(x) performance as significant effects are introduced even on the free-air (0mm) measurement due to the driver being installed on a small baffle (the 225x225mm baffle insert) and being placed magnet-down on a table which blocks any venting features of the motor. At this time I don't have the necessary setup to accurately and conveniently measure compliance equivalent volume (VAS) so I will not publish T/S parameters, however if I notice a large discrepancy in T/S parameters compared to what the manufacturer has published I will mention it in the comments below.

Ratings:

I've provided some subjective ratings as well as some commentary/interpretations of the measurement results below. I may choose to edit the ratings or comments over time as I re-evaluate the measurement results.

"Build/Construction" rates the overall build quality and mechanical construction of the driver. e.g. Add points for a sturdy cast frame designs that aid ventilation, large spiders and surrounds which allow for large excursion capability, pole piece and former venting. Subtract points for messy glue applications, frames which require tight baffle machining tolerances or unusually sized screws, ugly stamped steel frames, poor ventilation, flimsy parts which will break easily, tinsel lead arrangement which becomes overstretched or buzzes at high excursion, etc.

"Performance" rates the linear and non-linear distortion measurements. It is a measure of how usable the driver is. To achieve a 10/10 rating would require a ruler flat frequency response from 100Hz to at least 2kHz, high sensitivity and high power handling (subjective evaluation, not tested), non-existent harmonic distortion and T/S parameters which allow decent extension in a reasonably sized box (F3 no higher than 40Hz). Drivers lose more points for issues which severely restrict their usable bandwidth, and less points for issues that are benign or occur outside the usable frequency range. Large dips in frequency response or high harmonic distortion below 1kHz attract severe penalties as these are issues which cannot be easily fixed. A smoothly ramping response or benign peak in the frequency response does not attract a large penalty as these are issues which are easily fixed in the crossover. Large breakup peaks in the response above 6kHz do not attract severe penalties as they can be dealt with by the use of 4th order crossovers and/or notch filter. Large breakup peaks however do tend to trend into higher harmonic distortion at lower frequencies and this does attract a penalty if it exists. Drivers may be penalised for manufacturing consistency issues which affect performance.

"Value for money" evaluates the performance versus price. Listed prices are in US Dollars at the time of purchase

Results:

Comments: