Distortion test of drivers
For the PHOENIX main panel I use a set of 8" (21 cm) drivers, Scan Speak 21W/8554, which were selected some time ago for their low non-linear and linear distortion. With the current popularity of 7" (18 cm) drivers, and some new models available, it became time to re-asses the options. My selection was limited to drivers that I thought had potential or that had been recommended and were readily available. I missed that Accuton made drivers of this size, but have since added results for the C92/6. Audax and Morel could not compete in an earlier test. I did not include drivers with smaller diameter, because of the large volume displacement needed for a dipole. I have tested drivers like the SS 15W/8530K or Jordan, but they are just not suited for this application. There are many other drivers on the market and if you know of one that might stand out positively in the tests below, then please let me know, or better yet - send me a sample.
At the outset I must clarify that I am looking for the
midrange driver in a 3-way system. A dipole woofer with the Peerless 830500
could be crossed over as high as 150 Hz using a 4th
order acoustic transition. Compared to a 100 Hz crossover this would reduce the
volume displacement required from the midrange and thereby reduce its non-linear
Distortion tests are difficult to quantify by single numbers. This will be a relative comparison of drivers. Measurements are taken with the driver un-baffled and resting on its magnet, but without restricting air flow from any pole vent. Blocking the vent increases non-linear distortion considerably. The test microphone is at the center of the cone diameter and touches the plane which would be formed, if a flat sheet of paper were placed over the front of the driver.
Non-linear distortion test
A very effective test signal for measuring both harmonic and intermodulation distortion is a sinewave that is 100% amplitude modulated by another sinewave at 1/10th its frequency.
If set to 150 Hz its spectrum consists of a 150 Hz carrier
and sidebands at 135 Hz and 165 Hz with 6 dB lower amplitudes.
Without any test equipment for analysis this signal makes it easy to hear
non-linear distortion of a driver. (A four cycle burst of this signal is also
very useful for qualitative in-room woofer response analysis. Ref.1)
The drivers under test must be operated at the same volume
displacement for meaningful comparison of distortion produced. With the
microphone very near to the cone the close-up sound pressure must be adjusted in
proportion to the cone diameter so that the far-field SPL will be the
same for different size drivers. (e.g. D. B. Keele, JAES, Vol.22, No.3, April
Drivers 10 and 11 were added for comparison to a small cone diameter. Excluding these two drivers the best performance at 150 Hz was turned in by the SEAS W18E001 and the worst by the Vifa PL18WO-09.
At 800 Hz the best and worst performances were obtained from the Scan Speak 21W lab 43/95, a prototype 21W/8554 with SD-1 motor, and the 18W/8545.
While distortion differences between drivers are easily heard with the 150 Hz test tone, this is not the case at 800 Hz, because the signal itself is rather unpleasant and shrill sounding.
I have given each driver a place for its distortion performance though differences are often small and judgment is based on multiple observations.
Linear distortion test
The test signal consists of four cycles of a sinewave multiplied by a Blackman window (F. J. Harris, Proc. IEEE, Vol. 66, Jan.1978). This gives a fast rise time waveform with a spectrum that falls off rapidly at low and high frequencies. It makes a pinging sound. Ref.13
The drivers are tested at 800 Hz, 1200 Hz, and 1600 Hz in their cone breakup regions. I observe how rapidly the driver output decays and stored energy is released. The presentation is on a logarithmic amplitude scale and shows the envelope of the burst signal (the magnitude of the analytic signal). Below is the example of a tweeter with fast decay at 1600 Hz.
To avoid that the burst response is influenced by the driver's frequency response outside its intended operating range, I run the burst signal first through the 1440 Hz, 24 dB/oct PHOENIX crossover lowpass filter. Also, measurements are taken outdoors to eliminate room any reflection problems.
The Seas W18EX001 gave the best performance amongst the drivers tested at 800 Hz. The Vifa P21WO-20 was the worst.
To evaluate the burst performance I look at the shaded areas, giving less weight to, but not ignoring, the area below -30 dB.
At 1200 Hz the order was Vifa P13WH-00 first and Vifa P21WO-20 last.
At 1600 Hz the order was Scan Speak 18W/8545 first and again Vifa P21WO-20 last, together with the Vifa PL18WO.
The table lists the placement of all the drivers as they fall between the extremes shown above.
In addition to the above data close-up frequency response and terminal impedance was measured for all drivers, but the data were not directly taken into account for this survey. Midrange frequency response data.
Summary and conclusions
There is a considerable amount of data to consider. One way is to take the average of the two sets of non-linear distortion rankings and of the three sets of linear distortion. It shows that the order of the drivers is different for the two forms of distortion. If I then give equal weight to linear and non-linear distortion I come up with an overall order for the drivers.
A most remarkable driver is the SEAS W18EX001 because of its excellent linearity,
though its burst performance degrades somewhat at 1200 Hz and above, even with
no cone breakup below 3 kHz. It is well built with an open
cast basket, fully vented spider, and no pole vent.
The W18EX001 is the closest competitor for the 21W/8554 amongst the drivers tested. It might be worthwhile to experiment with it in a PHOENIX like system. Dipole peak equalization is likely to change and the baffle dimensions would have to be checked for optimum polar response.
Note 1: Since these tests were
done Seas introduced the W22EX001. Like the W18EX001 it has very low distortion,
but with 1.75 times the cone area it can move more air. This driver is my first
choice for any new open baffle speaker design. The large resonance peak at 5 kHz
requires a notch circuit. The PHOENIX pcb provides three
optional room equalization notch filters. One of them could be allocated to this
task and be patched into the midrange channel. The harmonic distortion of the
W22EX001 has a broad peak between 1.3 kHz and 3.5 kHz. The reason for this is
that distortion products generated at these frequencies are amplified by the
raised frequency response around 5 kHz. The notch filter will not help this
situation and the driver should ideally only be used below 1 kHz. The distortion
peak at 4.8 kHz in the Seas
data sheet is a measurement artifact. There is no increase in distortion.
Note 2: A low distortion driver for the 500 Hz to 3 kHz range is hard to find. There are a few large dome designs around, but they are displacement limited at the lower frequencies and poorly vented on the back side. The Seas W15CY001 looks promising if used below 2 kHz. I would investigate it, if I had a need for a 4-way speaker design.
Since non-linear distortion can always be reduced by using multiple drivers, while burst distortion remains constant, one could replace a single 21W/8554 with two 18W/8545, or with four P13WH-00 if levels are kept low. This approach is costly and quickly runs into polar response problems.
The test has given a snapshot of different drivers. It was by no means exhaustive, nor may I have done full justice to each, but I am quite confident that I caught the essential performance of the units. The survey was not taken, because I am unhappy with the 21W/8554 in the PHOENIX, but only because further reduction in the two types of distortion will give greater sonic payback than anything else. At this time I have no plans to revise the PHOENIX design.
FB: Years later, driver design has developed towards lower distortion levels. To see some LINKWITZ 22MG measurements, scroll down here
As part of my review of drivers for the PHOENIX I compared a Seas Excel T25CF002-06 Millennium tweeter to the SS D2905/9700 soft dome, the SS D2904/9800 Aluminum dome tweeter and the Vifa D26TG-35-06, by measuring stored energy at 3 kHz, 6 kHz, 12 kHz and non-linear distortion of a 1.6 kHz carrier with 100% AM of 160 Hz. Voltage levels were 5.6 V and 8.9 V rms.
Lowest non-linear distortion was measured for the T25CF, closely followed by the SS 9700. The SS 9800 had predominantly 2nd order distortion. The D26TG-35 showed a broad spectrum of distortion products from low to high frequency. Though lower in level than the 9800 2nd order products, the higher order components are more serious.
Stored energy was lowest for the T25CF, followed closely by the SS 9700. The SS 9800 stores energy above 10 kHz, probably due to the Helmholtz resonator formed by the cup in front of the dome. This feature is needed to keep the frequency response up. The D26TG ranked slightly ahead of the SS 9800 on bursts.
The measurements confirmed my preference, based on extensive
listening, for the D2905/9700 as the more accurate tweeter than the SS metal
dome. For any new design I would seriously consider the Seas T25CF002-06
tweeter, though I am not sure how directly the small improvements in measurement
data translate into audible benefits.
Addendum: I recently compared two Vifa XT25TG-30 to the Seas T25CF002-06 and Scanspeak D2905/9700 tweeters. Non-linear 2nd order distortion products of the XT25 were by far the worst of the three, 3rd order products were ok and higher order products were very low. Both units performed well on stored energy, better than the T25CF, which was slightly ahead of the 9700. Overall not a bad performance, but not quite the same as the Seas and Scanspeak, in my ranking, especially when the higher directivity of the XT25 is considered.
Vifa P13WH-00 (105 mm effective
A - Non-linear distortion was tested with 100% AM sinewave
at 120 Hz, 150 Hz, and 800 Hz. The amplitude was 5.6 Vrms (26 Vpp) and 2.8Vrms.
120 Hz, 5.6V P-V-S-A
B - Stored energy was tested at 800 Hz and 1600 Hz with a 4-cycle Blackman enveloped burst.800 Hz P-A-V-S
1600 Hz A-P-V-S
Overall ranking: A-P-V-S, all measured close to each other
These results would point to the Peerless 850488 as my first choice for best performance versus price. The four drivers were selected with some thought. They are well designed and have good sound quality. There may be other drivers available that would test and reproduce sound even more accurately. I would encourage you to use this test method and investigate for yourself, if you need a driver of this size.
When I developed my surround speaker I compared non-linear distortion of the Vifa P13WH to that of the PL14WJ09. In all cases the PL14 had lower distortion.
It should be apparent by now why stored energy and non-linear distortion are such important driver parameters. Stored energy is difficult to equalize for and non-linear distortion cannot be corrected. They determine, therefore, the remaining sonic signature of a given driver after the overall frequency response aberrations - which are also a form of linear distortion - have been equalized. Since accuracy of reproduction is my goal a driver should not noticeably contribute its own sound.
For large signal measurements see also: Triggered burst measurements of tweeters