Discuss Subwoofer Tests Explained in the Manufactured Speakers | Subwoofers forum. Welcome to the Subwoofer Tests forum!
This is the place to get accurate and comparable info how different subwoofers measure ...
This is the place to get accurate and comparable info how different subwoofers measure against each other. Many magazines (both paper and online) also measure subwoofers, but their time is limited and not too many subwoofers gets measured per year. It is also much harder to get one’s favourite subs measured in a magazine, not to mention DIY subs. Also different testing methodologies used by reviewers/magazines often makes comparing results almost impossible. But people are always interested to find out how different subwoofers measure against each other, and also because most specifications released by subwoofer manufacturers themselves are somewhat “unreliable” and hard to compare, I decided to start helping people (and of course to satisfy my own enquiring mind) to find out how they compare using accurate gear and more advanced and versatile tests than many magazines take (sad but true fact).
Since I don’t have access to an anechoic chamber, I'm performing my tests outdoors on a big field with no nearby large reflective surfaces (usually ~10 m distance is considered adequate for testing frequencies down to 10 Hz). Naturally testing ourdoors has its cons like random background noise, wind, ambient temperature, air humidity and atmospheric pressure. All these are unfortunately uncontrollable variables and therefore can and will affect on measurements. The key thing is try to keep their influence as small as possible.
Another great source for accurate subwoofer measurements can be found at AV Talk forum. It should be noticed that our testing methods and rigs are not identical, meaning one shouldn’t compare our results directly.
Naturally measurements are only one part of the subwoofer evaluation. If the particular subwoofer doesn’t sound good in your own room and to your ears, it doesn’t matter how good it looks on paper or sounds to some random reviewer. Measured performance and perceived performance/sound are of course tightly connected, which means that one can have a pretty good picture how a specific subwoofer will perform/sound by just looking at the measurements. Of course the final judgement should be done in your own room with your own ears.
One should not forget the importance of the final in-room frequency response and the integration between the sub and the mains, which IMO are the most important factors (if the max output level is high enough for your needs) when pursuing good bass reproduction. Especially the importance of good/strong mid-bass output from the main speakers is often underestimated. Pair of 5.25” or 6.5” woofers won’t be able to keep up with a strong subwoofer even at 15 dB below reference or higher (in a 15-25 m^2 room), when a typical 80 Hz crossover frequency is used. 80-300 Hz range must be balanced with the range below at all levels in order to have a solid and strong sounding bass. Subwoofer should be just the dot on the i on a system, not the dominating component which around the system is built to.
Laptop: Acer Aspire 5024WLMi (AMD Turion 64-bit 3000+, 512MB DDR, 80GB HD) Soundcard: Creative Labs SB Live! 24-bit External Microphone:IBF-Akustik EMM-8, serial number 31070504 SPL-calibrator:IBF-Akustik SC-1, serial number 0503003 Programs used: TrueRTA, SpectraPro, ETF5
I measured the frequency response of the soundcard using RMAA 5.5 program and it was +/- 0.3 dB 10 Hz - 15 kHz. Without a correction file, the frequency response of the microphone is +/- 1 dB 10 Hz - 20 kHz or +/- 0.3 dB 10 Hz - 5 kHz. With a correction file both soundcard and microphone frequency responses are corrected to flat. Microphone was calibrated on 31.7.2005 by IBF-Akustik, the reference microphone used was B&K 4133. The measured frequency response of the pre amp is +/- 0.1 dB 10 Hz - 20 kHz. The frequency response of the whole measuring chain is therefor better than +/- 1 dB 10 Hz - 20 kHz.
The SPL-calibrator gives a 94 dB @ 1 kHz signal with a +/- 0.5 dB accuracy. The calibrator was calibrated on 17.8.2005, the reference calibrator used was Quest CA-22. I did the calibration for the absolute SPL indoors at the ambient temperature of around 24 degrees Celsius. The SPL-calibrator was originally calibrated at this temperature. The atmospheric pressure was around 1035 hPa (Finnish weather service), which is very close to the 1007 hPa where the calibrator was originally calibrated. The atmospheric correction for the calibrator is around 1 dB / 200 hPa. The microphone is practically independent for variations in temperature and atmospheric pressure. A hairy windshield was used around the microphone, which helped to eliminate most wind noise (wind noise mainly occurs at very low frequencies).
Round 3 ->
Frequency response was measured crossover at its max and min settings and also bypassed if available. 10 second sine sweep from 200 Hz to 10 Hz was used. Level was matched at 90 dB at 50 Hz at 2 m distance. Mic was on the ground and the distance was measured from the acoustic center of the subwoofer. For sealed subs acoustic center lies right in front of the voice coil (front baffle was used for consistency). For vented and subs with passive radiator(s) acoustic center lies somewhere in between the driver(s) and the port(s)/PR(s). But because most of the output (vent/PR affects only near tuning frequency) comes from the driver, I measured the 2 m distance from the (active) driver also with vented/PR subs (using a directional line from the mic). If the sub had a port or PR on the back, the sub was rotated so both were equidistant to the mic. In some cases I also measured all combinations and choosed the one which gave better FR. Being consistant and fair to all subs regarding this matter is extremely difficult because of the various driver/port/PR combinations subwoofers have. If all subs were single driver, sealed, front-firing subs, there wouldn’t be any problems. The key thing was to measure the 2 m distance from the point where the most of the output comes from, but also keep the other sources (if there was one or more) as close as possible to the 2 m distance. Subs were kept in their normal operating positions, for example base plates weren’t removed or subs turned on their sides etc. Same orientation/position was used during all measurements. All graphs are using a full 1/24 octave resolution. +/- 3 dB points were calculated using maximum or bypass (if available) crossover setting.
Round 1 and Round 2
Subwoofer’s frequency response was measured with TrueRTA’s Quick Sweep signal. The signal is a short-duration, digitally synthesized logarithmic sine sweep from 0 Hz to 24 kHz (when using a 48 kHz sampling frequency). It takes only a few hundred ms’s to sweep the normal subwoofer operating range. Unfortunately this method isn’t as accurate as the normal sine sweep (which was the reason why I switched to it) under relatively loud background noise. Therefore small variations can be expected between these methods. +/- 3 dB points were calculated using maximum or bypass (if available) crossover setting.
+/- 3 dB points: 18.5 Hz - >200 Hz
The best possible anechoic FR depends of the room/space where the subwoofer will be put in. Usually gradually sloping (6-12 dB/oct.) response gives the flattest in-room response due to room gain. The bigger the room the lower the response can be flat without having problems with elevated low end response. Below can be seen two frequency responses. The first one is more suited for small/mid-sized rooms while the second one is more suited for larger spaces with less low frequency room gain. If the second sub will be put in a small (rigid walled) room, frequency response will be strongly elevated below 30-40 Hz, which will sound too "bottom heavy”.
Maximum long term output level was measured using a 30 second linear sine sweep from 100 Hz to 10 Hz. First sweep was level matched at 90 dB at 50 Hz. Drive level was raised by 5 dB after each sweep. Sweeps were taken up to the point where the output level exhibited clear compression. This is a very demanding test and the results should not be compared to other tests with faster sweeps etc. Power compression graph is showing the relative compression to the 90 dB sweep.
Round 3 ->
Sweeps were taken up to the point where the output level stopped rising (5 dB or more of compression). I also tried 2 dB as the last step, but usually it didn’t help.
Round 1 and Round 2
Maximum long term output level was evaluated with a 45 second linear sine wave sweep from 100 Hz to 10 Hz. Sweeps were conducted at progressively louder (3 dB increments) levels. If the sound of the subwoofer became too ”stressed”, the sweep was stopped. The levels shown are the very maximum levels, meaning the next sweep was totally compressed.
Max output should be as high as possible over as wide frequency range as possible. Power compression should be as low as possible.
Round 3 ->
The THD was measured at the same time with max level sweeps. The frequency resolution is around 1.4 Hz. Note that the SPL isn’t absolute but follows the frequency response and possible power compression instead! One should also notice that if the amplitude of the fundamental goes very low, the accuracy of the THD measurement is reduced. For example if the fundamental is at 70 dB, the random background noise at around 30 dB causes 1% THD. Measuring THD on the fly is more demanding for a subwoofer than measuring with single 2-3 second sine waves. Therefore the results are not perfectly comparable with each other.
Round 1 and Round 2
Subwoofer’s THD levels measured with single sine waves. We adjusted the gains on subwoofers so that we got 90 dB (+/- 0,5 dB) at 50 Hz - this was our starting point. Then we run the “sweeps” using 5 dB increments up to 105 dB. The program measures the THD level at 10 frequencies between 100 Hz - 16 Hz. The program interpolates the values in between. Each tone plays around 2.5 seconds, so the whole ”sweep” takes around 25 s. Notice that the absolute SPL is correct only at 50 Hz. For other frequencies one must look at the power compression sweeps and find the correct SPL.
THD levels should be as low as possible over the whole frequency range. THD at low frequencies (below 40-50 Hz) isn’t as harmful or audible as THD at high frequencies (above 40-50 Hz).
The individual levels for harmonics from second (H2) to sixth (H6) were manually plotted at 15 Hz, 20 Hz, 25 Hz, 32 Hz, 40 Hz, 50 Hz, 63 Hz, 80 Hz and 100 Hz. Some subwoofers didn't have enough output at the lowest frequencies, so those frequencies couldn't be plotted. The Y-axis scale is logarithmic instead of linear. The chosen sweep level wasn’t kept the same for all subs because of different max output levels. Usually the second highest sweep was chosen. That’s why you shouldn’t compare the absolute levels between different subwoofers, only relative levels between harmonics.
All harmonic components should be as low level as possible (at all levels). Due to masking effect, high-order harmonics are more audible than low-order harmonics, meaning same THD level with different kind of component distribution will sound different. It is also said that even-order harmonics (H2, H4, H6 etc.) are less likely to be perceived than odd-order harmonics (H3, H5, H7 etc.).
Mathematically group delay is the negative derivative of the phase response. Phase response was measured using the ETF-5 program which calculates it from the measured impulse response. Since frequency response and phase response correlate with each other, also FR and GD correlate with each other. Meaning flat or gently sloping FR results in low GD. That's why the final in-room frequency response should be as flat as possible. There are no studies defining the audibility of GD at low frequencies, but the threshold is suggested to be in the range of 1-1.5 cycles. Therefore I have included the 1 cycle curve on the graphs.
Ideally GD should be as low as possible, or at least gradually/gently rising.
Any sudden change (especially if high-Q) in frequency response always causes ringing (sometimes also referred as ‘stored energy’), and that’s why a smoothly sloping FR down to as close to direct current (0 Hz) as possible causes the least ringing and results in fast decay at all frequencies. Also same applies to the upper end of the frequency response. A steep low pass filter causes a rise in group delay and ringing near cut-off/crossover frequency. Spectral decay graph shows the frequency response after the input signal has stopped. I used a 200 ms gate time and a 40 ms slice spacing up to 200 ms. The spectral decay is evaluated down to -24 dB level from the highest amplitude point. The measurement itself was taken at around 90 dB level, which allowed a good S/N ratio but didn’t push subwoofers into compression. The same info is often presented with a ‘waterfall graph’.
Spectral decay should be as fast as possible over as wide frequency range as possible. Ideally the graph should only have one frequency response slice. Ringing at low frequencies isn’t as harmful or audible as ringing at higher frequencies.
”Spectral contamination is a graphic analysis of cross-modulation products (’self-noise’) produced by a system excited by a multi-frequency signal. Multi-tone tests are far more representative of the rigors of music or speech reproduction than traditional stimuli. The resulting non-linear distortion products (the noise between the tones, which are distortion products generated at frequencies where no excitation energy is present) correlates with subjective perceptions of quality, such as ’clarity’ and ’coloration’.” (Spectral Contamination Measurement, by Deane Jensen and Gary Sokolich, AES preprint 2725, 85th AES Convention, 1988).
This is a new test which hasn’t been used by subwoofer reviewers before. It has been used with full range speakers, but not with subwoofers. Since there wasn’t any info how this test should be performed on subwoofers, I and my good peer Ed Mullen had to start things from the scratch. Based on the SC tests performed on speakers, 10 stimulus frequencies ranging from 20 Hz to 77 Hz were chosen. A multiplier of 1.1618 was used between stimulus frequencies. 8 kHz sampling frequency and a 64k FFT size resulted in 0.122 Hz frequency resolution. I measured SC starting at 90 dB and then raised the level by 5 dB as long as the sub could produce a full 5 dB step. Three highest SPL screens are showed. Frequency range goes from 15 Hz to 200 Hz.
Spectral contamination test will show all non-linear distortions and mechanical ‘self-noises’ (panel vibrations, driver motor noise, port noise etc.). Also the limiting circuits often became very audible during this test. Many subwoofers started sounding like rumbling and popping V2 engines instead of steadily humming subwoofers. Some subs were already struggling at the lowest test level of 90 dB (combined power of all 10 tones, each individual tone at 80 dB). Interpretting the results isn’t very easy because as said earlier, these are the first tests ever published for large number of subwoofers. It is pretty clear that subwoofers with high output capabilities have also low overall spectral contamination at both low and high test levels. Whether this test will show anything usefull beoynd the ‘old tests’ is yet to be examined.
Ideally spectral contamination graph should only have the ten stimulus frequencies and nothing else between or around them (some amount of background noise and system noise is of course always present). The lower level the 'self-noise' is, the better.
The CEA-2010 (Consumer Electronics Association) is a standard which defines a method for measuring the performance of powered subwoofers. It present tone bursts centered at 1/3 octave frequencies in 20 - 63 Hz range. I also measured some additional frequencies below and above that range. The test is performed by increasing the input level until the SPL is limited by prescribed frequency-dependent distortion threshold (staircase function) or compression (won’t go any louder). The test signal is 6.5 cycle long, shaped (Hann window) sine wave burst. So the length of the tone burst stimuli goes longer as the frequency decreases. The staircase function defines the allowed distortion level for each harmonic. The staircase function allows higher SPL for harmonics closer to the fundamental, and lower for harmonics further to the fundamental. This is because human hearing has a decreased tolerance for distortion components at higher harmonics (based in part on studies of distortion audibility and masking [Shorter (BBC 1950’s), Harman]). Allowed SPL by harmonic (compared to the fundamental): 2nd -10dB; 3rd -15dB; 4th and 5th – 20dB; 6th – 8th -30dB; 9th and above -40dB. CEA-2010 rating is defined by calculating an average distortion or compression-limited SPL (GP @2m, dB RMS values) for each of two one-octave low-bass performance ranges - Ultra Low-Bass: 20, 25, 31.5Hz and Low-bass: 40, 50, 63Hz. If max SPL is not measurable or the S/N ratio becomes too low through some portion of ultra-low band, then the measurer may choose to state “NA” for that band.
IMD is also one of the variables which hasn’t been used in subwoofer reviews before. By definition the intermodulation distortion occurs when the non-linearity of a device or system with multiple input frequencies causes undesired outputs at other frequencies. IMD is measured by inputting two stimulus frequencies using certain spacing between them, a few common stardards are 250 Hz / 8020 Hz (with a 4:1 amplitude ratio), 60 Hz / 7000 Hz (with a 4:1 amplitude ratio) and 19 kHz / 20 kHz. These stardards were designed for measuring electrical devices such as amplifiers. Naturally they are not suitable for subwoofers, so again I and my peer Ed Mullen had to start from the scratch with this test too.
Eventually we ended up using fundamentals (or carrier signals) at 30 Hz and at 72 Hz. 30 Hz isn’t too low for most subwoofers to produce and 72 Hz is still below the commonly used 80 Hz crossover frequency. This combination also doesn’t share any THD/IMD harmonics. The Excel spreadsheet we made calculates all IMD harmonics up to 6th order, which was more than enough because most subs didn’t have any 5th or 6th order IMD harmonics (already below background noise floor at ~30-40 dB, system's noise floor is at much lower level). There isn’t a very large database yet, but these first results suggest that subwoofers with high output capabilities and low THD levels also have low IMD levels. One should carefully examine if IMD is even worth measuring with subwoofers due this relationship.
Ideally IMD should of course be as low as possible at all levels.