- Written by Dr. David A.Rich
- Published on 27 March 2013
The NHT B-10d Subwoofer On The Bench
The NHT is a sealed design, making it easier to attain accurately measure the near-field frequency response because the no port response needs to be accounted for. More significantly for the user, the absence of the port makes the ssubwoofer's frequency response repeatable in production. Fewer parameters are in the equation that predicts the response curve, meaning the response should be less sensitive to parameter variation in the production subwoofers.
Figure 1 shows the near-field frequency response of the subwoofer in both music and movie mode. The NHT B-10d looks absolutely flat in the near-field. The small deviations around 70Hz may be a measurement artifact. As you can see, movie mode bumps up the response for the sound effects. Many subwoofers only have movie mode and some are dramatically less flat. They can take on an almost bandpass response if a port is available and it is tuned to achieve this result.
The windowed near-field frequency response (figure 2) highlights the response at frequencies lower than the low frequency roll-off of the subwoofer. The response of the NHT rolls off quickly with a fifth-order response below 27Hz.
For a sealed system, a second-order (12dB per octave) roll-off is expected. The NHT is falling faster because a 2nd-order electrical high-pass filter in the DSP restricts the displacement of the cone. Non-linear driver behavior also results in a rolloff faster than the theoretical 12dB.
The slope of the rolloff is too fast for room effects to overcome the fast roll-off of the subwoofer. The in-room -3dB point could approach 24Hz, which is below the limit of standard instruments, although other rooms may just make the limit.
Figure 3 shows the full range near-field response with enhanced resolution of the vertical scale so the flatness of the response of the subwoofer is more apparent, as is the change in the movie mode. A filter in the DSP chip is programmed to create the shape of movie mode. Like the phase inversion switch, the switch for movie mode, on the back of the unit, sets a bit in the DSP. Make sure you have the switch in music mode when you calibrate with an electronic room EQ to prevent it from removing the peak. After calibration you can move the switch back for movie mode.
Figure 4 shows the near-field response of the subwoofer with the low-pass filter activated. This is a text book response: the cutoff setting of the control defines the -3dB point and the stop-band slope of a fourth-order (24dB per octave) is constant regardless of control position. Many subwoofers do not provide the correct transition band and stop-band shape for all settings of the cutoff knob. More electronics is required to keep the transition band shape independent of cutoff frequency. Recall that DSP filters in the system on a chip (SOC), used on the B-10d, produces these filter shapes. The control on the back sends a DC voltage to the DSP reflecting the user's desired rolloff frequency.
Using the LFE input the speaker remains flat to 140Hz and then rolls off monotonically. This is important for two reasons when using a room EQ. First, the bandwidth of the subwoofer during room EQ level calibration can produce an incorrect reading if the bandwidth of the subwoofer is too wide or narrow. Second, if the roll-off is not monotonic and exhibits output at higher frequencies, the room EQ may apply its second-order filters to clean up the response of the out of band response that could otherwise have been used to correct the modal response of the listening room.
Measuring Distortion in Subwoofers
Measuring distortion involves driving a low level voltage signal to the LFE input of the subwoofer and measuring SPL RMS out. Sound Pressure level (i.e., SPL, in dB) should not be confused with sound intensity, which is measured in watts per meter2. No power output is visible from the internal amplifier because it is internal to the system. When we make a distortion measurement at the speaker's output, we do not need to know if the driver is causing the distortion, if the amplifier is at the limit, or if electronic limiting that prevents damage to both the speaker and driver has been activated. Net-net, ignore power amp specifications supplied in subwoofer spec sheets.
THD versus level is a significant specification for a subwoofer. The Consumer Electronics Association (CES is only a small part of what they do) has developed a standard test method for a subwoofer (CEA 2010). Manufacturers of subwoofers for the home market have been slow to adopt it. Documentation is available from CEA at a cost; searching the internet provides a free summary.
CEA 2010 requires a ground-plane (microphone and speaker on the floor) measurement at 1 meter that approximates the response in an anechoic chamber, but results in measurements that are uniformly 6dB higher. Placing the subwoofer outside prevents room modes and boundary interactions with the room's perimeters from influencing the response when using the ground-plane technique. Setting up for CEA 2010 will, at a minimum, require a backyard. With the high level signals required for the tests, neighbors are not likely to allow this. I could not use the ground-plane technique and not just as a courtesy to the neighbors. My house backs up to a golf course.
The ground-plane technique has an advantage over near-field measurements where the microphone is placed near the cone(s) and then the port. Errors are introduced when combining the measurements with the near-field technique. Other systematic errors may be present in the near-field measurements that are not present in the ground-plane measurements. Ground-plane measurements are potentially sensitive to where the port is placed on the subwoofer relative to the active drivers and will produce different results dependent on how the subwoofer is oriented to the microphone. The engineer must angle the speaker so the measurement corresponds as closely as possible to the anechoic response.
CEA 2010 calls for special test tones and the resulting spectrum must fit below a mask that calls out limits from the second harmonic out to 1kHz. The tests at higher frequencies, outside the subwoofer's operational frequency, test for port noise or other out-of-band effects, such as cabinet resonance. Port noise is not an issue with the sealed NHT subwoofer. I did not observe any audible out-of-band cabinet resonance as I swept the frequency of a sine-wave oscillator at a high drive level because the LFE filter prevents internal sound pressure inside the box from occurring (see figure 4).
The shaped pulse in CEA 2010 may result in a higher measured SPL than a continuous-time distortion measurement. The pulse may replicate low real world performance with frequency transient events like a drum hit or car crash.
Since the ground plan measurements are 6dB higher than anechoic measurements care must be taken when comparing these measurements on a spec sheet that is not reporting distortion using the CEA 2010 standard. Some manufacturers (not NHT) report peak levels 3dB higher than RMS pressure on a level meter. In most cases, spec sheets fail to indicate whether the SPL is RMS or peak. The SPL numbers I report below are in RMS as would be measured in an anechoic chamber
To make my THD measurements, absent the ability to make a ground plane measurement, I placed the microphone in the near-field. Moving out to the far field in a room introduces room frequency response variations. How does this affect distortion? Consider what could happen as a result of room effects. The fundamental of the test signal could be in null with a response reduction compared to the average (say -3dB) and the second and third harmonics are located at peaks (say 6dB). The result is a reported distortion number amplified by a factor over 9dB.
Reporting the SPL measured in the near-field offers little information since it will be about 20dB higher than at 1 meter. Consequently, it is necessary to scale the SPL readings to represent the anechoic subwoofer output at 1 meter. When indoors, determining the scaling is difficult since the 1 meter SPL varies with frequency as a result of room effects. Absent an outdoors measurement, we must average the levels at 1 meter over several frequencies.
To get the most accurate scaling, NHT provided me with a specification to relate the input signal level at the LFE RCA plug (internal LPF disconnected as shown in Figure 4) to the anechoic subwoofer SPL at 1 meter. Their transfer function was applicable at one position of the level control, which they also specified. Adjusting the level control to the correct position, a 100mV RMS into the LFE input produced a 1 meter anechoic level of 90dB SPL RMS.
Since this is the first time I am reporting distortion in a loudspeaker let me provide a detailed roadmap for my approach. I first set the subwoofer level control to the specified value, supplied by NHT to produce 90dB RMS SPL peak level at 1 meter in an anechoic chamber. Next, I placed the microphone as close as possible to the driver to measure the near-field response that was not affected by the room boundary effects. (figure 1). For step 3, I turned to the Acoustisoft RplusD speaker measurement software. Its distortion measurement menu is in figure 6.
In this review, and others I have published, I have shown the performance of RplusD for making measurements in the time and frequency domain. RplusD also provides an affordable solution for readers who want to make distortion measurements for DIY projects. Distortion measurements can be made with something as simple as a Radio Shack sound meter stepping in for the microphone and mic preamp. I used my calibrated microphone and mic-preamp / USB sound card electronics designed exclusively for use with RplusD. RplusD produces tones for single tone THD, IM, or a special multi-tone measurement. RplusD does not do a sweep of THD with frequency. One measures each frequency independently, reads the THD value, and moves on to the next test frequency (figure 5). The panel shown above is set for an IM measurement. The two tones may be set to any values other than a harmonic multiple. Amplitude levels of the tones can also be set independently.
In its results panel, RplusD will not show the signal spectrum from the speaker; only the calculated values of the harmonics and IM spurs are displayed. The upper left side of the panel limits the frequencies used to calculate the distortion and allows measurement of higher-order IM products. The Tone 2 Level is set to a very low level for single tone tests. The panel does not tell me what value corresponds to 100mV RMS at the output of the sound card that drives the subwoofer. Instead, one measures the level at the subwoofer LFE input with a true RMS digital volt meter and adjusts the level value in the RplusD panel to produce 100mV RMS. Once calibrated, I created a 200mV signal by increasing the tone level on the RplusD panel by 6dB.
Now, the measurements proceed, and the distortion at each frequency is recorded. The microphone in the near-field is seeing a level about 20dB higher (it depends on the proximity to the cone) than at 1 meter. The SPL may be so high that nonlinearities in the microphone or mic preamp dominate over the subwoofers intrinsic distortion. I ran into this and had to displace the mic by two inches, thereby introducing a +/- 1.5dB error in the response of the subwoofer over its usable range. I judged this as inconsequential. All was well for the 100mV RMS test. Increasing the level to 200mV (96dB SPL RMS anechoic at 1 meter) did not result in mic or preamp distortion effects, but these effects did emerge at higher drive levels. This is why figure 5 is limited to a 96dB RMS SPL measurement.
NHT's engineers sent me a plot (Figure 7) of the distortion performance for the B-10d. NHT has highly-seasoned engineers who are well versed in the process and use different test equipment. The NHT plot provides data for 90dB SPL RMS and 100dB SPL RMS, which is 4dB higher than the limitation of my test equipment.
A comparison of the measurements in Figures 5 and 7 for 90dB SPL peak (1 meter anechoic) suggests my measurements are slightly more conservative, indicative of different samples of the speaker and the different test configurations. You can see my measurements at 96dB SPL RMS are below the 100dB RMS measurements supplied by NHT at a given frequency.
The dip in distortion at 50Hz, which my measurement system did not capture, is explained by Jay Doherty, Director of Engineering at NHT
"The dip in the distortion around 50Hz occurs when the impedance of the driver is at a maximum. The amp can provide more voltage where the impedance is high, so the subwoofer can play louder (increased SPL) in that region than otherwise would be possible. The DSP is programmed so the limiter value is increased in the frequency band of the impedance peak'
It is generally recognized that a 10% (-20dB) 2nd and 3rd harmonic distortion is not audible at low frequencies where the ear is less sensitive. Low-ordered harmonics are also masked by the human hearing system. Where the SPL level pushes the subwoofer to 10% is the tipping point. At 90dB SPL RMS, the subwoofer never hit 10% THD at 1 meter. At 100dB SPL RMS, the limit was 38Hz. I made a number of measurements of IMD at different frequency combinations and levels, but found no condition where the distortion differed significantly from the THD+N of the speaker at the lower frequency tone used in the IM test.
The distortion measurements are at 1 meter. If your listening seat is 3 meters at your back, the level is reduced by 9dB (the level is reduced by 6dB for each doubling of the distance in the far field); however, as explained in the introduction, most of the loss may be removed by room effects. Placement of the subwoofer in a room increases the SPL for 10% distortion compared to the anechoic numbers reported here provided the frequency is not at a null in the room's response. I again reference an article by Tom Nousaine which expounds on the issue of subwoofer response in real rooms with different subwoofer and seat placement:
As discussed in the introductory article Dr. Toole's text is also a good reference for single and multiple subwoofer placement (Toole, Sound Reproduction Loudspeakers and Room Focal Press. 2008; chapters 12 and 13).
I placed the subwoofer at the rear wall midway between the adjacent walls, and moved 3 meters back to my listening seat in a large room. I pushed the input level of the sine-wave oscillator to produce 100dB SPL at 30Hz on my sound level meter. From the measurements, taking into account the increased distance and room effect, this is the approximate 10% distortion level of the subwoofer at 30Hz. This was too loud for me without hearing protection, making assessment of the audibility of distortion impossible at this level. Lowering the level to a point at which I was comfortable revealed no audible distortion between 27Hz to 80Hz. Removing the subwoofer in one of my test setups that used a floor standing speaker in the $2000 range as the main channel did produce audible distortion below this level. I had to be careful not to overdrive with a tone too high in amplitude or too low in frequency to prevent damage. Smaller bookshelf speakers, run full range, showed more degradation in the SPL limit and minimum frequency before the cone movement (judged by feeling the cone) was becoming excessive.
NHT sent me some measurements at 117dB RMS SPL at 1 meter anechoic. These showed distortion at less than 20% from 40Hz to 100Hz, the level the limiters become active. This measurement is more relevant to the reproduction of sound effects than music.