I'm talking about a closed enclosure subwoofer. Traditionally, most people choose to build an enclosure with Qtc of 0.707 or Butterworth characteristic. I wonder what if I build the sub in other alignments. Two following choices are selected for comparison: Qtc of 0.5 and 1.0.
Has anybody ever heard the same drive unit in those two different alignments, Qtc = 0.5 and 1.0? How would you describe the difference of sound between them?
Finally, which one do you prefer?
Has anybody ever heard the same drive unit in those two different alignments, Qtc = 0.5 and 1.0? How would you describe the difference of sound between them?
Finally, which one do you prefer?
Well, the sound differences simply follow what the frequency response suggests.
What is more interesting is comparing different Q values and resonance frequencies (as dialed in by cabinet volume together with amplifier output impdedance), after EQing them all to the exact same frequency response.
As expected, to first or even second order they sound the same as they measure the same by definition. Assuming reasonable Qs with reasonable box volumes, there is not much difference in distortion either.
But looking closer, there is a slight difference. Whenever the cone does not exactly move as intended, we get an error signal. And this error signal undergoes the natural Q of the speaker, regardless of how it was EQd.
That means if you have a natual Q of 2, the speaker will ring with this Q in the error recovery (the simplest 'error signal' test is the knuckle test). That why I prefer to keep the Q at ~0.6. This gives a Bessel lowpass behavior on cone excursion, reacting as fast as possible to any error signal step change with no overshoot/ringing.
What is more interesting is comparing different Q values and resonance frequencies (as dialed in by cabinet volume together with amplifier output impdedance), after EQing them all to the exact same frequency response.
As expected, to first or even second order they sound the same as they measure the same by definition. Assuming reasonable Qs with reasonable box volumes, there is not much difference in distortion either.
But looking closer, there is a slight difference. Whenever the cone does not exactly move as intended, we get an error signal. And this error signal undergoes the natural Q of the speaker, regardless of how it was EQd.
That means if you have a natual Q of 2, the speaker will ring with this Q in the error recovery (the simplest 'error signal' test is the knuckle test). That why I prefer to keep the Q at ~0.6. This gives a Bessel lowpass behavior on cone excursion, reacting as fast as possible to any error signal step change with no overshoot/ringing.
Hi, do you have any error signals in mind what this helps for?
I mean is there any typical phenomenon that occurs, which then makes one Q alignment sound different from another when EQd the same? or is it dependent on driver or application for example? i'm interested if high Q was always bad or just in some cases, and then of course learn what those are 🙂
I mean is there any typical phenomenon that occurs, which then makes one Q alignment sound different from another when EQd the same? or is it dependent on driver or application for example? i'm interested if high Q was always bad or just in some cases, and then of course learn what those are 🙂
using digi EQ I had great freedom of low frequency design options.
If Q is too low you simply don't hear the bass.
If its too loud the sound gets slow and fat. If its lower it can sound fast and clean until its too low.
This is in an area of plus minus 2db.
After some time you find optimal loudness. Once you got used to the freedom with EQ I wonder how to make a high end speaker without.
Sometimes your ears play a game and then I turn some db up or down at certain frequency bands depending on daily listening condition. Mainly bass and highs because for the mids the tolerance is smaller like + - 1db.
I like for example having the mids attenuated 2 octaves wide by a half db. Sounds a bit more vivid.
If Q is too low you simply don't hear the bass.
If its too loud the sound gets slow and fat. If its lower it can sound fast and clean until its too low.
This is in an area of plus minus 2db.
After some time you find optimal loudness. Once you got used to the freedom with EQ I wonder how to make a high end speaker without.
Sometimes your ears play a game and then I turn some db up or down at certain frequency bands depending on daily listening condition. Mainly bass and highs because for the mids the tolerance is smaller like + - 1db.
I like for example having the mids attenuated 2 octaves wide by a half db. Sounds a bit more vivid.
One very typical error signal is dynamic DC offset, the tendency of the voice coil getting sucked in or popping out, which happens at medium to strong excursion at certain frequencies (IIRC, 2*fs is a problematic area as per some Klippel papers).
When you remove the drive signal the cone will move to its rest position and that movement is governed by the natural Q, not the Q of EQ'd SPL target. So if you have a high natural Q like 2 (pretty extreme, admittedly), the woofer would likely produce the dreaded "one-note bass" sound, ringing at resonance all the time when even just slightly overdriven. Behavior in, and recovery from full excursion overdrive is equally affected.
By chosing a moderate natural Q like 0.5...0.8 the woofer will return to nominal position as fast as possible with almost no "overhang". I wouldn't expect any significant differences -- after target EQ -- between a 0.5 and a 0.8 design goal for natural Q (if not constrained by other important considerations). Going much lower and increasing feedback (via negative output impedance, for example) might improve things or make them worse. Too much variables involved here. And sometimes real surprises.
- If you want a boomier sound, go for the QTC at 1.0 (Chebyschev), with the best efficiency.
- If you want a deeper sound, go for 0.707 (Butterworth), which is a compromise.
- With a QTC at 0.5, the bass will be subdued but even deeper, with the best transient.
These are general guidances, of course. The T/S parameters of the speaker you intend to use, combined to the value of the QTC can give you very different bass and transient responses, and indeed volumes.
QTC at 0.5 can give you very huge volumes, impossible to deal with in a living room...
QTC at 0.71 is a good - an possibly affordable - compromise between bass extension, transient response and volume.
QTC at 0.8 to 1 is a frequent value found in commercial sealed enclosures : "sellable bass" and moderate volume !
If I had to build a quite large sealed enclosure, I would go for a QTC at circa 0.71. That's what I choose for my 475L Isobaric project (Ext. HxWxD = 900x400x380mm) :
If it is for a Subwoofer... It depends on the place that you can deserve to it ! I think that QTC at 0.8 to 1 would give you more reasonable cabinet sizes - but again : it depends also on the T/S parameters and the kind / tone of bass extension you expect...
Personally, I went for the Ripole principle for my Subwoofer :
Deep bass extension, excellent damping, no boxy sound, very compact but moderate efficiency : IMHO not for Home Theater, but only music...
OK : it's another story, different from the subject of the thread : closed boxes ! 🙂 😉
Qtc 0.5, lots of cone area, and throw heaps of power and EQ at it. Cleanest and nicest-sounding bass I have ever heard, and not just because I built them!
Simulate them. The diagram you have shows them at one frequency but when you change the box volume the frequency should change too.
This is an interesting question, and I've tried to do some simulations assuming that the closed-box loudspeaker is a distortionless, linear time-invariant system. Such a system is relatively amenable to some straightforward modelling and analysis.
In the frequency response curves shown below, we have a closed-box system with resonance frequency f0 and a number of selected values of Qtc.
Let's take a look at the transient response of this system when excited by a 1.5-cycle tone burst with a frequency equal to f0. For the purpose of the simulation, f0 is chosen to be 50Hz, which is well within the capabilities of a closed-box loudspeaker system of moderate size. The transient responses are plotted for Qtc values of 0.5, 0.7071, 1.0, 1.4142, and 2.0.
Transient response for Qtc = 0.5
Transient response for Qtc = 0.7071
Transient response for Qtc = 1.0
Transient response for Qtc = 1.4142
Transient response for Qtc = 2.0
It seems quite apparent that, apart from the amplitude differences associated with the −3dB point of each Qtc frequency response curve, the majority of the transient responses are quite similar. The transient response for Qtc = 2.0 tends to oscillate most noticeably for a while after the input signal stops.
If we now go ahead and remove the amplitude differences, what will we see? The following plots show the transient responses where each of them has been normalised to a maximum absolute amplitude value of 1.
Once they are normalised, it seems that the transient responses for Qtc = 0.5, 0.7071, and 1.0 are very similar. It's only the ones for Qtc = 1.4142 and 2.0 where the post-input ringing becomes more noticeable.
Is there a sufficient difference in the results obtained for Qtc = 0.5, 0.7071, and 1.0 in order for the differences to be audible? I'm not sure. However, without normalisation, there would likely be amplitude-related audible differences, which would swamp any effects of the differences in the dynamic behaviour associated with the transient response itself.
Normalised transient response for Qtc = 0.5
Normalised transient response for Qtc = 0.7071
Normalised transient response for Qtc = 1.0
Normalised transient response for Qtc = 1.4142
Normalised transient response for Qtc = 2.0
Once they are normalised, it seems that the transient responses for Qtc = 0.5, 0.7071, and 1.0 are very similar. It's only the ones for Qtc = 1.4142 and 2.0 where the post-input ringing becomes more noticeable.
Is there a sufficient difference in the results obtained for Qtc = 0.5, 0.7071, and 1.0 in order for the differences to be audible? I'm not sure. However, without normalisation, there would likely be amplitude-related audible differences, which would swamp any effects of the differences in the dynamic behaviour associated with the transient response itself.
Normalised transient response for Qtc = 0.5
Normalised transient response for Qtc = 0.7071
Normalised transient response for Qtc = 1.0
Normalised transient response for Qtc = 1.4142
Normalised transient response for Qtc = 2.0
Nice comparisons Witwald.
The stop time also relates to the time getting started (building up levels in the resonance) since it relates to a change in either direction. What I'd imagine seeing, though it would take considerable time to prepare, is an animated flow diagram or colour coded dynamic schematic.
The stop time also relates to the time getting started (building up levels in the resonance) since it relates to a change in either direction. What I'd imagine seeing, though it would take considerable time to prepare, is an animated flow diagram or colour coded dynamic schematic.
Thank you first to Witwald for those visually lucid simulations.
I remain convinced that a short transient decay in subwoofers is largely what makes them - and excuse me for using these terms which I hate - 'tight' and/or 'fast'.
Much is made of waterfall plots when we go further up the spectrum where shorter decay is better, but little mention ever seems to be made of quick decay times of woofers and subwoofers. The decay in the Q = 2.0 looks somewhat like that in systems having high energy storage: ported; horn loaded; tapped horn etc, and which makes them all, to my ear, sound very muddied in the time domain.
Finally, I was told decades ago by a very experienced British speaker designer that the compression ratio of sealed enclosures (Vd/Vb) should not exceed about 5% before audible non-linearity becomes problematic. In reality this is quite difficult to exceed even at Q=0.5, except with drivers having excessive Vd for their size such as in the car audio scene. The 18s in my 55l subs are not the last word in excursion (8mm), and manage less than 2% at Xmax.
I remain convinced that a short transient decay in subwoofers is largely what makes them - and excuse me for using these terms which I hate - 'tight' and/or 'fast'.
Much is made of waterfall plots when we go further up the spectrum where shorter decay is better, but little mention ever seems to be made of quick decay times of woofers and subwoofers. The decay in the Q = 2.0 looks somewhat like that in systems having high energy storage: ported; horn loaded; tapped horn etc, and which makes them all, to my ear, sound very muddied in the time domain.
Finally, I was told decades ago by a very experienced British speaker designer that the compression ratio of sealed enclosures (Vd/Vb) should not exceed about 5% before audible non-linearity becomes problematic. In reality this is quite difficult to exceed even at Q=0.5, except with drivers having excessive Vd for their size such as in the car audio scene. The 18s in my 55l subs are not the last word in excursion (8mm), and manage less than 2% at Xmax.
I share your reluctance to use terms like this.
What does it actually sound like when you get the bass properly flat using a multi-sub arrangement? Tight and fast? Loose and slow? No, it simply sounds like the original instruments (and I wouldn't want it any other way).
Here is the controversial opinion.
As long as the speaker has suitable cone displacement volume, amplifier can supply enough voltage, and DSP is at hand, get the lowest Qes and Qms driver you can. Qms is quite tricky, as it can rob you from efficiency, not so with Qes.
What low Qes basically means is that it has high motor force. Meaning, the motor pushes the most Newtons per Watt at the cone. And there is no reason to not want that in general. If it was feasible, I'd buy bunch of 21IPALs.
As long as the speaker has suitable cone displacement volume, amplifier can supply enough voltage, and DSP is at hand, get the lowest Qes and Qms driver you can. Qms is quite tricky, as it can rob you from efficiency, not so with Qes.
What low Qes basically means is that it has high motor force. Meaning, the motor pushes the most Newtons per Watt at the cone. And there is no reason to not want that in general. If it was feasible, I'd buy bunch of 21IPALs.
Edgar M. Villchur (Acoustic Research) published an article titled "Another Look at Acoustic Suspension" in the January 1960 issue of AUDIO magazine (here).
In that article, he noted that when a volume of air is compressed and expanded it remains sufficiently linear under a volume change of ±5%. He also mentioned that the actual maximum air volume change in an AR-1 acoustic suspension loudspeaker cabinet was ±0.75%, which is one-sixth of that 5% range. This seems to tie in with your experience.
It finally dawned on me that I have some relevant audio samples that can be used to assess the audibility of variations in transient response.
Attached is a 96kHz Audacity project file that contains the transient responses of a 1.5-cycle 50-Hz tone burst (apodized using a Tukey window, r = 0.2) that has passed through 50-Hz 2nd-order high-pass filters with Q = 0.5, 0.7071, 1.0, 1.4142, and 2.0. The individual transient responses have been normalized so that they contain the same maximum magnitude, which it is hoped helps to make them sound equally loud. The individual WAV files are also supplied in another ZIP file for those wishing to use a different DAW.
The results for Q = 0.5, 0.7071, and 1.0 are the ones that sound much the same to me.
Last edited:
This is what I realised too during my ht build. Have as many as sealed sb bass drivers as possible for a smoother and cleaner bass. As individual cones have to move less vs 1,2 drivers. Now, towers with multi drivers for bass are getting my attention like the vintage IRS-V. May be I need to build one as large as this one day 😀
Attachments
It means not just high motor force but also high damping, which in turn can be so high that the driver can loose efficiency on its low-end. This can be seen even in a video that Bennett Prescott (of B&C) posted on youtube. If I interpreted correctly the graphs.
Sensitivity, not efficiency. I am afraid you didn't get that one correctly or it was presented in such a way it's not that easy to understand.
Yes, one has to drive the speaker with higher voltage, but that doesn't mean higher power compared to weaker motor driver.
Just imagine it. Layman terms, common sense. Less magnet and wire, less boom. It IS that simple (if all else is equal and the amp can deliver the voltage).
Yes, one has to drive the speaker with higher voltage, but that doesn't mean higher power compared to weaker motor driver.
Just imagine it. Layman terms, common sense. Less magnet and wire, less boom. It IS that simple (if all else is equal and the amp can deliver the voltage).