I don't believe that I was dismissing the step test at all. I was just questioning aspects of its efficacy for the job it was being put to. I can only agree that the step response is but one of many different valid tests of a loudspeaker.
Consider a signal whose dominant frequency content occurs at Fc, the resonance frequency of an example 2nd-order closed-box subwoofer loudspeaker system.
The point I was trying to make was that a Q=0.5 system at Fc has a response that is −6dB, while a Q=0.7071 system has a response that is −3dB at Fc. Although energy storage is happening, energy release is also occurring, otherwise, there would be no sound output (which is probably a bad thing for a loudspeaker, but I'm not entirely sure of my understanding 😉).
In the above example, the Q=0.50 system has 50% less power being output than the Q=0.7071 system, which psychoacoustically makes it seem like the Q=0.50 system possesses "tighter bass". However, it has distorted (lowered the intensity) of the output signal significantly in achieving that result. Any differences in "transient response" between Q=0.5 and Q=0.7071 are marginal, but that 50% decrease in power output is highly evident from both a hearing and tactile response perspective, and it reduces the accuracy of the system on music program material instead of enhancing it
Last edited:
Pre-ring is not inherent to FIR filters, just to linear-phase FIR filters. An FIR filter can also be, minimum-phase, non-minimum-phase or anything in between. Linear-phase requires that the filter kernel’s coefficients be symmetrical, which produces a symmetrical impulse response, featuring equal pre and post ringing.
The pre-ringing is what makes the filter linear-phase, so it is not inherently a problem. Anecdotal experiments appeared to indicate that the problem of unsatisfying ‘digital sound’ was related to how most FIR filters are implemented, and not because it featured linear-phase pre and post ringing.
Last edited:
You can compensate for pre-ring but if you drop the linear-phase of the filter then there's not one argument left to use FIR AT ALL!
Yes, it is! Since if the linear phase argument drops, there's not one single argument left to use FIR whatsoever.
For once i strongly disagree with you ICG: the way to go over compression ( mitigate it?) Is simple, just increase SD to the point you reach your target SPL at lowest freq planned and forget about huge power and amp ( well not really i must admit... still 500w will be commonly reached if you want 'standard' 83dbspl at listening location in medium room, that said it's mainly because of the transient nature of 'natural' low end and to be sure it can withstand a Drake or dubstep session... lol).
Something like what MrKlinky presented before and being half moked for. Of course you'll loose the advantage of cabinet size but it's a compromise i'm ok with ( ymmv of course!).
LT is theorical ok... but results using eq to compensate QTC are not.
The thing is there was enough example i met in control rooms to know it's not theorical but it works ( great) in practice. I think about NHT monitors and subs which were vastly superior to other commercial offer of that time on this point ( some 20 years ago) or other custom solutions i've seen used on the same principle.
Nht had access to drivers developped for them so they had an advantage over what we can do about this point, but we are diyers and can use other strategy to mitigate or even push boundary even more imho (talking about boundary... why not use them to increase output? You know, Allison's way 😉 ).
And i don't get the connector resistance as an issue either: you need to measure acoustic output to validate your eq so it's all taken into account in the end ime.
Of course amplifier power request end up being 'big' ( depend on your target spl, room,...) but nothing really different than a typical powered sub.
Don't take me wrong i'm not saying it's a cure for all disease or it doesn't have drawbacks but with a bit of planing it can ends up being very succesful approach and can rival other approach.
Still unconvinced, take a look at D&D 8c and size of the sub dedicated drivers... I think it's in Mitchba's test where i read he was surprised to not run out of headroom with them and they were on par with his horn loaded+2x15"+sub own system at it's regular listening level.
About preringing: yes it is present, yes it's an artefact. But like many users of FIR i now think there is a treshold to audibility of this and that if you don't use 500db/oct steep filters it is largely inaudible.
At 48db/oct in lows or >100db/oct in high freq i was never able to identify it with my Lake nor others which heard my mains ( but musicians had been bothered by latency when monitoring real time. Obviously! 🙂 ).
Last edited:
I agree on this completely! The problem is not that there's something wrong with LT but with how ppl implement it. And that was exactly my argument, nobody included the real life situation like the contact resistance etc, I insisted it has to be included in the formula but he was adamant his concept was flawless. If you ignore such an important fact then it isn't! I've included power compression etc just to show how much he ignored and that a big name doesn't 'prove' something is perfect! In a home subwoofer it often does not matter how much the power compression might apply but you have to be aware of it!
To have a great reputation or being dead does NOT make everything you did right nor perfect!
About the pre-ring: I don't claim it's significant audibly but to claim it's not existant and try to claim it depends on the filters is just plain wrong! That is a drawback of the principle and ignoring it doesn't make it go away! Note that I don't claim it is significant but to tell it's not there or depends on the frequency etc is something I'm viciously opposing!
Discuss the magnitude I'm perfectly fine with, I'm not boasting it's the end of the world, defending it not existing, I'm definitely not!
Maybe that's one drawback of using subwoofer(s), because it needs also a lowpass filter to crossover to main speakers, right in the middle of the bass range, which does similar things like a highpass filter.
@mark100 Many thanks for going to all this effort to get a nice set of measurements.
Below is the simulated response obtained when using a 4th-order Butterworth high-pass filter with an f3 of 27Hz. For the input signal, it is worth noting that the negative values on either side of the peak are due to the flat-top window that was applied.
The results shown below are somewhat similar to your measured response. As my results appear to be polarity inverted relative to yours, is there a chance that your measurements were obtained using an inverting amplifier?
Below is a simulation of a cosine-windowed 2.5-cycle 27Hz toneburst filtered by a 4th-order Butterworth filter with f3 at 27Hz. The simulated transient response dies away more quickly than the measured one.
Below is the simulated response to a 2.50-cycle 60-Hz toneburst with a cosine window applied when passed through a 4th-order Butterworth high-pass filter with –3dB cut-off at 27Hz. This result seems to be quite different to the one that was measured during the experiments.
For reflex and horn systems played at high SPLs, a steep high pass just below system resonance is essential.
But this adds even more group delay.
I found an elegant way to accomplish this with a 2nd order shelf filter, which (1) in some alignments is more effective at limiting excursion than a 4th order high pass, and (2) has half the phase shift of a 4th order. The end result is a 6th order system with 5-10dB higher max output than not using a filter.
I describe it in an AudioXpress article called "The DSP Assisted Reflex" and @pelanj discusses his application of it on DIYaudio here.
But this adds even more group delay.
I found an elegant way to accomplish this with a 2nd order shelf filter, which (1) in some alignments is more effective at limiting excursion than a 4th order high pass, and (2) has half the phase shift of a 4th order. The end result is a 6th order system with 5-10dB higher max output than not using a filter.
I describe it in an AudioXpress article called "The DSP Assisted Reflex" and @pelanj discusses his application of it on DIYaudio here.
Last edited:
Late to the party...is anyone have the psycho acoustic research that backs up the OP's claim?
Where is the title's claim from?
Where is the title's claim from?
here's a myth buster example
An electrical 3-way using xovers at 300Hz and 3KHz. Xovers are 96dB/oct complementary linear-phase
First, the 3-way's impulse. Note that the summed sections' response is essentially a textbook perfect Dirac pulse.
Mag and phase traces are shown for each of the sections.
Nest, summed step response.
There's clearly zero pre-ring in either impulse or step.
The key to that is fully complementary coupling between sections. Which is of course easy in 1D electrical space. And obviously more difficult in 3D acoustic space.
Fortunately, the only frequency ranges subject to less-than-fully-complementary summations, where pre-ring has potential, are the critical crossover summation regions. This is why a tone burst, or music content in general, that is not inside a critical xover summation region is immune to pre-ring potential.
And to the extent the speaker's acoustical design minimizes off-axis lobing, pre-ring potential in the critical xover summation regions drops towards nil.
A further aid in mitigating off-axis lobing potential, is steep xovers, which lower the width of frequency range subject to lobing.
Flat phase, reduced acoustic lobing, minimal if any acoustic pre-ring potential......
....what's not to like about steep complementary linear-phase xovers other than latency...
Not true in my experience.
Like Ken pointed out in #143, FIR can be used as anything, linear-phase, minimum-phase, maximum phase....combo phase.
For live-sound work where latency must be kept low, I use FIR in minimum-phase mode.
The advantages over slugging it out solely with IIR are two-fold.
First, a good auto-FIR generator will acoustic target match minimum-phase xovers and EQ's, with what amounts to having an essentially an unlimited number of filters available for corrections. Operator simply applies judgement to how close or loose they want to match to the acoustic target.
What DSP has an essentially unlimited number of IIR filters available for use? With FIR, you have that.
Second advantage, a FIR file works the same on any processor. IIR filters are processor specific and need be translated one to another.
Big pain in the butt.
Hey, I'm blessed with a processor that has decent FIR capability and even better IIR capacity.
I continually try out strategies using one or the other, most often a combo-of both (due to sometimes needing to minimize latency like for live, and always the need for the system high pas to be IIR).
FIR wins easy once above sub range, ime, ......
Hi witwald, I have to run for a while...will get back to your posts as soon as i can...thanks !
OK, then use eq. Or compensate with room acoustics. The audible difference is not due to levels in output which can easily be adjusted but due to the Q of the system. I’m sure many people are hearing less bass and think it’s due to lower Q or ‘faster’ woofers. But, as you pointed out, they are hearing the wrong thing. There is a very easy experiment you can do; build two boxes one for a Q of .5 and the other for a Q of .7. You can clearly hear the difference, well, I could hear the difference. And, no it wasn’t due to the volume. I equalized the levels. I could clearly hear ringing in the bass on staccato notes and bass drum. For movie soundtracks a higher Q might help with the rumble and sensation of low bass, but for music I prefer a Q of .5 for my subwoofers. I think it’s more accurate and more lifelike, which is the point of hifi.
I’m not sure how electronics affects all of this since the subwoofer still has a system Q, and I haven’t messed around with it much.
The title of this thread came from my observation of comments about Q factor on this forum, where some people say in their opinions that a Q of 0.5 sounds thin.
In itself, this thread is to discuss why some people consider a Q of 0.5 to be thin and weak, but it seems that this has already been resolved in this thread.
Really, as some have said here, without knowing the room, the system and the type of listening (such as: critical listening, non-critical listening, etc) of the listener, this question is somewhat or completely vague.
[Other] As I said in my first post, I consider Q 0.5 ideal for almost all situations, due to its characteristics of better step response and group delay. It may not be ideal for very large environments or outdoors, but nothing that a light equalization can't solve.
But there is, FIR filters are the best at delivering a sharp transition-band. Which is important for objectively separating an upper pass-band edge from a closely located lowest image frequency. Which is a narrow (22kHz - 20kHz = 2kHz) for 44.1ksps CD. It’s the subjective necessity of such a sharp transition-band for in-home music reproduction that is debatable.
I personally feel that it’s the subjective performance of the interpolation filter (if any) that most matters with music. Non-scientific listening experiments have led me to conclude that very high-performance FIR interpolation (such as, the PGGB s/w interpolator) and NOS (no interpolation filter at all) sound much more alike, and also more like natural sounding music than either sounds to ubiquitous, on-chip, half-band FIR filter. Indicating that typical, on-chip, FIR filters are less than audibly transparent with music, and so are inadequate.