DSP can get you pretty close, provided you can work with your room.
Transient or time coherent behaviour was high on my list of things to try (with sealed full range line arrays).
This is a graph showing how close I got:
This is a true measurement at the listening position.
(FIR) EQ is used to lift up the bottom end, 50x 3.5" drivers do all the work.
The step of left and right speakers combined at the listening position:
I have a downward sloping FR graph and corrected that to a minimum phase band pass behaviour.
Not by forcing the Phase, that wouldn't work well, but by treating first reflection points in the room
and correcting phase by evaluating the data in spectrogram and waterfall plots of that first wave front.
I left room effects in my left channel and right channel unchanged. Together they sum up to this coherent
shape. The biggest problems being a 60Hz resonance in the left channel (due to corner placement) that
is compensated for in the right channel.
Last edited:
Joined this thread late, but some comments on the latest additions...
I disagree entirely. Removing group delay via digital filtering has significant audible benefits - at least to my ears.
No. Localisation at low frequencies is perfectly fine, just compromised typically by room acoustics and increasing group delay due to the inevitable roll-off. For stereo we might also consider the effect of the 6dB/octave roll-off in the difference signal that is seldom compensated.
The effect is dominated typically by the system Q. Reflex (and other) designs are typically designed for maximally flat frequency responses which is not the minimally audible distorting response. The audible effects are different and separable.
You have the terms transient and attack around the wrong way. Transient describes the leading edge of a waveform and applies across all frequencies. Attack is a 'physco-acoustic' term that is generally effected by HF/MF effects.
Yes it can.
On an analyser yes, but audibly, system Q is the dominant factor.
Yes there is. I have found equalising group delay to 5Hz via the application of DSP to be very effective. Not only are bass transients perceived 'faster' (I hate the term too!) but for a system roll-off tuned to 40Hz, the effect of removing group delay I find much the same as listening to a similar Q speaker with a roll-off approaching 20Hz.
Of course there is a proviso to add here that such equalisation incurs a long delay that is sufficient to cause lip sync errors in video replay applications. But if you have the DSP platform available then I urge thread readers to try it. IMHO it is one of the few remaining secrets of the audio engineering world that the DIYer can exploit...
Firstly why would you need to? But secondly it depends on the method employed. Even a single, non-strategically placed speaker will produce a relatively large region of equalisation around a listener/measurement point at the low frequencies of interest here. Furthermore the room modes of interest will typically exhibit (in isolation) a minimum phase response and so be perfectly invertible.
Yes you can - but as with most applications of DSP it is better to sort such aberrations at source.
Again overall system Q is the primary audible factor, group delay second - plus some other oddities due to reflex tubes and the like. Room resonances are linear and so separable and not relevant to the argument here.
I disagree entirely. Removing group delay via digital filtering has significant audible benefits - at least to my ears.
No. Localisation at low frequencies is perfectly fine, just compromised typically by room acoustics and increasing group delay due to the inevitable roll-off. For stereo we might also consider the effect of the 6dB/octave roll-off in the difference signal that is seldom compensated.
The effect is dominated typically by the system Q. Reflex (and other) designs are typically designed for maximally flat frequency responses which is not the minimally audible distorting response. The audible effects are different and separable.
You have the terms transient and attack around the wrong way. Transient describes the leading edge of a waveform and applies across all frequencies. Attack is a 'physco-acoustic' term that is generally effected by HF/MF effects.
Yes it can.
On an analyser yes, but audibly, system Q is the dominant factor.
Yes there is. I have found equalising group delay to 5Hz via the application of DSP to be very effective. Not only are bass transients perceived 'faster' (I hate the term too!) but for a system roll-off tuned to 40Hz, the effect of removing group delay I find much the same as listening to a similar Q speaker with a roll-off approaching 20Hz.
Of course there is a proviso to add here that such equalisation incurs a long delay that is sufficient to cause lip sync errors in video replay applications. But if you have the DSP platform available then I urge thread readers to try it. IMHO it is one of the few remaining secrets of the audio engineering world that the DIYer can exploit...
Firstly why would you need to? But secondly it depends on the method employed. Even a single, non-strategically placed speaker will produce a relatively large region of equalisation around a listener/measurement point at the low frequencies of interest here. Furthermore the room modes of interest will typically exhibit (in isolation) a minimum phase response and so be perfectly invertible.
Yes you can - but as with most applications of DSP it is better to sort such aberrations at source.
Again overall system Q is the primary audible factor, group delay second - plus some other oddities due to reflex tubes and the like. Room resonances are linear and so separable and not relevant to the argument here.
Last edited:
Comment to bolded sentence:
Doesn't this alone really tell that the difference you heard had nothing to do with group delay?
Plus the group delay is actually very high in most horn subwoofers. Tapped horn has for instance clearly higher group delay than a vented box. I thought front loaded horn also had very high goup delay but perhaps it varies.
Another point. I find it unlikely that group delay, if to contrary belief it is really audible, that it can be audible in low freuencies before one has a very short decay in the room. Something almost no audiophiles have becuase most don't treat this area seriously. Indicating that any potential audible group delay would be masked by too long decay and ringing anyway. In other words, what might be audible with headphones may be very different from an actual room.
How are you EQing this? Just shaping frequency response? So just shaping a ported bass response to a target sealed response with desired Q?
Can you elaborate on that?
Can't suppress a chuckle when an abstract and logical-seeming argument begins with, "Our monkey ancestors needed this-or-that skill in order to hear fruit falling from trees, therefore we need....." and goes on to further conclusions arising from abstract logic rather than behavioural evidence.
BTW, big difference between someone saying " "fast" bass " versus somebody saying " fast bass ". Sad when people can't tell the difference.
B.
If one plays a 'live' drum set in the room, would it sound fast?
To get close to that, your speakers would need to be able to reproduce that impulse. However way you look at it, (excessive) group delay subtracts from that requirement.
I've shown what I think "fast Bass" means. It will sound different from a system with more group delay. I've had it adjusted to be even more linear phase down to lower numbers. That didn't sound as "natural" to me as having it follow the band pass behaviour of the frequency response.
The frequency extends to well down below 20 Hz in my case, though I couldn't play a full force note there due to lack of amplifier power and next would be x-max. I'm able to hit the notes found in regular music just fine, it's the Home Theatre stuff where I stumble upon its limits.
Basically I agree with (the late) John Dunlavy: http://www.stereophile.com/interviews/163/#M07YTT3Jclbvqycm.97
You've got to be careful though, adjusting the phase response of the speakers, not the room. The room will have to cooperate to make it work.
To get close to that, your speakers would need to be able to reproduce that impulse. However way you look at it, (excessive) group delay subtracts from that requirement.
I've shown what I think "fast Bass" means. It will sound different from a system with more group delay. I've had it adjusted to be even more linear phase down to lower numbers. That didn't sound as "natural" to me as having it follow the band pass behaviour of the frequency response.
The frequency extends to well down below 20 Hz in my case, though I couldn't play a full force note there due to lack of amplifier power and next would be x-max. I'm able to hit the notes found in regular music just fine, it's the Home Theatre stuff where I stumble upon its limits.
Basically I agree with (the late) John Dunlavy: http://www.stereophile.com/interviews/163/#M07YTT3Jclbvqycm.97
You've got to be careful though, adjusting the phase response of the speakers, not the room. The room will have to cooperate to make it work.
Last edited:
There are two separate discussions here...
Firstly there is an option to correct for a non-ideal system Q. Once the order is fixed, this can simply be a case of correcting a frequency response.
Secondly there is the option to correct for group delay. This requires a relatively long FIR filter capable of reversing the amplitude response - one of the few tasks we can definitely say that DSP is best for.
Once group delay is compensated, there is the option to implement a linear phase roll-off - and therefore one where Q is irrelevant. Instead we can adopt a maximally flat magnitude response without the problems of its minimum phase counterpart or make maximum use of the available driver displacement (which happen to be the same for a second-order system).
Although many cite problems with linear phase correction, at very low frequencies where the wavelengths are long, audible pre-ringing does not appear to manifest itself as it can do due to incomplete off-axis summations with linear phase crossovers, for example.
For a minimum phase roll-off, for both the cases of compensating for the system alignment and for compensating for the group delay, I would personally advocate a Bessel alignment in order to best preserve transient accuracy and minimise group delay distortions. I believe other system alignments best serve in making nice looking magnitude plots and technical specifications for sales people.
Perfectly answered!
There is also the dimension of peak SPLs to build in here too - as you point out - even for a grand piano let alone a drum kit. Peak limiting/clipping (devoid of consequences due to bad engineering) normally imparts a subjective dullness to audio reproduction regardless of Qts and group delay. Some might describe this as "slow"...
I'd like to interrupt this abstract discussion (however beneficial it is to have good theory... in its place) by talking about wood speakers.
Alas, all but universally, we play bass by obliging speaker systems to hoot through their resonance. When a BR box gets a signal in the vicinity of the somewhat broad and intrusive system resonance, it adds it own resonance to the music. As far as I know, you can't subtract something from the music signal to stop it from adding that tubbiness*. Advocates of using DSP to do so are, in my view, mistaken. The tubbiness is in the physical box design and won't go away by changing the input signal. (I think there is some confusion between shaping frequency response which is easy to do, and controlling resonance. And that arises because sims are sharply focussed on FR alone.)
"Tubbiness" is a suitable description and a more honest term for the sanitary-sounding euphemism, "group delay". But with kludges like an "aperiodic" port, you can tame the tubbiness.
I don't think you can make a BR box (or any box except pure true horns) without that broad resonant zone falling within your musical compass. A sealed box can keep it pretty low and damped.
Ben
*but motional feedback can correct it.
Alas, all but universally, we play bass by obliging speaker systems to hoot through their resonance. When a BR box gets a signal in the vicinity of the somewhat broad and intrusive system resonance, it adds it own resonance to the music. As far as I know, you can't subtract something from the music signal to stop it from adding that tubbiness*. Advocates of using DSP to do so are, in my view, mistaken. The tubbiness is in the physical box design and won't go away by changing the input signal. (I think there is some confusion between shaping frequency response which is easy to do, and controlling resonance. And that arises because sims are sharply focussed on FR alone.)
"Tubbiness" is a suitable description and a more honest term for the sanitary-sounding euphemism, "group delay". But with kludges like an "aperiodic" port, you can tame the tubbiness.
I don't think you can make a BR box (or any box except pure true horns) without that broad resonant zone falling within your musical compass. A sealed box can keep it pretty low and damped.
Ben
*but motional feedback can correct it.
Last edited:
Hello, poor transient response does not mean higher group delay. As someone already wrote here, group delay is a time, and a constant delay of all frequencies does not impair transient response. (It does impair time response, tho.) It is called group delay and not delay time because it is measured by running a cosine-amplitude-modulated sine, say a packet which contains mostly one frequency, thru the system and measuring time, until the packet arrives.
Transient response becomes distorted by group delay, which is not constant. Even a filter of first order does this, but the sum of each one hi- and lopass of first order with same corner frequency, say a cross-over of first order, does not.
Hipasses introduce delay within the stopband, lowpasses within the passband, allpasses belo mid frequency. Hence in a classical electrical system, hi frequencies may be damped but are not delayed. Only runtime due to wave propagation speed may delay trebles. Digital systems are just very ineffective ways of wave propagation. Yours, Uli
Transient response becomes distorted by group delay, which is not constant. Even a filter of first order does this, but the sum of each one hi- and lopass of first order with same corner frequency, say a cross-over of first order, does not.
Hipasses introduce delay within the stopband, lowpasses within the passband, allpasses belo mid frequency. Hence in a classical electrical system, hi frequencies may be damped but are not delayed. Only runtime due to wave propagation speed may delay trebles. Digital systems are just very ineffective ways of wave propagation. Yours, Uli
Yes you can and no they are not.
Yes it will if, as I presume, you are referring here to Qts and system resonance that should ideally be well or at least critically damped for minimum audibility. It can be compensated for by pre-filtering as it is a linear, single-series response. This is not, however, compensating for the audible effects of group delay that I have also discussed in this thread.
Motional feedback also cannot correct for port resonances, cabinet resonances and internal volumes resonances that radiate out through the port or the driver diaphragm, however, since it only corrects coil velocity. At least it cannot be done without other compromises and it is much better to fix such problems at source rather than presume DSP is a cure for all ills. Maybe one of these effects is your "tubbiness"?
No, it can't. Not in the whole room. It can optimized for a part of the room but since the different drivers aren't a point sources (well, some are, most not though) and they have a certain distance to each other, there will always be some differences in distance to a certain point in the room, that means a difference in phase, as soon as you move and you can't correct it everywhere perfectly at the same time. You can't equalize the frequency response flat in the whole room at once either.
You can delay the other parts/frequencies but you can't remove the decay time the speaker (or the room) itself adds to it. But you can cut off frequencies at which it gets worse.
Phase will only be correct at one spot. If you have speakers that keep their output pretty much the same whether you move left or right or even up or down, any phase manipulation will hold true in a much wider space.
Not that many speakers can do that though. Indeed I remove energy from a room effect, as my speaker setup isn't symmetrical I can add a little to the other side to make up for it as it is around 70 Hz. I'm still considering adding 2 subs to be able to have them catch up so I'd have a more perfect left and right balance. Right now it looks like this:
Left, right and stereo sum at the listening position.
As can be seen, there's a dip in the left channel at 70 Hz, still some room reflections between 100-200 Hz in the right side and that side has more difficulty at the real low end around 30 Hz, so the left array makes up for that from it's corner position.
If anyone wants to see how they are doing, just download the APL_TDA software and have a look in your room.
I can account for just about any wiggle still left in that blue sea.
It should be a deep blue see, but hey, it's still a pretty normal living room.
Not that many speakers can do that though. Indeed I remove energy from a room effect, as my speaker setup isn't symmetrical I can add a little to the other side to make up for it as it is around 70 Hz. I'm still considering adding 2 subs to be able to have them catch up so I'd have a more perfect left and right balance. Right now it looks like this:
Left, right and stereo sum at the listening position.
As can be seen, there's a dip in the left channel at 70 Hz, still some room reflections between 100-200 Hz in the right side and that side has more difficulty at the real low end around 30 Hz, so the left array makes up for that from it's corner position.
If anyone wants to see how they are doing, just download the APL_TDA software and have a look in your room.
I can account for just about any wiggle still left in that blue sea.
It should be a deep blue see, but hey, it's still a pretty normal living room.
Last edited:
Yes it can in the context of my original answer and in the context of this thread. It was never implied that equalisation should occur over the entire room volume - and nor is that particularly relevant in this case...
My answer concerned equalisation at a nominal listening point which, as we are talking specifically about long wavelengths, will serve as the centre of a large spatial region of equalisation even where a single loudspeaker is attempting to equalise the room response due to itself over over local region (and not therefore necessarily fully coupled to any particular mode such as is the case in a more global equalisation scheme such as CABS, for example).
Very nearly you can. The aforementioned CABS scheme is one solution that tends towards global equalisation throughout the room volume.
It is a relatively simple matter to compensate for a non-ideal decay time. With the addition of a delay, it is further possible to produce a linear phase system. There is no need to "cut-off" any frequencies.
Last edited:
It was the opposite - it was never mentioned you were talking about one point in the room only! And the thread starter didn't mention that at all either!
That's only if you assume a box-type room. Once you have a more complex structure (L-shape or roof slope etc.) it can change much faster. You are automatically assuming a lot of things without ever mentioning it and concealing a lot of conditions which have to be fulfilled for that to apply.
I don't know the term 'CABS', care to explain?
It's not 'simple'. You have to measure the room and have to add a lot of filters and eq's and to add to that, you have to use FIR filters, which i.e. introduce pre-ringing which in turn worsen the transient response itself. So it's by far not a perfect solution at all.
Oh and I never said you have to cut it off. I said it is a simple way to improve it and that's with very little effort and costs - and it does exactly what I said.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.