I do quite a lot of processing in my home system -- most of it not dedicated to the crossovers, but that other processing that adds several tens of milliseconds of delay to the signal. I have a problem with lip-sync when watching television.I'll give you that but for home audio/video, not so much.
Not sure what you mean by offset...maybe physical offset?Yes, but isn't that to compensate for offset, not to linearize phase?
I've seen Meyer instructional vids that showed two uses of all-pass.
One was to flatten phase between CD and woofer like Subsonics mentioned, using a UPA-1p as an example.
and the other was to better match phase traces between their various system families/products.
Meyer is big on phase matching.
I would like to think it is never too late to change your mind. Your thinking in this seems just as wrong to me as the original premise of this this thread. I don't have the time or energy to expend on trying to convince anybody of anything.OK, this is all just me and a big so what.....
but pls understand, i think optimizing to a listening position in a room is all well and good, it's just hard for me to get excited about....
Re the sound of a linear phase system, to me there seems to be a less compressed sound with the bass frequencies tending to sound less muddy/not masking the higher frequencies so much. My system is not a home hifi type system, more to cover 500-1000 people size outdoor venues for live music events. With a system like this, your not only hearing with your ears, various parts of the body are a part of the picture and levels don't have to be huge for this to be the case, for example, Evlyn Glennie is a highly regarded percussionist, plays with some of the worlds best orchstras and she is deaf, she does so the feel. One moment that really made real impression on how I listen to sound systems is the experience of being a sound tech when we had the BBC Philharmonic Orchestra playing the closing show of a festival I was working at, I was asked to sit by the conductor during sound check to be available to adjust the microphones if needed, now that was not just heard but properly felt as well, classical is generally not my thing, recorded at least but experiencing it as the conductor does was altogther something else.
When you adjust the phase so as the impulse response is intact over a decently wide litening window, the overall peak is higher and is not smeared over time as much so can only add to the visceral experience. It also does the opposite to the signal tge amplifier seesb peaks not generally being as high, so more headroom.
When you adjust the phase so as the impulse response is intact over a decently wide litening window, the overall peak is higher and is not smeared over time as much so can only add to the visceral experience. It also does the opposite to the signal tge amplifier seesb peaks not generally being as high, so more headroom.
I get how you think.I would like to think it is never too late to change your mind. Your thinking in this seems just as wrong to me as the original premise of this this thread. I don't have the time or energy to expend on trying to convince anybody of anything.
But it's not about changing my mind....it's about how do i change a great deal of experience?
All i can say is have you ever heard a very high quality setup outdoors, one that has clean linear and strong SPL, all the way down to the bottom?
It makes room optimizing pursuits seem a bit trivial to me....in terms of the quest for best sound ever heard.
Last edited:
Yes!Re the sound of a linear phase system, to me there seems to be a less compressed sound with the bass frequencies tending to sound less muddy/not masking the higher frequencies so much. My system is not a home hifi type system, more to cover 500-1000 people size outdoor venues for live music events. With a system like this, your not only hearing with your ears, various parts of the body are a part of the picture and levels don't have to be huge for this to be the case, for example, Evlyn Glennie is a highly regarded percussionist, plays with some of the worlds best orchstras and she is deaf, she does so the feel. One moment that really made real impression on how I listen to sound systems is the experience of being a sound tech when we had the BBC Philharmonic Orchestra playing the closing show of a festival I was working at, I was asked to sit by the conductor during sound check to be available to adjust the microphones if needed, now that was not just heard but properly felt as well, classical is generally not my thing, recorded at least but experiencing it as the conductor does was altogther something else.
When you adjust the phase so as the impulse response is intact over a decently wide litening window, the overall peak is higher and is not smeared over time as much so can only add to the visceral experience. It also does the opposite to the signal tge amplifier seesb peaks not generally being as high, so more headroom.
I think that hearing goes beyond our ears, and is an integration of our skin and body cavity vibrations, combined with our ears.
We can sense/feel the rightness of a clean in-phase impulse repulse, by both rhythm and dynamics.....or at least i think i do 🙂
This has been a really great thread even if 51% over my head. Thanks to all contributors.
@mark100 I have followed your build threads with great interest. They are always great reads. I wonder has anything discussed here resulted in you making any positive, even if only subjective, changes to your pre-this-thread syn11?
@mark100 I have followed your build threads with great interest. They are always great reads. I wonder has anything discussed here resulted in you making any positive, even if only subjective, changes to your pre-this-thread syn11?
I don't know if you thought of how the compensation could be frequency scaled for a 1kHz LPF, where it might matter a bit more than at 100Hz.But, if I can throw a wrench into all this, I'm not quite sure what benefit extending the linear phase region into the stop band has from an acoustic point of view. Suppose I use this filter on a woofer ..... there isn't much difference in below 100 Hz because the GD is dominated by the woofer low frequency cut off.
Also, if I were to use the final result (last graph) of the below as the input to yet another iteration, then it may be possible to linearise the phase across 20kHz, which would be very useful for high-pass filters, that have the same phase characteristics (only with a mirrored magnitude).
Now, put that HPF with an LPF, you'd have a linear phase crossover system that uses only biquads !! However, I think some type of optimisation must be done to maximise the width of the linear phase band, in the first place.
I haven't tried it, but I think the mathematics look feasible, as a single iteration alone would linearise upto 5x-6x the -6dB crossover frequency. So, if one were to start with an 800Hz filter (typical for 15" drivers), you'd need only two iterations to reach 20kHz !!!
No you don’t. As this has nothing to do with the thread there seems no point in continuing.I get how you think.
Yes, physical offset resulting in mismatch phase through the crossover. The lattice is used to add delay to time align.Not sure what you mean by offset...maybe physical offset?
I've seen Meyer instructional vids that showed two uses of all-pass.
One was to flatten phase between CD and woofer like Subsonics mentioned, using a UPA-1p as an example.
and the other was to better match phase traces between their various system families/products.
Meyer is big on phase matching.
Re the sound of a linear phase system, to me there seems to be a less compressed sound with the bass frequencies tending to sound less muddy/not masking the higher frequencies so much. My system is not a home hifi type system, more to cover 500-1000 people size outdoor venues for live music events. With a system like this, your not only hearing with your ears, various parts of the body are a part of the picture and levels don't have to be huge for this to be the case, for example, Evlyn Glennie is a highly regarded percussionist, plays with some of the worlds best orchstras and she is deaf, she does so the feel. One moment that really made real impression on how I listen to sound systems is the experience of being a sound tech when we had the BBC Philharmonic Orchestra playing the closing show of a festival I was working at, I was asked to sit by the conductor during sound check to be available to adjust the microphones if needed, now that was not just heard but properly felt as well, classical is generally not my thing, recorded at least but experiencing it as the conductor does was altogther something else.
When you adjust the phase so as the impulse response is intact over a decently wide litening window, the overall peak is higher and is not smeared over time as much so can only add to the visceral experience. It also does the opposite to the signal tge amplifier seesb peaks not generally being as high, so more headroom.
Big difference between outside and a room. In a room I can't see how linear phase in the low frequency , modal region can make a nificant difference. I certainly don't hear it. At higher frequencies, perhaps. When you look at the impulse and see a higher peak it is generally because the higher frequency components are more faithfully reproduced in the direct sound. So maybe the intial attack would be cleaner, but not so much in the over hang and decay. Think of the room as a filter. At any listening position in the room there is a source to listener transfer function, T(s,l), where s and l are the source and listening position, respectively. Just as a crossover filter, that transfer function has both amplitude and phase. So even if the speaker is linear phase to DC, what you hear is the convolution of the source response and T(s,l). I can assure you that T(s,l) is not linear phase at any reasonable listening distance. Typically, it's not even minimum phase, and it's different at every point in the room.
Sorry, not quite the case, most probably because the intermediate result is nowhere close to a Bessel.I haven't tried it, but I think the mathematics look feasible, as a single iteration alone would linearise upto 5x-6x the -6dB crossover frequency. So, if one were to start with an 800Hz filter (typical for 15" drivers), you'd need only two iterations to reach 20kHz !!!
But the initial slope of the pulse is still first to arrive, that doesn't change even if listening from another room.Big difference between outside and a room. In a room I can't see how linear phase in the low frequency , modal region can make a nificant difference. I certainly don't hear it. At higher frequencies, perhaps. When you look at the impulse and see a higher peak it is generally because the higher frequency components are more faithfully reproduced in the direct sound. So maybe the intial attack would be cleaner, but not so much in the over hang and decay. Think of the room as a filter. At any listening position in the room there is a source to listener transfer function, T(s,l), where s and l are the source and listening position, respectively. Just as a crossover filter, that transfer function has both amplitude and phase. So even if the speaker is linear phase to DC, what you hear is the convolution of the source response and T(s,l). I can assure you that T(s,l) is not linear phase at any reasonable listening distance. Typically, it's not even minimum phase, and it's different at every point in the room.
To add to above as missed the edit window:
In the case of both Mark100's boxes and mine, the off axis step response stay's very much intact over quite a wide angle with a gradual reductiin in the sharpness of the slope, both in the horizontal and vertical, the step response also holds on mine at approx 2 baffle lengths to over 50' away. I have a pair on my workbench at home.
In the case of both Mark100's boxes and mine, the off axis step response stay's very much intact over quite a wide angle with a gradual reductiin in the sharpness of the slope, both in the horizontal and vertical, the step response also holds on mine at approx 2 baffle lengths to over 50' away. I have a pair on my workbench at home.
I don't know if you thought of how the compensation could be frequency scaled for a 1kHz LPF, where it might matter a bit more than at 100Hz.
View attachment 1132846
Also, if I were to use the final result (last graph) of the below as the input to yet another iteration, then it may be possible to linearise the phase across 20kHz, which would be very useful for high-pass filters, that have the same phase characteristics (only with a mirrored magnitude).
Now, put that HPF with an LPF, you'd have a linear phase crossover system that uses only biquads !! However, I think some type of optimisation must be done to maximise the width of the linear phase band, in the first place.
I haven't tried it, but I think the mathematics look feasible, as a single iteration alone would linearise upto 5x-6x the -6dB crossover frequency. So, if one were to start with an 800Hz filter (typical for 15" drivers), you'd need only two iterations to reach 20kHz !!!
The question I would pose (not to you but in general) is that what is the point in developing linear phase filters when you then apply them to filter drivers which are minum phase? You end up with a response from each driver in which the amplitude of the driver is modified by the filter but the phase is not (other than adding a delay). The acoustic output is therefore neither linear phase or minimun phase, (or minimum phase + delay). To me the idea of linear phase is a speaker who's acoustic outout at the design point is linear phase. When I was involved in the development of PC software to do that the approach taken was to start with the driver's response, say a woofer and tweeter in a two way. Then apply minimum phase equalization to flatten the response ober some frequency range that extended well into the stop bands. Then a text book minimum phase filter was applied, like and LR4. At that point the summed response would yield a typical LR4 system. If linear phase was desired a phase correction filter was applied to each driver making it linear phase and the summed response was then also linear phase. Caveat being that the base crossover had to be of the LR type. If a Butterworth type x-o was used the phase linearization has to be applied to the system, not the individual drivers.
Here is an example. The red is the woofer response. The green the desired LR4 LP filtered acoustic response. MP EQ is applied between the noted frequency limits making the response flat.
Then that EQ is convolved with the LR4 minimum phase LP filter. The resulting filter looks like this where HBT(s) is the EQ part and T(s) the LR4.
The final filtered acoustic response then looks like a minimunm phase band pass. with the LP roll off closely matching the LR4 target.
The net filter F(s) = HBT(s) x T(s) is emulated using FIR and being minimum pohase, is causal. The acoustic output is than O(s) = F(s) x D(s) where D(s) is the driver response. To linearize phase X(s) is found as X(s) = O(s)/ |O(s)| which is just a band pass filter that has the same phase as O(s). Then the impulse respons of X(s), x(t) is found by iFFT. While X(s) is not minimum phase it's impulse is causal. Anyway, to linearize phase, you need to convolve F(s) with 1/X(s) and the linear phase acoustic output is O'(s) = (F(s)/X(s)) x D(s).
Measured output looks like this. This was an in room measurement so only extends to 400 Hz to window out room reflections.
Impulse,
Frequency and phase,
Obviously, you can start with a speaker's resonse, S(s), and find the EQ, HBT(s), drop T(s) fromt he picture and just have O'(s) = HBT(s)/X(s) x S(s) and equalize the speaker while corrrecting phase with a single FIR filter. And the thing is that you only need to comupte the convolution of F(s)/X(s), or in the speaker case HBT(s)/X(s), once and store it. Applying phase linearization to the complete speaker is required if the crossover type does not sum flay with phase alignment, as with a Butterworth crossover.
This was all done on a Windows XP PC and, yes there was processing latency.
Anyway, when all is said and done, I applied this to a variety of speakes and that's when I lost interest in linear phase and so called transient perfect speakers. Maybe they sounded a little different, but I could never quantify them as sounding better.
But the initial slope of the pulse is still first to arrive, that doesn't change even if listening from another room.
Yes, but the point I was making was that the initial slope, the rise, is a function of the high frequency component in the signal. Maybe this will help. The top 3 plots show the response of a linear phase system with 100 Hz crossover with LR4 amplitude response. The system has flat amplitude and linear phase. At the left is the system response to a 0.1 msec wide pulse. Center is the contribution from the HP section and at thre right that from the LP section.
Below that is a comparison of the response of the LP section to a 1 msec wide pulse when the LP section is minimum phase and linear phase with a delay equa;y to the DC GD of the minimum phase response.
Fluid, sorry if i gave any offense....you truly rate as one of the most knowledgeable, considerate, and helpful posters. Sincerely meant, and thank you.No you don’t. As this has nothing to do with the thread there seems no point in continuing.
Agreed, as to the off-topic nature of our mini-discussion.
What you describe has been my standard technique on perhaps 30 various DIY projects over the last half-dozen years. Although I always apply high order (96dB/oct) standard LR linear-phase xovers, after the minimum-phase driver by driver flattening stage.To me the idea of linear phase is a speaker who's acoustic outout at the design point is linear phase. When I was involved in the development of PC software to do that the approach taken was to start with the driver's response, say a woofer and tweeter in a two way. Then apply minimum phase equalization to flatten the response ober some frequency range that extended well into the stop bands. Then a text book minimum phase filter was applied, like and LR4. At that point the summed response would yield a typical LR4 system. If linear phase was desired a phase correction filter was applied to each driver making it linear phase and the summed response was then also linear phase. Caveat being that the base crossover had to be of the LR type. If a Butterworth type x-o was used the phase linearization has to be applied to the system, not the individual drivers.
Here's my latest ongoing 5-way project using such:
It's an experimental test bed with a lot of latitude for choosing various xover points.
Since the xovers are all linear phase, its super easy to change their xover frequency and continue to get flat mag and phase summations.
So I do that, and examine off-axis responses, to find an optimal set of xover points.

Using the same technique, whenever I'v compared low order (24 dB/oct or less) minimum phase LRs vs the same order linear-phase LRs, I couldn't hear much of a difference between the two.Anyway, when all is said and done, I applied this to a variety of speakes and that's when I lost interest in linear phase and so called transient perfect speakers. Maybe they sounded a little different, but I could never quantify them as sounding better.
Where I hear an improvement over either of those low order usages, is with high-order lin-phase.
Whether the improvement is due to the relative easier task of achieving driver flattening without too much EQ, or perhaps less off-axis lobing potential from narrower summation ranges, I dunno.
I just know the technique with steep linear-phase LRs is easy to implement with almost guaranteed good results.
(High-order minimum phase are of course a non-starter, so no way to compare to steep lin-phase.)
Hi Kevomoso, kind words for all 🙂This has been a really great thread even if 51% over my head. Thanks to all contributors.
@mark100 I have followed your build threads with great interest. They are always great reads. I wonder has anything discussed here resulted in you making any positive, even if only subjective, changes to your pre-this-thread syn11?
The thread opening measurement mistake I was making, has changed how I'll be making outdoor sub measurements. And since syn11 includes sub drivers, will effect it when spring comes and I can set syn11 up outdoors. The last measurement project I was working on before winter cam, was trying to achieve as clean a step response as possible, using an IIR system high-pass. It was wasn't working as well as I expected, but i now see it was probably just the same measurement error being made then.
So a very helpful thread for me, and thx to all.
Otherwise, i don't plan any change in how I've been going about tuning.
If anything, I feel even more confident with about the processing strategy I've been using. It seems very straightforward with little room for error, compared to more complicated filter alternatives.
Plus, "it takes time to fix time" implies to me, that all roads end up with the same constant delay for achieving the same degree of desired phase flattening......so why bother with anything other than the simplest method to enact....?
That's very encouraging to the proponents of FIR filters. Yes, FIR filters can calibrate out the effects of everything including the speakers themselves. No pun intended, but in my opinion, a perfect output would be achieved when the variations in speaker parameters due to temperature etc. are also exactly modelled and compensated.The question I would pose (not to you but in general) is that what is the point in developing linear phase filters when you then apply them to filter drivers which are minum phase? You end up with a response from each driver in which the amplitude of the driver is modified by the filter but the phase is not (other than adding a delay). The acoustic output is therefore neither linear phase or minimun phase, (or minimum phase + delay). To me the idea of linear phase is a speaker who's acoustic outout at the design point is linear phase.
However, there's still a problem, that is, the mixing of the many a music titles have not been carried out on group delay corrected monitor systems, and that could introduce differences between what we hear and what was intended.
No offence taken, I should not let frustration get the better of me.Fluid, sorry if i gave any offense....
You may well be right with regards to linear phase in the modal region in a room, I haven't dealt with the GD of the subs on my system with DSP as would mean to much delay so have tuned the system low instead. Also most of what I have been doing with the system is outdoors in spaces to big to be in the modal region except for my mid tops on my workbench which are very much in the nearfield. I do think though that using filters is the way to go and notice a major improvement in the realism of my system, there has been a lot of chipping away at polar response, time domain and distortion products overall though so will take some working out what is what. I also think it takes a lot of things to drop into place before some of the more subtle effects really shine, for example, working on a random PA system can be quite hard to mix on, frequently look for what I'm cutting or boosting on a live mix on multiple channels and then add it as a global filter, eventually it is possible to make very small changes audible.Yes, but the point I was making was that the initial slope, the rise, is a function of the high frequency component in the signal. Maybe this will help. The top 3 plots show the response of a linear phase system with 100 Hz crossover with LR4 amplitude response. The system has flat amplitude and linear phase. At the left is the system response to a 0.1 msec wide pulse. Center is the contribution from the HP section and at thre right that from the LP section.
Below that is a comparison of the response of the LP section to a 1 msec wide pulse when the LP section is minimum phase and linear phase with a delay equa;y to the DC GD of the minimum phase response.
- Home
- Loudspeakers
- Multi-Way
- Why not IIR filters + a global phase linearization by FIR