Hi,
For a while I have been trying miniDSP with various configuration of active filters on a couple of my DIY systems.
What I have noticed is subjective listening fatigue when 48db/oct filters are used vs lower order ones like 24db/oct.
The sound is subjectively "cleaner" with these high order filters but after a while it becomes very uncomfortable to listen. The feeling is like the sound is "strange" to me. I have tried to use them both as a subwoofer filter and also between bass/mid/treble in various configurations in a 3-way with the same result. Frequency response analysis did not reveal any problems.
I have some possible explanations for this:
a) Very abrupt change in polar pattern causing sharp changes in power response.
b) Significant phase shift so harmonics of instruments are shifted too much
c) Psycho acoustics
d) Something specific to my systems or electronics
Am I lone there or someone else has similar experience?
Regards,
Lukas.
For a while I have been trying miniDSP with various configuration of active filters on a couple of my DIY systems.
What I have noticed is subjective listening fatigue when 48db/oct filters are used vs lower order ones like 24db/oct.
The sound is subjectively "cleaner" with these high order filters but after a while it becomes very uncomfortable to listen. The feeling is like the sound is "strange" to me. I have tried to use them both as a subwoofer filter and also between bass/mid/treble in various configurations in a 3-way with the same result. Frequency response analysis did not reveal any problems.
I have some possible explanations for this:
a) Very abrupt change in polar pattern causing sharp changes in power response.
b) Significant phase shift so harmonics of instruments are shifted too much
c) Psycho acoustics
d) Something specific to my systems or electronics
Am I lone there or someone else has similar experience?
Regards,
Lukas.
Have you checked off-axis response? Ideally, you'd want both drivers to have wide dispersion near the crossover for steep slopes to work. Without it, there may be abrupt change in off-axis response, causing poor timbre and ultimately inaccurate sound.
There is nothing wrong with steep slopes... depends on implementation and design.
There is nothing wrong with steep slopes... depends on implementation and design.
Same experience for me. I don't pretend to know why, but the lower order filter sounds more 'relaxed'.
When I have a free afternoon I need to try switching from LR24 to B18.
Bill
When I have a free afternoon I need to try switching from LR24 to B18.
Bill
It wouldn't but phase and impulse response probably would as higher order filters screw both of them up and this may be what you're hearing.
Measure at the output of the board for each filter type from 1st to 8th order and you'll see what I mean.
I always used higher order filters but currently running a 4 way + subs with 1st and 2nd order filters.
Never sounded better.
Sigh...
Its like people forget what they're trying to accomplish, it seems to me. So many people just "run after the numbers".
The ONLY purpose of a crossover, yes ... the SINGLE incontrovertible purpose, is to "steer" power at the various drivers in frequency ranges that corresponds to their reasonably efficient and accurate sound-reproduction range. Period. Truly, the whole story in a nutshell.
So, let's look at drivers. Its not like they have 48 dB/octave rolloffs outside their sweet-spot range! Most tend to ROLL off first at 2-3 dB/octave for an octave, then 6 for another one or one and a half, then another 10 for another, and so on... a complex roll-off that is very hard to model with either single-stage single-pole filters, or double pole, or even multipole in Chebychev, Bessel, Butterworth and all the other configurations.
But again - what are these drivers doing, outside their sweet (relatively flat) range? Well, they're making sound, but not doing so especially efficiently. And way outside, they're doing so especially inefficiently. Moreover, usually when they're in their "inefficient" sidebands, they're also likely to be distorting the sound, and causing all sort of spurious sonic effects that are ... generally non-musically interesting. [he puts up kevlar shield to deflect tomatoes lobbed by electric-guitar rockers!]
Anyway, the idea was/is: "well then, why not put together some filters, either passive, or active that confine the frequency spectrum applied to each speaker, more or less in its sweet (relatively linear) frequency band!"
Passive crossovers, active ones, makes no difference - this is the goal. So... the next (and by far hardest) thing to do is to figure out then what the response of the drivers are, in the enclosure that has been engineered for them, in a "normal" listening situation that is statistically likely to represent the average consumer use of the speakers ... and when that's figured out, to tailor a set of filters that splits up the signal power into bands that allow the smoothest transition between the drivers' responses ... across the whole band.
As I said, this edges toward the mythically-hard science and art. Purists will rail endlessly on their soapboxes about phase distortion, transient delays, responsiveness, energy storage and so on ... but in truth, that which seems most apparently important over the years of designing and listening to these things is the gentleness of the transition between frequency ranges.
In other words, I am entirely unsurprised that a lower dB/octave roll-off would sound better. This is because the transition between the drivers overlaps more, and allows their specific 'speed' and 'attenuation' to mix into a more interesting (complex phase, yet relatively flat power) far field.
And its not like with modern amplifiers, that one can't afford to "waste" a bit of power overdriving the skirts of each driver's cross-over frequency band, in order to achieve "sweet" and "musical" and all the rest of the words of cherished nature we use to describe perfect systems.
So... using near-50 years of experience: Choose a the lowest order filters that accomplish the cross-over, and moreover, if using multipole, choose non-ringing filters, such as subcritical Butterworth or even Bessel filters. Space the poles out in a chain, to further unsharpen the cutover frequencies. The result WILL be musically pleasing.
This advice can be followed whether one is bi/tri amping, or whether one is simply making passive cross-overs with low Q within the speakers themselves.
GoatGuy
Its like people forget what they're trying to accomplish, it seems to me. So many people just "run after the numbers".
The ONLY purpose of a crossover, yes ... the SINGLE incontrovertible purpose, is to "steer" power at the various drivers in frequency ranges that corresponds to their reasonably efficient and accurate sound-reproduction range. Period. Truly, the whole story in a nutshell.
So, let's look at drivers. Its not like they have 48 dB/octave rolloffs outside their sweet-spot range! Most tend to ROLL off first at 2-3 dB/octave for an octave, then 6 for another one or one and a half, then another 10 for another, and so on... a complex roll-off that is very hard to model with either single-stage single-pole filters, or double pole, or even multipole in Chebychev, Bessel, Butterworth and all the other configurations.
But again - what are these drivers doing, outside their sweet (relatively flat) range? Well, they're making sound, but not doing so especially efficiently. And way outside, they're doing so especially inefficiently. Moreover, usually when they're in their "inefficient" sidebands, they're also likely to be distorting the sound, and causing all sort of spurious sonic effects that are ... generally non-musically interesting. [he puts up kevlar shield to deflect tomatoes lobbed by electric-guitar rockers!]
Anyway, the idea was/is: "well then, why not put together some filters, either passive, or active that confine the frequency spectrum applied to each speaker, more or less in its sweet (relatively linear) frequency band!"
Passive crossovers, active ones, makes no difference - this is the goal. So... the next (and by far hardest) thing to do is to figure out then what the response of the drivers are, in the enclosure that has been engineered for them, in a "normal" listening situation that is statistically likely to represent the average consumer use of the speakers ... and when that's figured out, to tailor a set of filters that splits up the signal power into bands that allow the smoothest transition between the drivers' responses ... across the whole band.
As I said, this edges toward the mythically-hard science and art. Purists will rail endlessly on their soapboxes about phase distortion, transient delays, responsiveness, energy storage and so on ... but in truth, that which seems most apparently important over the years of designing and listening to these things is the gentleness of the transition between frequency ranges.
In other words, I am entirely unsurprised that a lower dB/octave roll-off would sound better. This is because the transition between the drivers overlaps more, and allows their specific 'speed' and 'attenuation' to mix into a more interesting (complex phase, yet relatively flat power) far field.
And its not like with modern amplifiers, that one can't afford to "waste" a bit of power overdriving the skirts of each driver's cross-over frequency band, in order to achieve "sweet" and "musical" and all the rest of the words of cherished nature we use to describe perfect systems.
So... using near-50 years of experience: Choose a the lowest order filters that accomplish the cross-over, and moreover, if using multipole, choose non-ringing filters, such as subcritical Butterworth or even Bessel filters. Space the poles out in a chain, to further unsharpen the cutover frequencies. The result WILL be musically pleasing.
This advice can be followed whether one is bi/tri amping, or whether one is simply making passive cross-overs with low Q within the speakers themselves.
GoatGuy
Bazukaz,
You can try to isolate the amplitude slope issue (your "a" explanation) from the phase shift issue ("b") by using rePhase to linearize the phase of the crossover (or maybe just linearizing some of it, like 360°, to end up with a 24dB/oct-like phase shift)
But of course before going any further in your investigations (and whereas you choose to linearize the phase or not) you should make sure that your crossover is properly setup: symmetrical acoustical slopes, and phase coherency between drivers around the crossover point. The phase coherency is the most important (and most overlooked) part IMHO, and if messed up it can have an impact on the slopes you will prefer: shallower slopes might typically become more "acceptable" in such situations, as it will tend to "blur" any phase issue...
You can try to isolate the amplitude slope issue (your "a" explanation) from the phase shift issue ("b") by using rePhase to linearize the phase of the crossover (or maybe just linearizing some of it, like 360°, to end up with a 24dB/oct-like phase shift)
But of course before going any further in your investigations (and whereas you choose to linearize the phase or not) you should make sure that your crossover is properly setup: symmetrical acoustical slopes, and phase coherency between drivers around the crossover point. The phase coherency is the most important (and most overlooked) part IMHO, and if messed up it can have an impact on the slopes you will prefer: shallower slopes might typically become more "acceptable" in such situations, as it will tend to "blur" any phase issue...
Hi,
Anyone could suggest a program for that? I have found some but not for 48db/oct slopes, and rePhase can only generate but not re-process existing files.
I am using a symmetrical filter, and drivers have adequate dispersion at crossover region. IMO phase should overlap but I did not measure this except for frequency response(to check if drivers are summing properly).
However there will always be variations in phase difference from different listener positions as drivers are offset from each other. Is it important to optimize in the front at 1m?
Regards,
Lukas.
Thanks for idea. That is interesting. Now I think a phase shift filter(Q=2, right?) could be used to process some music and listened through headphones for "fatiguing" and "audible" difference in a blind test.
Anyone could suggest a program for that? I have found some but not for 48db/oct slopes, and rePhase can only generate but not re-process existing files.
I am using a symmetrical filter, and drivers have adequate dispersion at crossover region. IMO phase should overlap but I did not measure this except for frequency response(to check if drivers are summing properly).
However there will always be variations in phase difference from different listener positions as drivers are offset from each other. Is it important to optimize in the front at 1m?
Regards,
Lukas.
Optimally driver centers should be located <1/4 wavelength apart to minimize lobing/off axis. Measurement distance should be several times the enclosure's width, and gated/windowed to remove reflections from room.
I don't recall Zaph on this, but how does reducing out of band signal add to driver demand? This is nonsense.
I don't recall Zaph on this, but how does reducing out of band signal add to driver demand? This is nonsense.
Yes I understand this well. However quarter wavelength at 2.5 kHz is just ~3.5 cm and this is obviously almost never possible to achieve between driver centers.
Also when the listener is moving in the room phase will be shifting constantly and different depending on frequency; combined with offset drivers this gives phase corrections across a wide range of positions mostly meaningless IMO.
Accepting this in design, and assuming vertical driver axis then lobes are vertical too.
Steering main lobe horizontal at listener to +/- 5 degrees requires timing control better than 6 microseconds, regardless of crossover slope and driver spacing.
With DSP each driver may be EQ flat through crossover region yielding better acoustical sum with crossover.
Steering main lobe horizontal at listener to +/- 5 degrees requires timing control better than 6 microseconds, regardless of crossover slope and driver spacing.
With DSP each driver may be EQ flat through crossover region yielding better acoustical sum with crossover.
Scott was speaking of the demand in amplitude just before the slope. I have read this on the Zaph site as well (although I can't recall where) but he mentions that he decided to go with a second order slope instead of fourth order because the slope started earlier on the 2nd order, as in a lower Q.
I believe it was pertaining to a tweeter.
Exactly.
Take, say, LR4 and LR6. Since LR4 begins its rolloff at at a higher frequency, amplitude demands on the tweeter are reduced compared to LR6 which forces it to work harder to a lower frequency. Which is the better tradeoff? YMMV on that score, especially given the host of other factors at work.
Here's the above mentioned example: it's talking LR2 & LR4 but you get the idea: Zaph|Audio
Edit: Finally found it. See the XO section & the remarks on tweeter power handling: http://www.zaphaudio.com/BAMTM.html
Take, say, LR4 and LR6. Since LR4 begins its rolloff at at a higher frequency, amplitude demands on the tweeter are reduced compared to LR6 which forces it to work harder to a lower frequency. Which is the better tradeoff? YMMV on that score, especially given the host of other factors at work.
Here's the above mentioned example: it's talking LR2 & LR4 but you get the idea: Zaph|Audio
Edit: Finally found it. See the XO section & the remarks on tweeter power handling: http://www.zaphaudio.com/BAMTM.html
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.