Hello, now that drivers have been chosen for my center channel prodject and have designed baffle for edge deffraction problems, I was wondering about time delay of the seperate drivers? with a 24db per octave slope and being electronic, this still can't do anything for the offset of acoustical centers of the drivers themselves right??? If this "is" the case then how does one implement correcting for time differences created from seperate drivers when using an active setup. Or is this even worth the trouble?
Thanks,
Dallaire
Thanks,
Dallaire
You will have to decide if it is worth the trouble, but the optimal way to correct for differences in air path delay is with a delay line. The convenient way to create an adjustable delay line is in the digital domain. There are professional (sound reinforcement) equipment that include both adjustable crossover options and adjustable delay lines.
Mark
Mark
Do you want a create a coherent source to reproduce transients? If you test with a transient signal, a multi-way system using asymmetrical electrical filters will still test as multiple discrete sources. Asymmetrical electrical filters will not correct for differences in air path delay in multi-way systems.
It would be wonderful if such a simple fix would work, but it doesn't. Sorry!
Be a diyer. Test it for yourself. Easy to do.
Mark
It would be wonderful if such a simple fix would work, but it doesn't. Sorry!
Be a diyer. Test it for yourself. Easy to do.
Mark
Of course, if I was patronizing, then you could dismiss all that I say.
To continue the actual discussion I will mostly ignore the first two lines of sreten's reply.
I notice that you (sreten) became specific in the third line. This is a statement that can advance discussion.
Not all asymmetrical electrical crossovers are L/R crossovers. The L/R fourth order is a specific design for a specific purpose. The L/R fourth purposely induces on-axis errors to correct a specific off-axis problem. I believe this is fairly well accepted since the on-axis problem part was part of the original paper.
The question for this thread (as I read it) was not about this off-axis problem. In relation to the question posed, the L/R fourth will not correct for differences in apparent acoustic centers (or differences in air path delay). It will, however, cause problems on-axis.
For me, the reason to be worried about differences in apparent acoustic centers (differences in air path delay) is to achieve an on and off-axis response that is as close to coherent and flat as possible. The L/R fourth will not accomplish this.
While sreten accused me of being patronizing (whether I was or not), I also said how Dallaire (or anyone else) could test what I said. I did not expect Dallaire to believe what I said just because I said it. An interesting difference.
Mark
As far as I see, there are two things: phase and time. You can correct the phase "error" introduced by the time delay with an easy all pass net. Going analog you can not correct the time delay.
I'm not too much concerned about the time alignment per se, because if you look at the pulse through the filter, it's kind of Quasimodo
I'm not too much concerned about the time alignment per se, because if you look at the pulse through the filter, it's kind of Quasimodo
Provided that all drivers are measured correctly with both amplitide and phase, or the phase is generated by FFT and the acoustic centres are perfectly offset in simulation software, such as SpeakerWorkshop, which we use to model the crossover. Does "align the phase" mean at the crossover point the phases of both drivers need to be the same, such as both are at 30 degree?
Hi Raka,
I agree with you about the look of an impulse at the output of an analogue filter. A transient test of an analogue filter will not look good.
What if the acoustic impulse is important? It is not a problem to have a crossover network that does not sum electrically if when added to the inherent slopes of the transducers, it sums acoustically. It is not a problem to have an impulse response that looks incoherent at the output of the filter if it becomes coherent when added to the "filters" implicit in the imperfect transducers.
One problem is that you cannot produce a coherent acoustic impulse if all you do is play with the electrical filters. A related problem is that if you then correct for air path delay when you test acoustically, your acoustic output will not sum. Heyser used to do this in his Audio magazine loudspeaker reviews. He produced some very nasty looking frequency response graphs for most of the loudspeakers of those days.
Another problem is that there is a problem or blind spot with FT testing. If you have a multi-way loudspeaker and the air path delay differences are large and one driver has stopped voicing before the next driver begins to voice and so on, all you are looking at in the FT is the magnitude sums. Most of the possible time relation information is lost. Any possible interference is ignored. This makes designing and claiming this or that response much easier. Still, there are lots of errors in the actual loudspeaker. Compensate for the differences in air path delay and you will then be able to see the interaction problems.
Back to Dallaire, Raka has stated that she is not too much concerned with time alignment per se. In contrast, I am for time alignment for the reasons given above (and many more).
If you go with the digital delay and crossover, once again, you will be able to test for yourself the differences between a loudspeaker system that is time aligned and one that is not, but that does use asymmetrical filters or all-pass filters to adjust only for phase.
Mark
I agree with you about the look of an impulse at the output of an analogue filter. A transient test of an analogue filter will not look good.
What if the acoustic impulse is important? It is not a problem to have a crossover network that does not sum electrically if when added to the inherent slopes of the transducers, it sums acoustically. It is not a problem to have an impulse response that looks incoherent at the output of the filter if it becomes coherent when added to the "filters" implicit in the imperfect transducers.
One problem is that you cannot produce a coherent acoustic impulse if all you do is play with the electrical filters. A related problem is that if you then correct for air path delay when you test acoustically, your acoustic output will not sum. Heyser used to do this in his Audio magazine loudspeaker reviews. He produced some very nasty looking frequency response graphs for most of the loudspeakers of those days.
Another problem is that there is a problem or blind spot with FT testing. If you have a multi-way loudspeaker and the air path delay differences are large and one driver has stopped voicing before the next driver begins to voice and so on, all you are looking at in the FT is the magnitude sums. Most of the possible time relation information is lost. Any possible interference is ignored. This makes designing and claiming this or that response much easier. Still, there are lots of errors in the actual loudspeaker. Compensate for the differences in air path delay and you will then be able to see the interaction problems.
Back to Dallaire, Raka has stated that she is not too much concerned with time alignment per se. In contrast, I am for time alignment for the reasons given above (and many more).
If you go with the digital delay and crossover, once again, you will be able to test for yourself the differences between a loudspeaker system that is time aligned and one that is not, but that does use asymmetrical filters or all-pass filters to adjust only for phase.
Mark
Wow, my head is spinning! you guy's obviously know alot more than myself, trust me!
Perhaps I can elaborate a bit more as to my setup. I am making a WMTMW array. My crossover setup is the Marchand XM1's, speaker will be all electronicly crossed. I am using, and I hate to say it cause I don't want to get audio rocks thrown at me, but I am using the Behringer 2496 units for all channels in my dedicated HT room, but they are just the parametric EQ's no x-overs or other bells/whistles.
I think alot of time when someone says "time correctiing" people start calling this "phase" are'nt the two different? When I sat "time" I am talking about the voice coils being in line in the physical sense.
So now knowing my x-over setup and the proccessors I have, it sounds like I "would" then benifit from alligning voice coils of drivers.
The only thing that I am concered with is I have spent alot of time placeing drivers on baffle using "Edge" program to get the baffle step to look nice with regard to indended cross points and such. I now I start building drivers forward on the baffle, I think this would "almost have to change the deffraction estimate"! don't you think...?
Anyway, this is where I'm currently at.
Vince
Perhaps I can elaborate a bit more as to my setup. I am making a WMTMW array. My crossover setup is the Marchand XM1's, speaker will be all electronicly crossed. I am using, and I hate to say it cause I don't want to get audio rocks thrown at me, but I am using the Behringer 2496 units for all channels in my dedicated HT room, but they are just the parametric EQ's no x-overs or other bells/whistles.
I think alot of time when someone says "time correctiing" people start calling this "phase" are'nt the two different? When I sat "time" I am talking about the voice coils being in line in the physical sense.
So now knowing my x-over setup and the proccessors I have, it sounds like I "would" then benifit from alligning voice coils of drivers.
The only thing that I am concered with is I have spent alot of time placeing drivers on baffle using "Edge" program to get the baffle step to look nice with regard to indended cross points and such. I now I start building drivers forward on the baffle, I think this would "almost have to change the deffraction estimate"! don't you think...?
Anyway, this is where I'm currently at.
Vince
Hmmm.........
Well I do not like being patronised and it being implied i do not
understand the obvious with an argument that is inconsistent.
All should agree that a digital delay will correct the driver offsets, this is obvious.
The offset causes assymetric lobing due to the phase tracking error it
introduces to the theoretically perfect phase tracking of a 4th order L/R
and by definition a delay will fix this.
I was in effect agreeing with the first statement of post 2.
What is it that you are actually fixing ? lobing ? phase tracking ? transient response ?
For the last case for 4th order L/R "transient coherency" cannot be
restored because perfect 4th order L/R is not transiently coherent.
Is there any point to perfect phase tracking (coherency would be
a misnomer if used here) of 4th order L/R ? Not really, so it seems
that an assymetric near 4th order L/R that corrects the phase error
at the crossover point is a simpler alternative approach to a delay.
That it will not restore the physical offset of the driver in the time
domain should be obvious to the least informed. What is not at all
obvious is whether restoring the physical offset is actually worth
doing for 4th order L/R, even if it is trivial in the digital domain.
I cannot answer that question from listening experience,
but I suspect given typical tweeter offsets and 4th order
electrical filters (indeterminate acoustic order) probably not.
/sreten.
The original SL design introducing 4th order L/R filters :
http://www.linkwitzlab.com/x-sb80-3wy.htm
uses a 36mm / 115uS delay at the c/o point.
Well I do not like being patronised and it being implied i do not
understand the obvious with an argument that is inconsistent.
All should agree that a digital delay will correct the driver offsets, this is obvious.
The offset causes assymetric lobing due to the phase tracking error it
introduces to the theoretically perfect phase tracking of a 4th order L/R
and by definition a delay will fix this.
I was in effect agreeing with the first statement of post 2.
What is it that you are actually fixing ? lobing ? phase tracking ? transient response ?
For the last case for 4th order L/R "transient coherency" cannot be
restored because perfect 4th order L/R is not transiently coherent.
Is there any point to perfect phase tracking (coherency would be
a misnomer if used here) of 4th order L/R ? Not really, so it seems
that an assymetric near 4th order L/R that corrects the phase error
at the crossover point is a simpler alternative approach to a delay.
That it will not restore the physical offset of the driver in the time
domain should be obvious to the least informed. What is not at all
obvious is whether restoring the physical offset is actually worth
doing for 4th order L/R, even if it is trivial in the digital domain.
I cannot answer that question from listening experience,
but I suspect given typical tweeter offsets and 4th order
electrical filters (indeterminate acoustic order) probably not.
/sreten.The original SL design introducing 4th order L/R filters :
http://www.linkwitzlab.com/x-sb80-3wy.htm
uses a 36mm / 115uS delay at the c/o point.
Hello, I am trying to fix acoustic time delay of the drivers so that soound is recieved to the ears from all drivers at the same time. If this is even a acchievable goal. I think, or hope that the lobing is taken care of by the D' Appolito(hope that's spelled correct) array conf. on the baffle. What do you think using "round" stand off spacers that match frame shape of drivers will do to my edge diffraction estimate??? This build out of drivers must also to something to the tweeters dispertion caricteristics I would think, being sandwiched between the mids, and set back in relation to the mid drivers within such close proximity.. I just don't want to get one portion of design nice and then screw someting else up trying to make something else correct. Perhaps I need to implement some sort of DIY built time delay for my situation?
Vince
Vince
Hi,
My point is 4th order L/R is not time coherent and whatever you do
all signals will not arrive at the same time even if the drivers are
equidistant from your ear.
The MTM array does force symmetrical lobing.
Phase matching does become a different issue for MTMs.
The centre speaker : WMTMW used horizontally or vertically ?
sreten.
My point is 4th order L/R is not time coherent and whatever you do
all signals will not arrive at the same time even if the drivers are
equidistant from your ear.
The MTM array does force symmetrical lobing.
Phase matching does become a different issue for MTMs.
The centre speaker : WMTMW used horizontally or vertically ?
sreten.Physical offsets can be a good thing
A physically offset driver is a good way to get good driver integration regardless of crossover type (even DSP), especially if the acoustic offset is significant. The advantage to a physical offset is that the integration of the drivers in the off-axis (especially the horizontal) is maintained better than for non-offset drivers with any type of time delay (DSP, all-pass delays or non-symmetrical slopes). The reason for this is that the non-offset drivers require a differing amount of delay with each change in the off-axis angle whereas physically aligned drivers do not. The actual significance of this will vary, of course, as it does for the on-axis acoustic offset phase delta. Small acoustic offsets may not be audibly significant, such as those between small midranges and tweeters.
As you point out, a physical offset will present an additional obstruction that will add reflection/diffraction problems that can be significant. Without some form of treatment, the "cure" of the offset may be worse than the original "symptom" of inter-driver acoustic delay. It is possible to ameliorate the diffraction to large degree, but at the expense of aesthetics. The one other possibility for the acoustic delay on a flat baffle is for the design axis to be on the lobe. That is, either tilt the baffle or use an inverted configuration. I've used the latter in cases of a 2-way with a largish mid-woofer. The off-axis droop in the tweeter response is not a big issue in my experience. The listening axis being on the mid-woofer's axis has its advantages.
A couple of sections at my site may be of interest. The first one is an early page when I first experimented with felt for diffraction. The drivers are not offset, but the midranges I tend to use have a small acoustic offset relative to tweeters.
Felt treatment for diffraction
If you have the June 2005 issue of audioXpress, they published an article of mine that has a much more detailed presentation for diffraction control for an system with an offset midrange. I've yet to add that to my pages. The design is at my site, though the boxes shown are unfinished mdf.
3-way with offset mid and diffraction control
It's not a pretty box, but it demonstrates the issue. This includes the M/T box used the audioXpress article.
Hope this helps.
Dave
dallaire said:
A physically offset driver is a good way to get good driver integration regardless of crossover type (even DSP), especially if the acoustic offset is significant. The advantage to a physical offset is that the integration of the drivers in the off-axis (especially the horizontal) is maintained better than for non-offset drivers with any type of time delay (DSP, all-pass delays or non-symmetrical slopes). The reason for this is that the non-offset drivers require a differing amount of delay with each change in the off-axis angle whereas physically aligned drivers do not. The actual significance of this will vary, of course, as it does for the on-axis acoustic offset phase delta. Small acoustic offsets may not be audibly significant, such as those between small midranges and tweeters.
As you point out, a physical offset will present an additional obstruction that will add reflection/diffraction problems that can be significant. Without some form of treatment, the "cure" of the offset may be worse than the original "symptom" of inter-driver acoustic delay. It is possible to ameliorate the diffraction to large degree, but at the expense of aesthetics. The one other possibility for the acoustic delay on a flat baffle is for the design axis to be on the lobe. That is, either tilt the baffle or use an inverted configuration. I've used the latter in cases of a 2-way with a largish mid-woofer. The off-axis droop in the tweeter response is not a big issue in my experience. The listening axis being on the mid-woofer's axis has its advantages.
A couple of sections at my site may be of interest. The first one is an early page when I first experimented with felt for diffraction. The drivers are not offset, but the midranges I tend to use have a small acoustic offset relative to tweeters.
Felt treatment for diffraction
If you have the June 2005 issue of audioXpress, they published an article of mine that has a much more detailed presentation for diffraction control for an system with an offset midrange. I've yet to add that to my pages. The design is at my site, though the boxes shown are unfinished mdf.
3-way with offset mid and diffraction control
It's not a pretty box, but it demonstrates the issue. This includes the M/T box used the audioXpress article.
Hope this helps.
Dave
While LR is indeed not time-coherent its amplitude response would still suffer from time-delays (correctly named that way by the original poster) since such a delay causes phase-lag and therefore the correct summing would suffer.
If you want to use an LR crossover and you therefore don't care about the temporal response you could use an allpass to achieve correct summing.
If you want to use a time-aligned crossover and still have a flat baffle then there is no way around implementing a delay for correcting a delay. This is indeed best achieved digitally. If you can live with a restricted frequency range where the correction has to work then you could also use a high-order analog allpass network, ideally with Bessel response.
I have once placed the simulation results for such a filter. Its order is 10, the delay is 100 us but starts to drop above 30 kHz.
My search only found the following though I am sure that I once posted more about that.
http://www.diyaudio.com/forums/showthread.php?postid=858255#post858255
Regards
Charles
If you want to use an LR crossover and you therefore don't care about the temporal response you could use an allpass to achieve correct summing.
If you want to use a time-aligned crossover and still have a flat baffle then there is no way around implementing a delay for correcting a delay. This is indeed best achieved digitally. If you can live with a restricted frequency range where the correction has to work then you could also use a high-order analog allpass network, ideally with Bessel response.
I have once placed the simulation results for such a filter. Its order is 10, the delay is 100 us but starts to drop above 30 kHz.
My search only found the following though I am sure that I once posted more about that.
http://www.diyaudio.com/forums/showthread.php?postid=858255#post858255
Regards
Charles
Just an added thought on why you might want to correct differential time delays for a coherent acoustic output. It is a way to avoid compression.
Most percussive instruments reproduced by multiple drivers will suffer three-dB peak magnitude compression if the output from the various drivers are not close to coherent when they arrive at the listener's ear. The lower the crossover frequency the more instruments that will be impacted by this peak level compression. Three-dB of compression is easily audible.
Mark
Most percussive instruments reproduced by multiple drivers will suffer three-dB peak magnitude compression if the output from the various drivers are not close to coherent when they arrive at the listener's ear. The lower the crossover frequency the more instruments that will be impacted by this peak level compression. Three-dB of compression is easily audible.
Mark
How do you define coherent?
It would be helpful if you would provide your definition of coherent. As sreten and phase_accurate have pointed out, no L-R crossover is "coherent" by any definition of which I am aware.
Also, your point about lower crossover frequency is a bit of a contradiction with regard to inter-driver phase relationships. The lower the Fc of any crossover point between any two drivers, the less the phase delta for any non-zero acoustic offset. The summed magnitude response is closer to the target for that crossover, thus the impact of relative acoustic offset is reduced with lower Fc.
If by compression you refer to power response, then the opposite of your conclusion is the correct one. The lower the Fc, the better the phase relationships, therefore the closer to flat the power response in the crossover area.
Compression, though, does not refer to power response issues.
Dave
MarkMcK said:
It would be helpful if you would provide your definition of coherent. As sreten and phase_accurate have pointed out, no L-R crossover is "coherent" by any definition of which I am aware.
Also, your point about lower crossover frequency is a bit of a contradiction with regard to inter-driver phase relationships. The lower the Fc of any crossover point between any two drivers, the less the phase delta for any non-zero acoustic offset. The summed magnitude response is closer to the target for that crossover, thus the impact of relative acoustic offset is reduced with lower Fc.
If by compression you refer to power response, then the opposite of your conclusion is the correct one. The lower the Fc, the better the phase relationships, therefore the closer to flat the power response in the crossover area.
Compression, though, does not refer to power response issues.
Dave
I seem to have misplaced my masters degree in quantum physics
All I want to do is build the perfect speaker !! Well the heck with the acoustic center thing, its sound like with my 24db per octave electronic x overs, it would still not have proper allignment even "if" the voice coils "were" in line! I will go with design and then measure with R plus D and see what it looks like. Sounds like this is the best I can do for now. I need to bone up more on all the various crossover designs, seems "EVERTHING" is a tradeoff!!!
Just curious, can I measure the driver delay's after building, using R plus D? When taking impulse measurments it lines the impulse with T=0 even if I measure all drivers seprate from one another.
Or does one need an ossiliscope to do this properly?
Thanks,
Vince
All I want to do is build the perfect speaker !! Well the heck with the acoustic center thing, its sound like with my 24db per octave electronic x overs, it would still not have proper allignment even "if" the voice coils "were" in line! I will go with design and then measure with R plus D and see what it looks like. Sounds like this is the best I can do for now. I need to bone up more on all the various crossover designs, seems "EVERTHING" is a tradeoff!!!Just curious, can I measure the driver delay's after building, using R plus D? When taking impulse measurments it lines the impulse with T=0 even if I measure all drivers seprate from one another.
Or does one need an ossiliscope to do this properly?
Thanks,
Vince
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.