What is the best/easiest way to digitally time align multiple loudspeakers, for instance, subs with mains? Before you respond that time alignment is unnecessary, especially at low frequencies, let me say that I am specifically dealing with a case that has significant path length and/or phase differences, and that I have the ability to add arbitrary delay to any and all channels.
I can measure step responses, but I'm not sure aligning the step response is really meaningful--would it be better to configure the loudspeakers so they are out of phase at crossover, find the largest null, and then then reverse the phase? The only thing that concerns me is knowing whether I am at 0 degrees or 360 degrees! (I guess this is impossible at subwoofer frequencies, but it could be harder to determine in the midrange and up)
Going farther afield, is there benefit to doing this at higher-than-sub frequencies? My midbass/midrange crossover is not time aligned (perhaps this should be called impulse aligned?), yet the phase and FR sum nicely on and off axis. The mid/high transition is a little more ragged off axis, but given the rapid phase rotation at HF I'm not sure time alignment at a given point could help...
Thanks for any input.
I can measure step responses, but I'm not sure aligning the step response is really meaningful--would it be better to configure the loudspeakers so they are out of phase at crossover, find the largest null, and then then reverse the phase? The only thing that concerns me is knowing whether I am at 0 degrees or 360 degrees! (I guess this is impossible at subwoofer frequencies, but it could be harder to determine in the midrange and up)
Going farther afield, is there benefit to doing this at higher-than-sub frequencies? My midbass/midrange crossover is not time aligned (perhaps this should be called impulse aligned?), yet the phase and FR sum nicely on and off axis. The mid/high transition is a little more ragged off axis, but given the rapid phase rotation at HF I'm not sure time alignment at a given point could help...
Thanks for any input.
For example, here is the phase response of ESP's 36dB/octave rumble filter. Add this to the changing phase of the driver itself near resonance and I think calculations won't cut it...some kind of iterative or other measurement process is needed.
I know there are other solutions (allpass filters) but if the problem can be directly solved, is there a reason to much around with something like that?
I know there are other solutions (allpass filters) but if the problem can be directly solved, is there a reason to much around with something like that?
Attachments
Well, the problem is that there does not appear to be widespread agreement on what time alignment means. To me, it ideally means two things: that the start of the impulses of two transducers are aligned. And two, that they are in phase at their crossover point.
To me this poses several questions. The first is whether the impulse start times mean anything if they are less than 1 cycle off and the drivers are in phase at crossover. Reviewers often use this as the only criteria for "time alignment."
Now, a delay (positive or negative) can affect both impulse start time and (effective) phase, although not independently. Unless I am mistaken, an all-pass can alter phase but has no effect on impulse start time.
I am not confident in these assertions though...I think there are people here that can respond with more authority. If you'd like you can replace "crossover" above with "frequency of interest", since adding filters to the mix will probably hoplessly complicate this discussion.
To me this poses several questions. The first is whether the impulse start times mean anything if they are less than 1 cycle off and the drivers are in phase at crossover. Reviewers often use this as the only criteria for "time alignment."
Now, a delay (positive or negative) can affect both impulse start time and (effective) phase, although not independently. Unless I am mistaken, an all-pass can alter phase but has no effect on impulse start time.
I am not confident in these assertions though...I think there are people here that can respond with more authority. If you'd like you can replace "crossover" above with "frequency of interest", since adding filters to the mix will probably hoplessly complicate this discussion.
The crux of the problem is that just about anything affects phase, including all the electronic components and even the wire in the system. That's why the only thing that is generally worried about is pathway differentials, because those can be easily fixed. Getting both that and perfection of phase alignment is next to impossible, especially in the bass. where room interactions also glitch everything up phase wise.
That bolsters my suspicion that the empirical technique I offered above might be the easiest solution--do you agree? I think your "fix the easy stuff" suggestion is valid, but even if worst-case phase summation errors only take 6dB off of your low end that is equivilent to losing half your subs!
The reason this piqued my interest was Harman's white paper on multiple subs. They standardized around 2-4 in part because the destructive effects of higher numbers of transducers...but they specifically noted that they were not considering the possibilities of aligning the phase of individual drivers.
The reason this piqued my interest was Harman's white paper on multiple subs. They standardized around 2-4 in part because the destructive effects of higher numbers of transducers...but they specifically noted that they were not considering the possibilities of aligning the phase of individual drivers.
If you measure the location of the voice coils for say the high frequency horn and the bass driver and input that difference in inches into the delay you would have a starting point for your time alignment. If your delay isn't capable of using that format then you will need to input the delay time in ms. To properly set the delay you will need some specialized high dollar equipment that I doubt that you have or have access to.
Your idea of adding a delay for each device per channel is asking for more trouble than what it is worth. Any delay will add noise and distortion to the system. Your best bet is to borrow a pink noise generator, an analyzer, calibrated microphone, warble generator and plotter and go to work on properly eqing your system. This will be of far more benefit to you than worring about a few inches that it may be off.
Your idea of adding a delay for each device per channel is asking for more trouble than what it is worth. Any delay will add noise and distortion to the system. Your best bet is to borrow a pink noise generator, an analyzer, calibrated microphone, warble generator and plotter and go to work on properly eqing your system. This will be of far more benefit to you than worring about a few inches that it may be off.
Since the apparent acoustic center changes over the bandwidth of a driver, I mainly concentrate on transient coherence through the crossover region. If you do this, while you will have more "phase shift" (I hate that concept of phase in acoustics, it is so sine wave like) there can be a smooth time and magnitude transition through the crossover region.
While the air path delay is important, the critical factor is the lag introduced by the low pass filter. The lower the crossover frequency, the more delay will be needed on the high pass driver or drivers.
It can be measured, it can be done, it may even have some beneficial effect. The effect, however, is at best subtle and probably not worth doing for any reason other than doing it. If your loudspeakers are transiently accurate and the cones are quiet above 300 Hz, you will love the dynamic onset sound. The sustains may still be bloated and ring like crazy because of room mode excitment or enclosure problems, but time aligning will not help that no matter how you time align them.
Good designing and good building,
Mark
While the air path delay is important, the critical factor is the lag introduced by the low pass filter. The lower the crossover frequency, the more delay will be needed on the high pass driver or drivers.
It can be measured, it can be done, it may even have some beneficial effect. The effect, however, is at best subtle and probably not worth doing for any reason other than doing it. If your loudspeakers are transiently accurate and the cones are quiet above 300 Hz, you will love the dynamic onset sound. The sustains may still be bloated and ring like crazy because of room mode excitment or enclosure problems, but time aligning will not help that no matter how you time align them.
Good designing and good building,
Mark
Really?burnedfingers said:
That definitely isn't true. Taking a digital signal and delaying it to some outputs and not others won't add any noise or distortion.
burnedfingers said:
Mark, please forgive my ignorance, because I really think you are a great resource here. I understand how to align the transient impulses in the absolute sense, but I have no idea how it applies specifically to "transient coherence through the crossover region." This goes to the heart of my original question, which really was "what is the relationship, if any, between aligning the initial transients, and the phase coherence through crossover?" Could I ask you to elaborate at all on this?
Quote:
That definitely isn't true. Taking a digital signal and delaying it to some outputs and not others won't add any noise or distortion.
Well it surely won't add distortion to a channel it isn't added to.
All digital delays input a certain amount of digital noise/hash. I consider this to be distortion. Analog delays also contribute to the distortion of a system as do ANY electronic components. Its very simple...the more components in the signal chain the more distortion.
Quote:
To properly set the delay you will need some specialized high dollar equipment that I doubt that you have or have access to.
Quote:
Really
Most people don't have a "Crown" Tef analyzer, software and calibrated microphones at their fingertips. Do you?
That definitely isn't true. Taking a digital signal and delaying it to some outputs and not others won't add any noise or distortion.
Well it surely won't add distortion to a channel it isn't added to.
All digital delays input a certain amount of digital noise/hash. I consider this to be distortion. Analog delays also contribute to the distortion of a system as do ANY electronic components. Its very simple...the more components in the signal chain the more distortion.
Quote:
To properly set the delay you will need some specialized high dollar equipment that I doubt that you have or have access to.
Quote:
Really
Most people don't have a "Crown" Tef analyzer, software and calibrated microphones at their fingertips. Do you?
My signals are already digital, so no, there is no loss of quality in using delays. I know that isn't always the case.
From what I found in a search the TEF analyzers are for automated FR, polar response, waterfall, etc plots. I would suspect that Speaker Workshop + other measurement gear could be used...with a lot more elbow grease I am sure.
From what I found in a search the TEF analyzers are for automated FR, polar response, waterfall, etc plots. I would suspect that Speaker Workshop + other measurement gear could be used...with a lot more elbow grease I am sure.
Quote:
From what I found in a search the TEF analyzers are for automated FR, polar response, waterfall, etc plots. I would suspect that Speaker Workshop + other measurement gear could be used...with a lot more elbow grease I am sure.
The setup of the Tef Analyzer takes less than the time it would take for you to boot up a lap top computer. I can complete the process in less than 5 min. Talk about elbow grease.
You could try Speaker Workshop or some other primative program out there but the results will be disappointing.You can't expect those programs to compete with a 10K piece of equipment. Those of us that do this as a normal part of our job do know what brings results.
We wouldn't even consider the gains
of what your trying to do.
I suggest you apply your time and considerable knowledge trying to tackle something that will bring better results. Borrow a calibrated microphone, Audio control analyzer(affordable unit),Pink noise generator,warble generator, and plotter. This will give you the necessary equipment needed to properly EQ your system. Start there first before trying to attempt time alignmment.
From what I found in a search the TEF analyzers are for automated FR, polar response, waterfall, etc plots. I would suspect that Speaker Workshop + other measurement gear could be used...with a lot more elbow grease I am sure.
The setup of the Tef Analyzer takes less than the time it would take for you to boot up a lap top computer. I can complete the process in less than 5 min. Talk about elbow grease.
You could try Speaker Workshop or some other primative program out there but the results will be disappointing.You can't expect those programs to compete with a 10K piece of equipment. Those of us that do this as a normal part of our job do know what brings results.
We wouldn't even consider the gains
of what your trying to do.I suggest you apply your time and considerable knowledge trying to tackle something that will bring better results. Borrow a calibrated microphone, Audio control analyzer(affordable unit),Pink noise generator,warble generator, and plotter. This will give you the necessary equipment needed to properly EQ your system. Start there first before trying to attempt time alignmment.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.