I can't say I have done it directly. I have used DSP (MiniDSP) for years, like most DSP, this allows active EQ and crossovers and time delay per channel. I only have the most basic test equipment (REW + a mic) and understand frequency testing but not so much about phase. I think it is more important to be time aligned ( = valid at all frequencies) rather than phase aligned ( = valid only at certain frequency). With DSP you can do anything, including things you don't understand 
Nope
Interesting question Erik.
Much talk is abound the forum about the advantages of DSP, and the claimed audible response of either FIR IIR or more ye old world analogue active filtering (say S & K op amp filters)
I have heard some reasonable time aligned DSP based monitors (tbh I don't even know what brand/model)
I have heard my own system, analogue active, and passive speaker level filters. The phase in both at XO is aligned within reason.
Passive version is a typical 2nd order, inverted tweeter 2 way. I didn't like having to invert the tweeter.
So...in the op amp filters I used 4th order. Phase is within 40' of a full cycle (so either just shirt of wrapping phase, or just over a cycle wrap, I can't recall)
The active version is so much more coherent on instruments with short decays, and percussion.
I would like to time align and see if I can tell the difference.
It would be a little difficult for me to A-B compare, without another quad of amps identical to the ones I have already...
Still it would be very interesting to directly compare the analogue active filters against a DSP emulation with the appropriate delay to negate the 320' or 400' phade shift, and hear it on one set of speakers.
I'm just of the uninformed opinion that the improvement, if any, is going to be less than the improvement I experienced when going from passive to active analogue. filter (it's been debated here many times...)
The effort:reward ratio just becomes small.
Interesting question Erik.
Much talk is abound the forum about the advantages of DSP, and the claimed audible response of either FIR IIR or more ye old world analogue active filtering (say S & K op amp filters)
I have heard some reasonable time aligned DSP based monitors (tbh I don't even know what brand/model)
I have heard my own system, analogue active, and passive speaker level filters. The phase in both at XO is aligned within reason.
Passive version is a typical 2nd order, inverted tweeter 2 way. I didn't like having to invert the tweeter.
So...in the op amp filters I used 4th order. Phase is within 40' of a full cycle (so either just shirt of wrapping phase, or just over a cycle wrap, I can't recall)
The active version is so much more coherent on instruments with short decays, and percussion.
I would like to time align and see if I can tell the difference.
It would be a little difficult for me to A-B compare, without another quad of amps identical to the ones I have already...
Still it would be very interesting to directly compare the analogue active filters against a DSP emulation with the appropriate delay to negate the 320' or 400' phade shift, and hear it on one set of speakers.
I'm just of the uninformed opinion that the improvement, if any, is going to be less than the improvement I experienced when going from passive to active analogue. filter (it's been debated here many times...)
The effort:reward ratio just becomes small.
I'm with you here, turk 182. Maybe I just don't get the question. To me phase is the indicator of time behavior.
What is meant by phase aligned in this question?
I've done all I can to get my system time coherent, but I don't have any crossovers. To me (in other words: in my opinion) everything matters. But there's lots of ways to get it wrong. It should be right at the listening position, but you can't just unwrap the phase with FIR and expect it to be good enough. A frequency dependent window filter would help to pick the timing where you want the phase to be correct. All in my opinion of coarse.
One way to find out how you're doing is to run the demo of APL_TDA and measure at the listening position. That will show you the timing of your signal at that listening spot.
Mine looks like this:
Compared to a direct loop back of my DAC:
The plot of the speakers in 2D view:
This is the combined result of both left and right speaker at the listening position.
@ mondogenerator, if you have the option, try running a Harsch crossover. The tweeter/woofer will still be positive but this will give you way less phase rotation. That way you could test if you like it or not. Not sure if you have the possibility to try it.
What is meant by phase aligned in this question?
I've done all I can to get my system time coherent, but I don't have any crossovers. To me (in other words: in my opinion) everything matters. But there's lots of ways to get it wrong. It should be right at the listening position, but you can't just unwrap the phase with FIR and expect it to be good enough. A frequency dependent window filter would help to pick the timing where you want the phase to be correct. All in my opinion of coarse.
One way to find out how you're doing is to run the demo of APL_TDA and measure at the listening position. That will show you the timing of your signal at that listening spot.
Mine looks like this:
Compared to a direct loop back of my DAC:
The plot of the speakers in 2D view:
This is the combined result of both left and right speaker at the listening position.
@ mondogenerator, if you have the option, try running a Harsch crossover. The tweeter/woofer will still be positive but this will give you way less phase rotation. That way you could test if you like it or not. Not sure if you have the possibility to try it.
Last edited:
Sorry for the confusion. What I meant was, take a traditional box shaped enclosure with a tweeter and woofer.
Using analog filters (either active or passive) without time delay compensation, the tweeter output will arrive at the listener's ears before the woofer. A good crossover will be phase matched through the crossover region, but still the tweeter arrival will be ahead of the woofer, causing the step response to show the sharp peak of the tweeter, followed by the tall and wide output of the woofer.
In what I am calling a time-aligned system, the tweeter's output would be delayed by a millisecond or more (usually) so that the arrival of the driver's output at the listener's location is coincident, causing a step response to look like a single sharp rise followed by a triangle-like decay.
Of course, there are two ways to achieve time-alignment. One is by delaying the tweeter electronically, the second is by physically moving the tweeter away from the front so that the acoustic planes (distance) of the drivers are the same.
What I'm curious about is whether anyone has used DSP or some other electronic means to compare merely phase-aligned speakers to phase and time-aligned. I'm not worried about minimum phase in the question. A direct A/B comparison seems easiest with DSP active crossovers, so that you compare:
1 - 2 way crossover designed without any time delay
2 - 2 way crossovers designed so that the tweeter is delayed in such a way as to reduce the acoustic offset to 0. Usually this is the time equivalent of 1.2" to 1.8" depending on the tweeter or woofers being used. These are just example values!
The reason I ask is that I've seen some promote time alignment (version 2) as being superior. I'm just curious how it sounds. It's easy to see it's better on the scope. I'm less interested in that as in the subjective differences.
Yes, I am aware that acoustic offset of the drivers will naturally change the correct filters used. I'm just trying to find out if there are any tangible benefits of time alignment, and if so just how big it is.
Hope that helps.
Best,
Erik
Using analog filters (either active or passive) without time delay compensation, the tweeter output will arrive at the listener's ears before the woofer. A good crossover will be phase matched through the crossover region, but still the tweeter arrival will be ahead of the woofer, causing the step response to show the sharp peak of the tweeter, followed by the tall and wide output of the woofer.
In what I am calling a time-aligned system, the tweeter's output would be delayed by a millisecond or more (usually) so that the arrival of the driver's output at the listener's location is coincident, causing a step response to look like a single sharp rise followed by a triangle-like decay.
Of course, there are two ways to achieve time-alignment. One is by delaying the tweeter electronically, the second is by physically moving the tweeter away from the front so that the acoustic planes (distance) of the drivers are the same.
What I'm curious about is whether anyone has used DSP or some other electronic means to compare merely phase-aligned speakers to phase and time-aligned. I'm not worried about minimum phase in the question. A direct A/B comparison seems easiest with DSP active crossovers, so that you compare:
1 - 2 way crossover designed without any time delay
2 - 2 way crossovers designed so that the tweeter is delayed in such a way as to reduce the acoustic offset to 0. Usually this is the time equivalent of 1.2" to 1.8" depending on the tweeter or woofers being used. These are just example values!
The reason I ask is that I've seen some promote time alignment (version 2) as being superior. I'm just curious how it sounds. It's easy to see it's better on the scope. I'm less interested in that as in the subjective differences.
Yes, I am aware that acoustic offset of the drivers will naturally change the correct filters used. I'm just trying to find out if there are any tangible benefits of time alignment, and if so just how big it is.
Hope that helps.
Best,
Erik
Last edited:
What kind of crossover are we talking about that gives that second kind of STEP response? Just to be clear of what you are talking about.
You seem to be tacitly referring to direct radiators only. If you were talking about a multiple-entry horn, then your question would be answered...probably like this:
In the crossover region, either the drivers are aligned in phase/time and their summed response is flat, or it isn't.
In the case of impulsive response, I've found that time/phase misalignments are extremely audible--as little as 45-90 degrees of relative phase--to the point that it will make you want to leave the room if you don't fix it after dialing the alignment off using a digital crossover. Timbre shifts and impulsive "smearing" is very audible with separate drivers/horns, and somewhat less so with multiple entry designs (i.e., there are no "lobes"). Dialing in the time alignment of a tweeter-midrange, or a midrange-bass bin isn't difficult to do by ear-only with a digital crossover, assuming that you're within 180 degrees of phase to start with.
YMMV.
Chris
In the crossover region, either the drivers are aligned in phase/time and their summed response is flat, or it isn't.
In the case of impulsive response, I've found that time/phase misalignments are extremely audible--as little as 45-90 degrees of relative phase--to the point that it will make you want to leave the room if you don't fix it after dialing the alignment off using a digital crossover. Timbre shifts and impulsive "smearing" is very audible with separate drivers/horns, and somewhat less so with multiple entry designs (i.e., there are no "lobes"). Dialing in the time alignment of a tweeter-midrange, or a midrange-bass bin isn't difficult to do by ear-only with a digital crossover, assuming that you're within 180 degrees of phase to start with.
YMMV.
Chris
Hi Wesayso,
I'm a little confused because I thought I was asking something very simple. Let me use some fake data.
First, a two way, phase aligned but not time aligned step response:
As you can see, the tweeter arrives at t=0, before the woofer about t=0.3 milliseconds, causing the classic spike, then triangle. Now, in a time aligned system, the step response would look more like this:
Obviously, this is a mock up, but the point is, the initial arrival of the woofer and tweeter are coincident. They merge to form the single rise at t=0.
What I'm curious about is, given the same drivers, if there's a significant audible difference, and if so what exactly that is. Another way to ask this is, what audible benefits are there by moving the tweeter away from the woofer? Better imaging, more sparkle? Spaciousness?
I'm also aware that electrical vs. acoustical delays are not equivalent in terms of the polar reponse. I'm not interested in that right now. Just want to know, WHY bother time aligning a tweeter to a woofer?
Thanks,
Erik
I'm a little confused because I thought I was asking something very simple. Let me use some fake data.
First, a two way, phase aligned but not time aligned step response:
As you can see, the tweeter arrives at t=0, before the woofer about t=0.3 milliseconds, causing the classic spike, then triangle. Now, in a time aligned system, the step response would look more like this:
Obviously, this is a mock up, but the point is, the initial arrival of the woofer and tweeter are coincident. They merge to form the single rise at t=0.
What I'm curious about is, given the same drivers, if there's a significant audible difference, and if so what exactly that is. Another way to ask this is, what audible benefits are there by moving the tweeter away from the woofer? Better imaging, more sparkle? Spaciousness?
I'm also aware that electrical vs. acoustical delays are not equivalent in terms of the polar reponse. I'm not interested in that right now. Just want to know, WHY bother time aligning a tweeter to a woofer?
Thanks,
Erik
Attachments
Last edited:
Time alignment in a conventional two way can offer practical benefits in one or two dimensions just as the unity concept can offer practical benefits in three dimensions. Understanding this, as well as putting faith in published group delay audibility threshhold studies can help to side-step concerns related to this excess group delay.
Erik,
Whether time-aligned and phase aligned, or only phase aligned, the alignment is crucial for a flat frequency response, which is sonically a "bigger deal" than actual time alignment.
With a careful A/B test one (someone-not me) may notice the time aligned speaker sounds slightly more "real" than one simply phase aligned. I say "may" because the perception of the difference is not "big", and is confounded by the Haas effect- what is heard first sounds louder, so the tweeter leading the woofer could make the high frequency transients seem more "present" or "forward" or "brighter" than a time aligned speaker.
The subtle difference is also masked because although we can hear timing differences of much shorter duration at high frequencies than low frequencies, the wavelength duration of high frequencies are much shorter than we can perceive a discreet echo, so the portion of a wavelength needed to align phase using analog filters and/or polarity reversal is short compared to perception. The upper echo threshold varies in length from just a few milliseconds to as much as 50 ms depending on the individual listener and type of stimulus.
With such a large difference in people's perception, whether time alignment will make a difference to an individual can only be determined by making the test yourself.
Merlign van Veen created a video in which you can determine for yourself the audibility of phase shift:
https://www.merlijnvanveen.nl/en/study-hall/100-the-harmonics-lead-the-fundamental
And more discussion:
http://forums.prosoundweb.com/index.php/topic,156515.20.html
Art
Last edited:
Hi Wesayso,
I'm a little confused because I thought I was asking something very simple. Let me use some fake data.
First, a two way, phase aligned but not time aligned step response:
As you can see, the tweeter arrives at t=0, before the woofer about t=0.3 milliseconds, causing the classic spike, then triangle. Now, in a time aligned system, the step response would look more like this:
Obviously, this is a mock up, but the point is, the initial arrival of the woofer and tweeter are coincident. They merge to form the single rise at t=0.
What I'm curious about is, given the same drivers, if there's a significant audible difference, and if so what exactly that is. Another way to ask this is, what audible benefits are there by moving the tweeter away from the woofer? Better imaging, more sparkle? Spaciousness?
I'm also aware that electrical vs. acoustical delays are not equivalent in terms of the polar reponse. I'm not interested in that right now. Just want to know, WHY bother time aligning a tweeter to a woofer?
Thanks,
Erik
Not trying to confuse you
, but may I ask what kind of crossover you simulated in the second example?
Wesayso,
First image is of my current main speaker simulation.
For the second I just wanted to present what I've seen presented as the step of a near-ideal time aligned speaker.
Truth is it's an XSim ideal speaker with a high pass filter on it to make it reasonably bandwidth limited. I just didn't want to get into actual crossover design. I wanted to ask if there is a benefit of spending the time and money to get from the first to the second.
Best,
Erik
First image is of my current main speaker simulation.
For the second I just wanted to present what I've seen presented as the step of a near-ideal time aligned speaker.
Truth is it's an XSim ideal speaker with a high pass filter on it to make it reasonably bandwidth limited. I just didn't want to get into actual crossover design. I wanted to ask if there is a benefit of spending the time and money to get from the first to the second.
Best,
Erik
FWIW, I have tried this two ways:
1. Create a FIR filter that reproduces the time-domain response of a 3-way system I have (modified B&W CM9). Make sure the frequency response is exactly flat with unity gain. Using headphones (Grado) that do a reasonable facsimile of a time-aligned response, I cannot hear any difference between filter and no filter.
2. Create a FIR filter that corrects the time-domain response of the CM9s. Again the filter response is exactly flat with unity gain. With this I get nice square waves from the speaker over a rather small volume in 3D space (i.e. the sweet-spot for time-alignment is small). There might be a slight difference between filter and no filter, but I'm not certain. The way I had things configured there was a small click when the filter was enabled, so the transition was identifiable. If there was a difference, it was very subtle. Imaging did not seem to be affected. Note that this method of time correction does not alter the speaker's radiation pattern, frequency response or power response. That would not be true when switching from a conventional configuration to, say, a Harsch.
1. Create a FIR filter that reproduces the time-domain response of a 3-way system I have (modified B&W CM9). Make sure the frequency response is exactly flat with unity gain. Using headphones (Grado) that do a reasonable facsimile of a time-aligned response, I cannot hear any difference between filter and no filter.
2. Create a FIR filter that corrects the time-domain response of the CM9s. Again the filter response is exactly flat with unity gain. With this I get nice square waves from the speaker over a rather small volume in 3D space (i.e. the sweet-spot for time-alignment is small). There might be a slight difference between filter and no filter, but I'm not certain. The way I had things configured there was a small click when the filter was enabled, so the transition was identifiable. If there was a difference, it was very subtle. Imaging did not seem to be affected. Note that this method of time correction does not alter the speaker's radiation pattern, frequency response or power response. That would not be true when switching from a conventional configuration to, say, a Harsch.
Sadly it isn't as simple as switching to active crossovers and use the proper time alignment to get from case 1) to case 2). There are a lot of options to get there though. Is it worth it? To me it is.
If you are really interested in seeing what it can do check out the http://www.diyaudio.com/forums/multi-way/285030-bookshelf-multi-way-point-source-horn-126.html#post4644738 thread where I asked bushmeister to check 2 of his speakers at the listening position. He posted the results from the synergy/unity horn which was the subject of that thread and compared it to his prestige speakers. In post #1329 you can see the results of FIR correction of that synergy speaker with the Harsch x-over and what that does for the measurements. You will have to determine for yourself if it's worth it to try something like that. Pay notice to the early waterfall plot differences between the corrected and uncorrected speaker. (though I believe the uncorrected speaker isn't presented right in that first plot)
Even though this wasn't exactly a perfect test, the speaker had been moved, it's still remarkable what the differences look like, aside from a cute looking STEP. These tests were done at the listening position.
Personally I believe the synergy/unity concept is a very good start to do this kind of experiment with. Think about how that step is going to change if you move left or right, or even up or down. Now think about a more conventional design. Think in distances to the source of the sound.
Hope this get's you thinking a little more
. I wouldn't try it on just any speaker, some have better potential than others by design. That doesn't mean you couldn't try it with other types of speakers. But I do know why I picked my particular design.** the FIR correction used is my own recipe, based on the DRC-FIR software, it wasn't done with standard templates **
How did thou come to the first plot, Erik? A linear-phase crossover, such as of first order or digital ones, with drivers much closer together than lambda has the step response of the second plot. As soon as one wants to equalize analogue phase distortions, and time distortion being a sort of phase distortion, one needs computers such as cascaded allpasses, bucket chains or binary ones.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.