Hi Everyone,
I haven't been here in a while but a recent built by Troels Gravesen has made me think about building a new, 3-way active center. Mostly I'm intrigued by the Illuminator 12MU he used.
Here's my question. Assuming I go fully active with a similar design, is there any benefit to digitally delaying the tweeter or tweeter/mid to co-locate the acoustic plane for all 3 sets of drivers?? In simulations I've tried before using 2-way speakers it seemed the only real benefit was smoother off-axis response. Assuming I went with a 3-way active amp/crossover is it going to be worthwhile to attempt to simulate this using digital delays?
Thank you!
Erik
I haven't been here in a while but a recent built by Troels Gravesen has made me think about building a new, 3-way active center. Mostly I'm intrigued by the Illuminator 12MU he used.
Here's my question. Assuming I go fully active with a similar design, is there any benefit to digitally delaying the tweeter or tweeter/mid to co-locate the acoustic plane for all 3 sets of drivers?? In simulations I've tried before using 2-way speakers it seemed the only real benefit was smoother off-axis response. Assuming I went with a 3-way active amp/crossover is it going to be worthwhile to attempt to simulate this using digital delays?
Thank you!
Erik
PS - So how I understand it is that even co-locating the drivers acoustically doesn't automatically give me a perfectly time-aligned speaker, a'la Thiel or Vandersteen, which I think I'm OK with since none of my other speakers are that way.
And, a little off topic for my heading, sorry: What do you all think of using an under-sized, sealed cabinet and correcting with the EQ??
An under-sized, sealed cabinet corrected with EQ works fine if the drivers excursion and thermal potential meet the SPL required.
The length of digital delay required to co-locate the acoustic plane for all 3 sets of drivers as depicted in the OP would be very little, so the benefit would also be fairly little, but as there are no downsides in use of delay in DSP, no reason to not use it.
The length of digital delay required to co-locate the acoustic plane for all 3 sets of drivers as depicted in the OP would be very little, so the benefit would also be fairly little, but as there are no downsides in use of delay in DSP, no reason to not use it.
Yes, we're talking small distances here, 3" or less. I just wonder, since it doesn't really make the crossover slopes any easier how much worth while it would be to measure the acoustical difference and then compensate for them. Of course, being done digitally this compensation would only work on the forward axis.
To answer your question, the result is no different whether you use digital delay adjustment or the conventional way of doing it using a passive crossover.
The great thing about a DSP is that you could load the crossover in slot, and then load another in another slot, this time with the delay, and test/listen for yourself.
The smoother off-axis response, if it results in a smoother in-room response, should result in an improvement.
How audible is a different matter? A night and day difference, or more like late morning and early afternoon difference?
I haven't seen any evidence of acoustic benefits, but keep an open mind.
The smoother off-axis response, if it results in a smoother in-room response, should result in an improvement.
How audible is a different matter? A night and day difference, or more like late morning and early afternoon difference?
I haven't seen any evidence of acoustic benefits, but keep an open mind.
If you can create a proper delay network with passive crossover (that's the hard part), then it is the same as the digital delay, otherwise not. I mean, not just altering the phase-rotation (this is the usual case case with passive crossovers), but precisely delaying the whole band of a driver, these are different things.
"
It should be smoother on and off-axis due to the phase?
But I have to say I never A/B tested without it.
The above is from me inverting the tweeter and adding small amounts to the delay. The top line is with the tweeter back to normal and with the delay.
To flip the question, do you want the drivers to be -5db or -30db in phase?
Is digital delay worth it in an active speaker?
Yes.It should be smoother on and off-axis due to the phase?
But I have to say I never A/B tested without it.
The above is from me inverting the tweeter and adding small amounts to the delay. The top line is with the tweeter back to normal and with the delay.
To flip the question, do you want the drivers to be -5db or -30db in phase?
Last edited:
YSDR, erik's first question I wanted to answer is regarding polar differences, and since electrical variations don't change one axis without changing the others, all crossovers do the same thing there.
It's the same thing. Phase is just one frequency's contribution towards it, that's why when phase varies you get group delay.
It is still not the same thing, because to alter the phase relation with passive crossover, you need to change the frequency response/crossover slope, so the exact phase matching occurs only at one frequency. With digital delay, you can time align the drivers to get the exact phase relation you want at all frequencies without changing the frequency response.
For simple example, we want a 24dB/octave Linkwitz-Riley crossing. If the acoustic centers of the drivers are not in the same plane, then you need to use slightly assymmetrical slopes (this determines the slight phase misalignment above and below the crossover point) in the low- and high-pass to align the phase at the crossover region to get a forward acoustical lobe. This can be achieved relatively easy with passive crossover.
But to if you want a symmetrical crossing (this is the only way to get exact phase matching at all frequencies without FIR filtering), you need to use symmetrical acoustical low- and high-pass slopes which can be phase aligned with time delay, like stepped baffle or digital delay.
So in the end, passive crossover aligning and digital time delay aligning are similar but not the same thing.
For simple example, we want a 24dB/octave Linkwitz-Riley crossing. If the acoustic centers of the drivers are not in the same plane, then you need to use slightly assymmetrical slopes (this determines the slight phase misalignment above and below the crossover point) in the low- and high-pass to align the phase at the crossover region to get a forward acoustical lobe. This can be achieved relatively easy with passive crossover.
But to if you want a symmetrical crossing (this is the only way to get exact phase matching at all frequencies without FIR filtering), you need to use symmetrical acoustical low- and high-pass slopes which can be phase aligned with time delay, like stepped baffle or digital delay.
So in the end, passive crossover aligning and digital time delay aligning are similar but not the same thing.
Last edited:
In short, a digital time delay delays all frequencies by the specified time. Whereas (passive) crossover phase altering (by altering the crossover slopes) are not.
Last edited:
No, it is possible with any kind of crossover to change phase without changing response.
Here is a simulation of a phase shift, showing that the wave is delayed.
Here is a simulation of a phase shift, showing that the wave is delayed.
The two waves looks different. With digital delay, they would not differ from each other just one would be delayed, compared to the other.
This example shows that they are different things.
This example shows that they are different things.
Last edited:
What YSDR is saying it’s not easy to do with a passive crossover for crossovers like acoustic LR2. To do it properly you need a ladder delay network. And so it uses a lot do components.
Here’s a well documented example:
paragraph 5 from
(Copy)
There are 3 passive solutions to the problem of mismatched acoustic center offsets common to LR2 designs. The first and most common is asymmetrical crossover slopes and the other common solution is using a slanted baffle. These work but are not optimal. Asymmetrical crossover slopes cause poor phase tracking around the crossover point which can remove a level of coherency depending on how severe the phase angles don't line up. (this is debatable and unproven however) There's no problem with crossovers that are mildly asymmetrical, but if you have to go more than one order in either direction on either slope, there can be some problems with the drivers not summing to flat. Slanted baffles represent a couple a different problems. First off, the drivers are off axis, which affects vertical off axis response and rolls off the top end response of both drivers.
… A ladder delay network is the best solution in this system. It allows us to use a flat baffle, getting us the most out of the (tweeter’s) top end and giving us good phase tracking around the crossover point and symmetrical lobing when hitting LR2 targets perfectly.
reference:
http://zaphaudio.com/ZD5.html
Here’s a well documented example:
paragraph 5 from
(Copy)
There are 3 passive solutions to the problem of mismatched acoustic center offsets common to LR2 designs. The first and most common is asymmetrical crossover slopes and the other common solution is using a slanted baffle. These work but are not optimal. Asymmetrical crossover slopes cause poor phase tracking around the crossover point which can remove a level of coherency depending on how severe the phase angles don't line up. (this is debatable and unproven however) There's no problem with crossovers that are mildly asymmetrical, but if you have to go more than one order in either direction on either slope, there can be some problems with the drivers not summing to flat. Slanted baffles represent a couple a different problems. First off, the drivers are off axis, which affects vertical off axis response and rolls off the top end response of both drivers.
… A ladder delay network is the best solution in this system. It allows us to use a flat baffle, getting us the most out of the (tweeter’s) top end and giving us good phase tracking around the crossover point and symmetrical lobing when hitting LR2 targets perfectly.
reference:
http://zaphaudio.com/ZD5.html
Yes, that's what I thought. If we can do a proper passive delay network, then it is the same as a digital delay, otherwise it is not.
Hi, yes it will definitely be worthwhile to use digital delays if you have gone multi-way active.
It's one of the clear superiorities of DSP vs passive......the ability to implement very fine increments of true fixed time delay.
Delays can range in time from a single sample (about 0.02ms @48kHz or a fat 1/4" distance; to x number of feet, and still with 1/4" distance precision).
Good luck trying to achieve either that precision, or range of distance, with a passive or analog ladder network.
3" btw, can be huge. What is the xover frequency where the 3" distance error resides?
Evaluate how many degrees of phase rotation 3" equals at that freq, and use the well known phase summation wheel to get an idea of combing/lobing.
Using digital delay to equalize time-of flights for driver's raw acoustic measurements, before adding xovers, makes tuning a speaker much easier with active processing.
I find that simulations are almost a waste of time. Get delays, polarities and levels right, then simply EQ each drivers' min phase response, and add xovers.
The closer to acoustic complementary that is achieved with the min phase EQ's, the less timing shift will occur post xovers. No timing shift at all is possible when adding complementary linear phase xovers.
The biggest tell i've found that correct timing matters, is that i'll make a new system tuning and after listening a bit sometimes go hmm...sounds good, but somethings off.
Ok, let's remeasure. Freq response looks fine....how about phase? Whoops, there it is....bad delay setting....
Hi @YSDR
I guess you are hitting the crux of the matter. If I build everything on 1 plane, regardless of active/passive I can't use ideal slopes becuase my phase alignment wont' be ideal. Your example of using LR4 really helped me.
Let me see if I understand you. Assuming all active/digital for a bit, if I set delays = 0 for all, I can't use ideal slopes, I somehow have to jimmy (tweak) the slopes and crossover points and maybe even EQ the drivers into mating well, but if I use the digital delays I can use ideal LR4 alignments and my life gets easier? Is this true for other alignments??
I guess you are hitting the crux of the matter. If I build everything on 1 plane, regardless of active/passive I can't use ideal slopes becuase my phase alignment wont' be ideal. Your example of using LR4 really helped me.
Let me see if I understand you. Assuming all active/digital for a bit, if I set delays = 0 for all, I can't use ideal slopes, I somehow have to jimmy (tweak) the slopes and crossover points and maybe even EQ the drivers into mating well, but if I use the digital delays I can use ideal LR4 alignments and my life gets easier? Is this true for other alignments??
Thank you @mark100 , That is exactly the kind of advice I was asking about. 🙂 I'm all for easier, and cleaner crossover design. Having worked with passive crossovers, and tweaking with XSim I didn't think it was too difficult to do these slope/phase adjustments but if using digital delay to compensate for the physical offsets is the nicest way to go then that's what I'll do.
One thing I didn't take into account is that using passive components, and a zobel or other tricks one can make rather fine adjustments to the electro-acoustical roll-off slope of a driver. From my experience with miniDSP though things are a bit rougher. Only precise 6 dB/octave slopes and EQ settings are allowed, so it seems based on the advice above that having precise time alignment will make things much easier for me.