Thanks again for all input.
I think I now understand how it works, and that my previous attempt to first align the "global phase" based on distance only was based on wrong assumptions. As soon as you introduce the LR4 filtering and the phase shifts this introduce, anything like a global phase becomes completely irrelevant.
All phase alignment focus will then turn towards aligning the crossover sections and let the phase "warp" in between. Keeping the HP/LP frequencies together makes sense based on the LR4 theory of a flat SPL response, which instead lead to the conclusion to use the delay as the tool to achieve the alignment.
Makes sense and also explains the the design of the Wilson Audio WAMM where each element can be "delayed" individually in relation to the listening position. This makes it possible to fine tune the delay alignment depending on the exact listening position, but also to compensate for component tolerances in the passive crossover.
With a DSP based design like the Hypex plate amps, delay is easy to implement, and the limitation is the step size of approx. 11 microseconds. I will experiment to see if this brings me close enough or if I gain anything by fine tuning the frequencies.
Will get back with results after the experiments are done.
I think I now understand how it works, and that my previous attempt to first align the "global phase" based on distance only was based on wrong assumptions. As soon as you introduce the LR4 filtering and the phase shifts this introduce, anything like a global phase becomes completely irrelevant.
All phase alignment focus will then turn towards aligning the crossover sections and let the phase "warp" in between. Keeping the HP/LP frequencies together makes sense based on the LR4 theory of a flat SPL response, which instead lead to the conclusion to use the delay as the tool to achieve the alignment.
Makes sense and also explains the the design of the Wilson Audio WAMM where each element can be "delayed" individually in relation to the listening position. This makes it possible to fine tune the delay alignment depending on the exact listening position, but also to compensate for component tolerances in the passive crossover.
With a DSP based design like the Hypex plate amps, delay is easy to implement, and the limitation is the step size of approx. 11 microseconds. I will experiment to see if this brings me close enough or if I gain anything by fine tuning the frequencies.
Will get back with results after the experiments are done.
As you've discovered, phase can be manipulated by crossover frequencies and/or delay. In the case of a LR4 filter small corrections in crossover frequencies can be enough vs delay. However, larger frequency adjustments to align phase, in the LR case, can negatively effect frequency response around the crossover. In that case, delay may be more appropriate. ... or, another filter type such as a 4th order BW. That would give you additional leeway in spreading frequencies at the crossover while maintaining a flat(er) overall frequency response.
Bear in mind that EQ used to flatten a drivers individual response also effects the phase of that driver in relation to the others.
Bear in mind that EQ used to flatten a drivers individual response also effects the phase of that driver in relation to the others.
Intermediate report from "the field" (< 1hour evaluation):
I now tried the "delay alignment" approach with some success. Ended up in:
Woofer 0 delay
Mid 32 microseconds
Tweeter 53 microseconds
(listening based, using interval halving)
The assumption that further frequency based fine tuning could improve the alignment was tested. I changed the low pass filters side ending up in 401Hz (was 400) and 2495 Hz (was 2500), based on listening tests (using interval halving down to single Hz level). High pass sides remaining at 400 and 2500.
Listening impressions so far:
More flat SPL curve when the frequencies are kept together, as expected. A significant improvement.
The PRaT factor slightly decreased with the new delay approach, but very small differences.
Conclusion:
The SPL curve flatness improvements are bigger than the decrease of PRaT, so the delay approach is most likely here to stay. I will listen to then both for a while, letting family and friends into the selection, before a final decision.
I now tried the "delay alignment" approach with some success. Ended up in:
Woofer 0 delay
Mid 32 microseconds
Tweeter 53 microseconds
(listening based, using interval halving)
The assumption that further frequency based fine tuning could improve the alignment was tested. I changed the low pass filters side ending up in 401Hz (was 400) and 2495 Hz (was 2500), based on listening tests (using interval halving down to single Hz level). High pass sides remaining at 400 and 2500.
Listening impressions so far:
More flat SPL curve when the frequencies are kept together, as expected. A significant improvement.
The PRaT factor slightly decreased with the new delay approach, but very small differences.
Conclusion:
The SPL curve flatness improvements are bigger than the decrease of PRaT, so the delay approach is most likely here to stay. I will listen to then both for a while, letting family and friends into the selection, before a final decision.
I'm not sure if you are in the right ball park? It should be under a millisecond if not a horn? My 4cm distance between the voice coils got only 0.12 ms with mic measurment.
and seems to be right in a calculation as well:
Frequency, Wavelength, Time and Distance Calculator
Do you have a flat preset to flip between? sound memory is only 3 sec
I can notice a slight improvement.
and seems to be right in a calculation as well:
Frequency, Wavelength, Time and Distance Calculator
Do you have a flat preset to flip between? sound memory is only 3 sec
I can notice a slight improvement.
Attachments
Last edited by a moderator:
The delay serves two purposes, distance compensation and phase alignment. They are then "melted" together into a single number.
My listening tests indicate 75 microseconds distance delay between woofer and mid, and 160 microseconds for the tweeter.
Adding phase alignment to this ends up in a total of 32 and 53 microseconds.
At a crossover frequency of 2500Hz, each full wave length is ~137mm. Each 3.7mm step corresponds to approx. 10 degrees phase shift (137/360 = 0.38mm).
My listening tests indicate 75 microseconds distance delay between woofer and mid, and 160 microseconds for the tweeter.
Adding phase alignment to this ends up in a total of 32 and 53 microseconds.
At a crossover frequency of 2500Hz, each full wave length is ~137mm. Each 3.7mm step corresponds to approx. 10 degrees phase shift (137/360 = 0.38mm).
The LR4 has a flat response, but it is not of minimum phase (allpass behaviour). It introduces a group delay of one wavelength at the crossover point, which remains nearly constant for all lower frequencies.
For 4kHz it will be 130us, which has to be considered when crossing over to the bass speaker.
The LR is popular because of minimum overlap, so you can combine drivers close to the lower and upper reproduction limits, in fact one of the BW2 can be the driver and the second only needs to come from the x-over.. With an asymmetric_shelf filter you can tailor the lower and higher end (as important!) of the speaker to your desired x-over frequencies.
Never-the-less you pay the price with a horrible pulse response as shown in one of the replies.
There are subtractive crossover circuits apart from the 6dB filter which maintain a perfect pulse response and flat frequency, but a lot of knowledge and experience is needed to implement them.
So it is a good idea to start with the simple LR4 and after some time to program more sophisticated filters, so you can hear yourself the difference
For 4kHz it will be 130us, which has to be considered when crossing over to the bass speaker.
The LR is popular because of minimum overlap, so you can combine drivers close to the lower and upper reproduction limits, in fact one of the BW2 can be the driver and the second only needs to come from the x-over.. With an asymmetric_shelf filter you can tailor the lower and higher end (as important!) of the speaker to your desired x-over frequencies.
Never-the-less you pay the price with a horrible pulse response as shown in one of the replies.
There are subtractive crossover circuits apart from the 6dB filter which maintain a perfect pulse response and flat frequency, but a lot of knowledge and experience is needed to implement them.
So it is a good idea to start with the simple LR4 and after some time to program more sophisticated filters, so you can hear yourself the difference
Bansuri,
No way! Are you really in Dolceacqua? I've been there! I bought some wine from Cantina di Dolceacqua, and a bottle from a little old lady who sold wine from the front door of her "house". The owner of C di D graciously gave us a tour of his facilities. Really nice guy. That is a very good memory from many years ago!
No way! Are you really in Dolceacqua? I've been there! I bought some wine from Cantina di Dolceacqua, and a bottle from a little old lady who sold wine from the front door of her "house". The owner of C di D graciously gave us a tour of his facilities. Really nice guy. That is a very good memory from many years ago!
Last edited:
Thank you all for sharing your knowledge and experience, and for teaching me how to do the LR4 crossover phase alignment properly.
As bansuri writes, this is an OK starting point, but most likely not the final destination. And since I'm using a DSP, new experiments are inexpensive (not counting the time...).
As bansuri writes, this is an OK starting point, but most likely not the final destination. And since I'm using a DSP, new experiments are inexpensive (not counting the time...).
My experience with Hypex fusion amp DSP crossovers has been that LR4 has sounded best. I will caveat this statement by saying I am using direct radiator (not horns or waveguides), and my drivers are ceramic/aluminum cones which do not have a nice smooth roll-off in the high frequencies.
However, even after equalizing down the high frequency peaks in the woofer and mid, I found the 1st order, LR2, and BW3 to be less satisfying than the LR4. At the upper 2k crossover, the differences were so small I really could not tell much difference... they all sounded excellent. But at the lower 200 Hz crossover, definitely preferred the LR4.
I also implemented a Harsch crossover as described by xrk971, and this sounded almost as good as the LR4... which is ironic and mysterious, because if we can hear phase, the Harsch and 1st order should have sounded best, followed by LR2 and BW3, with LR4 being last. But if flat magnitude is the criteria, the Harsch should have been last. In short, I have not been able to correlate my subjective preference with any particular phase theory.
However, even after equalizing down the high frequency peaks in the woofer and mid, I found the 1st order, LR2, and BW3 to be less satisfying than the LR4. At the upper 2k crossover, the differences were so small I really could not tell much difference... they all sounded excellent. But at the lower 200 Hz crossover, definitely preferred the LR4.
I also implemented a Harsch crossover as described by xrk971, and this sounded almost as good as the LR4... which is ironic and mysterious, because if we can hear phase, the Harsch and 1st order should have sounded best, followed by LR2 and BW3, with LR4 being last. But if flat magnitude is the criteria, the Harsch should have been last. In short, I have not been able to correlate my subjective preference with any particular phase theory.
One thing that is readily audible to me at low crossover frequencies is driver delay.
At my 200 Hz crossover, I can clearly hear the effects of changing the delay of the mid driver by 2 inches (150 micro seconds). I found that when the delay is properly optimized, the sound becomes more three dimensional, more immediate. It is very noticeable with cello, piano, male voice, sax, drums, trombone, bass clarinet. changing the delay from optimum by 2 inches results in a noticeable degradation.
With my 200 Hz LR4 implementation, the woofer and mid will share the output from about 150 to 300 Hz. In this region, I have found it is critically important for both the woofer and mid to be in tight phase alignment along the listening axis.
At 200 Hz, the wavelength is 67 inches. A 2 inch delay represents only about 10 degrees of phase angle. If I model two drivers at 200 Hz and introduce a 10 degree phase difference, the change in frequency response is trivially different. So what is going on here?
I do not know, but I speculate that I my ear-brain is processing the blend of direct and reflected sound... With tight phase alignment, the ear-brain is fooled into believing there is just one sound source.... With with 2 inches of wrong delay, the ear-brain detects two sound sources, and the illusion of real sound is damaged.
At my 200 Hz crossover, I can clearly hear the effects of changing the delay of the mid driver by 2 inches (150 micro seconds). I found that when the delay is properly optimized, the sound becomes more three dimensional, more immediate. It is very noticeable with cello, piano, male voice, sax, drums, trombone, bass clarinet. changing the delay from optimum by 2 inches results in a noticeable degradation.
With my 200 Hz LR4 implementation, the woofer and mid will share the output from about 150 to 300 Hz. In this region, I have found it is critically important for both the woofer and mid to be in tight phase alignment along the listening axis.
At 200 Hz, the wavelength is 67 inches. A 2 inch delay represents only about 10 degrees of phase angle. If I model two drivers at 200 Hz and introduce a 10 degree phase difference, the change in frequency response is trivially different. So what is going on here?
I do not know, but I speculate that I my ear-brain is processing the blend of direct and reflected sound... With tight phase alignment, the ear-brain is fooled into believing there is just one sound source.... With with 2 inches of wrong delay, the ear-brain detects two sound sources, and the illusion of real sound is damaged.
The white paper for the Grimm LS1 by Putzeys goes into some details about that: https://www.grimmaudio.com/wordpress/wp-content/uploads/speakers.pdf
Updated report from the field:
A listening session with some friends ended up in that everyone, including myself, preferred the "frequency aligned" approach compared to the "delay aligned" approach even with the clearly audible frequencey response drawback.
New experiment:
I tried to align the elements better by slightly tilting the whole speaker backwards toward the wall by putting an 8mm plate under the front "legs". The tilt is approx. 35mm at tweeter level, => ~29mm for the mid and ~20mm for the woofer.
Then I repeated the frequency based phase alignment with new basic delay settings due to the tilt (64 and 128 microseconds), ending up in 0 Hz separation for the mid-woofer alignment (= spot on!), and 124.5 (!) Hz "alignment offset" for the mid-tweeter (tweeter at 2440 Hz and mid at 2564.5 Hz).
The result was chocking!!!
The improvement by far exceeded all my expectations, and I just couldn't stop listening to song after song.
I will have to analyze this more before I make any conclusions.
Please share you thoughts and experiences.
A listening session with some friends ended up in that everyone, including myself, preferred the "frequency aligned" approach compared to the "delay aligned" approach even with the clearly audible frequencey response drawback.
New experiment:
I tried to align the elements better by slightly tilting the whole speaker backwards toward the wall by putting an 8mm plate under the front "legs". The tilt is approx. 35mm at tweeter level, => ~29mm for the mid and ~20mm for the woofer.
Then I repeated the frequency based phase alignment with new basic delay settings due to the tilt (64 and 128 microseconds), ending up in 0 Hz separation for the mid-woofer alignment (= spot on!), and 124.5 (!) Hz "alignment offset" for the mid-tweeter (tweeter at 2440 Hz and mid at 2564.5 Hz).
The result was chocking!!!
The improvement by far exceeded all my expectations, and I just couldn't stop listening to song after song.
I will have to analyze this more before I make any conclusions.
Please share you thoughts and experiences.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.