So I can now get ruler-flat phase from the driver's impulse response using my homebrew active crossover software, with a selectable degree of smoothing. But what is the rationale behind correcting the phase and not the amplitude? My best efforts with corrected phase-only seem to give me improvements(?) in the stereo image but (and I may be imagining this because I'm half-expecting it) I'm not convinced that it's not also producing some 'uneasiness' in my ears. Perhaps transients are less 'solid'.
If I correct the amplitude as rigorously as the phase, I'm not sure I like the sound as much as the uncorrected (it is testing measurement technique and microphone quality in a much more clearly audible way than phase alone). However, and again I may be imagining it, it does seem to get rid of the 'uneasiness'.
The definition of minimum phase says that phase and amplitude are inextricably linked. Does correcting one without the other mean that the system is no longer minimum phase and our ears detect this as unnatural? The phase is then effectively floating around arbitrarily compared to the amplitude, and vice versa with no way for the ear/brain to link the two together.
If I correct the amplitude as rigorously as the phase, I'm not sure I like the sound as much as the uncorrected (it is testing measurement technique and microphone quality in a much more clearly audible way than phase alone). However, and again I may be imagining it, it does seem to get rid of the 'uneasiness'.
The definition of minimum phase says that phase and amplitude are inextricably linked. Does correcting one without the other mean that the system is no longer minimum phase and our ears detect this as unnatural? The phase is then effectively floating around arbitrarily compared to the amplitude, and vice versa with no way for the ear/brain to link the two together.
Attachments
What are your crossover points?
Are you doing the whole filtering with phase linear crossovers?
If you only correct the allpass of your acoustical crossovers (with the "filter linearizion" tab if you use IIR crossovers, and some additional phase EQ to get it right, as the acoustical slopes are never completely textbook) and let the rest untouched (the small variations along the passband and the bass rolloff) then you should end up with something similar to a minimum phase "perfect" fullrange driver response-wise (on axis...).
Then if you also correct the bass rolloff the system becomes non causal, as the (unavoidable) magnitude rolloff does not come together with its "natural" phase shift...
Same thing for all the small phase glitches that you might correct along the passband (and that you should maybe let alone, as well as the magnitudes ones, if you are not sure they are not artifacts of a particular measurement and are not repeatable along several measurement points).
I should really include minimum phase EQ in rePhase but I have been to busy (err lazy) to put any serious work on it lately...
Are you doing the whole filtering with phase linear crossovers?
If you only correct the allpass of your acoustical crossovers (with the "filter linearizion" tab if you use IIR crossovers, and some additional phase EQ to get it right, as the acoustical slopes are never completely textbook) and let the rest untouched (the small variations along the passband and the bass rolloff) then you should end up with something similar to a minimum phase "perfect" fullrange driver response-wise (on axis...).
Then if you also correct the bass rolloff the system becomes non causal, as the (unavoidable) magnitude rolloff does not come together with its "natural" phase shift...
Same thing for all the small phase glitches that you might correct along the passband (and that you should maybe let alone, as well as the magnitudes ones, if you are not sure they are not artifacts of a particular measurement and are not repeatable along several measurement points).
I should really include minimum phase EQ in rePhase but I have been to busy (err lazy) to put any serious work on it lately...
Have you applied these techniques to small full range setup?
Information transmission line with periodic IR is correctable. Desired property is output matching input for all in band frequencies both in amplitude and phase. Passage of in band signal representing minimum phase IR is then perfect.
The vast majority of multiway speaker systems are not minimum phase to start with. Equalizing such systems to have perfect band pass response using minimum phase equalization will therefore not correct phase. So whether you use MP EQ or nonMP eq in such cases makes little difference. If MP Eq is used the amplitude correction will also impose phase corrections. If you use amplitude only corrections the phase will remain unaltered. The two approaches with yield different system phase but in neither case with the EQ'ed system be MP.
Yes but in case of a multiway loudspeaker only linear phase crossovers can give a minimum phase system (summed responses)
There are lots of non-linear phase crossovers that sum to minimum phase systems. Examples are, of course, 1st order crossovers, the B&O filler drivers approach, there is a class of 2nd order crossovers, and the family of subtractive crossovers. We don't see many of them implimented because they don't always have the best polar response.
Fair enough John, so remove the "only" in my previous post 
Using phase minimum EQ in the driver passband (and around) and then using linear phase (or linearized) complementary crossovers looks like a good practical way to obtain a minimal phase system from a multiway loudspeakers.
Of course it is not always easy to get the phase right in the passband with only minimum phase EQ, as it requires lots of EQ well outside of the passband, down low (Linkwitz transform).
Using phase minimum EQ in the driver passband (and around) and then using linear phase (or linearized) complementary crossovers looks like a good practical way to obtain a minimal phase system from a multiway loudspeakers.
Of course it is not always easy to get the phase right in the passband with only minimum phase EQ, as it requires lots of EQ well outside of the passband, down low (Linkwitz transform).
Yes, that is one pretty good approach. That is the UE approach, but actually extending the MP well into the stop bands. But there is an advantage to using minimum phase filters (or targets) and then using phase linearizion to eliminate GD associated with the low frequency cut off of the system. That is that the individual pass bands with have causal response. Thus, there will be no pre-ringing in the individual pass bands which may become audible off axis. I've never been able to hear the pre-ringing, though I can measure it. Other claim it is very audible.
Could you elaborate?
As I see it linearizing the high pass of the system (BR + optional subsonic) will give the so called pre ringing, as the impulse of the high pass will become symmetrical.
I don't see how this would prevent the other crossovers in the system to show pre ringing as well wherever the crossed-over drivers are not properly summed in phase (somewhere off axis).
Yes. But if there is nothing outside the pass band, you will not see any pre-ring. I make this point in "Is Linear Phase Worthwhile" and so does Dick Heyser & Bode.
That's why I don't mind 'linear phase' antil-aliasing filters. The frequency distribution of real life sounds means you will only see pre-ring with signals which are likely to leave you with a headache for 24hrs.
However, all reliable instances of excess phase detection have been at LF (& mid freq) so there is likely to be audible effects of "correcting" the LF phase. My instinct is to leave the LF phase alone. It will require a VERY large FIR and is unlikely to make things sound 'better'.
If anyone wants to repeat my 1980 Blind Listening Tests with better equipment & sources, I'm happy to advise.
If you 'correct' the LF phase due to the roll-off you will have pre-ringing regardless of how things sum.
If you DON'T correct the LF phase and the crossovers are Min. Phase .. they will be causal (no pre-ring) but not MP .. and this is also the case off the design axis when they don't sum perfectly.
Last edited:
Hi,
This paper may be of some help:
http://www.bodziosoftware.com.au/Pre_Post_Ringing_IR_And_Pulses.pdf
BeEst ReEgards,
Bohdan
This paper may be of some help:
http://www.bodziosoftware.com.au/Pre_Post_Ringing_IR_And_Pulses.pdf
BeEst ReEgards,
Bohdan
Thanks for this Bohdan.
I personally know all the people associated with the listening tests in The Audibility of Loudspeaker Phase Distortion at Celestion and from talking to them, its obvious the test wasn't carried out properly.
I actually carried out the same test at Wharfedale with the Essex Equaliser before they did the test at Celestion. To our surprise, in my Double Blind bla bla tests, all those who could reliably tell the difference preferred the non-phase corrected version.
We traced this to the fact that the MP EQ was done with a IIR which was self-dithering, while the excess phase correction was via a FIR which wasn't.
I pointed out both these points to Greenfield & Hawkesford.
They later presented another paper On the Dither Performance of High-Order Digital Equalization for Loudspeaker Systems
I was miffed cos they didn't acknowledge that we had observed the phenomena first and told them about it.
____________________
I'm interested in your comments about what you heard with your LP EQ vs MP EQ.
What bandwidth was the 18" subwoofer you were listening on? Was this the 20Hz 4th order system?
I know from my own Blind Listening Tests that a 2ms pulse can differentiate Excess Phase but I've not been able to find musical or real life sounds which lead to reliable detection.
You might also like to look at On the Audibility of Midrange Phase Distortion in Audio Systems though I have some reservations about the speakers & headphones they used.
____________________
My $0.02
- Excess Phase distortion as introduced by speakers is audible under some circumstances by some people.
- The jury is still out over whether it makes things sound worse cos the very small number of people who can detect it.
- It's probably not worth taking special efforts to correct it unless it can be done at little or no cost & complexity.
- The benefits of 21st century DSP are better utilized to improve more traditional speaker faults like Directivity and Delayed Resonances. These are known to be audible to even the unwashed masses.
Hi kgrlee,
Thanks for your pointers. I am familiar with Hawkesford works.
I have collected my impressions on linear phase loudspeakers and summarized them in:
http://www.bodziosoftware.com.au/Attributes_Of_Linear_Phase_Loudspeakers.pdf
If you are specifically interested in subwoofers, they are described in:
http://www.bodziosoftware.com.au/LP_MP_Subwoofer_Tests.pdf
The 18" subwoofers after HBT linearization and linear-phase are 16Hz, 4th order systems.
Best Regards,
Bohdan
Thanks for your pointers. I am familiar with Hawkesford works.
I have collected my impressions on linear phase loudspeakers and summarized them in:
http://www.bodziosoftware.com.au/Attributes_Of_Linear_Phase_Loudspeakers.pdf
If you are specifically interested in subwoofers, they are described in:
http://www.bodziosoftware.com.au/LP_MP_Subwoofer_Tests.pdf
The 18" subwoofers after HBT linearization and linear-phase are 16Hz, 4th order systems.
Best Regards,
Bohdan
Bohdan, we are on the same page.
I work with Farina's Kirkeby inverse instead of HBT correcting system responses.
Thanks for links. Yes, time domain performance is needed for realism.
In domestic listening space, room response of sub system is inseparable from room. Room is speaker.
I've aligned sub setup in room from microphone at dust cap to >2m with little change in results.
I get impression that people who talk about audibility of linear phase acoustic response pre-ringing are generating impulse response, playing it back and going "Oh, there hear that."
It's as silly as asking for interconnects with minimum phase response.
Most don't get concept of convolving continuous time program with periodic IR of transmission system to get final response.
Indeed, coupling capacitors forming low frequency high pass filters mess up phase response, and then woofer finishes it off.
A square wave is asking driver to slew rapidly between two states and then hold position. But this is suppose to be SPL. Pressure wave is moving away from driver, so driver needs to keep moving to maintain flat SPL, right up to next slewing event to opposite pressure state.
Above is cardioid sub response as "U" frame stacked with sealed, corrected in room. Original post is in thread for Linkwitz inspired two-way DSP Pluto Clone.
I've linked to this in other threads here, most just don't get it at all.
I work with Farina's Kirkeby inverse instead of HBT correcting system responses.
Thanks for links. Yes, time domain performance is needed for realism.
In domestic listening space, room response of sub system is inseparable from room. Room is speaker.
I've aligned sub setup in room from microphone at dust cap to >2m with little change in results.
I get impression that people who talk about audibility of linear phase acoustic response pre-ringing are generating impulse response, playing it back and going "Oh, there hear that."
It's as silly as asking for interconnects with minimum phase response.
Most don't get concept of convolving continuous time program with periodic IR of transmission system to get final response.
Indeed, coupling capacitors forming low frequency high pass filters mess up phase response, and then woofer finishes it off.
A square wave is asking driver to slew rapidly between two states and then hold position. But this is suppose to be SPL. Pressure wave is moving away from driver, so driver needs to keep moving to maintain flat SPL, right up to next slewing event to opposite pressure state.
An externally hosted image should be here but it was not working when we last tested it.
Above is cardioid sub response as "U" frame stacked with sealed, corrected in room. Original post is in thread for Linkwitz inspired two-way DSP Pluto Clone.
I've linked to this in other threads here, most just don't get it at all.
Actually at LF, it is simple to get pressure response flat. The smaller the room, the easier it is.
There are 3 regions for EQ of speaker in rooms.
- At HF, the speaker is sufficiently far from boundaries to enable EQ of the 'anechoic' response. This has been the focus of most Digital EQ efforts in the last decade or so. It is known to give good results.
- At LF, when room dimensions are the same or greater then the wavelength, the room is a 'point' and very simple EQ gives good results.
- The difficult part is the 'midrange' where there are significant reflections which can't be distinguished from the direct response but cause complicated response changes.
It is the 3rd region where I believe room & speaker EQ still needs a lot of research.
We know that speakers in anechoics sound terrible. So what are we trying to achieve? With respect to Bohdan and other workers in this field, I think the jury is still out on this question.
I have not done serious work on this for well over a decade so am interested in what others have found. This millenium, the DSP power available to us is far greater than the early 90's. But what should we be using it for?
Barley, how different are the EQs for the 2 channels on your final system? Can you post a curve(s)?
You mean you can get linear phase interconnects?
All the interconnects I've measured have been Minimum Phase. They satisfy one of the necessary AND sufficient criteria.
Last edited:
You are mistaken. At such frequencies the response in in the modal region and every position in the room has a different response. At such frequencies what matters are the near field response of the woofer, the listening position, and the source to listener transfer function. This last item is a function of the room and the positions of the source and listener. Change either and the response changes. It is not possible to ever equalize a room to have more than a single point with flat low frequency response let alone linear phase.
It is only when the wave length is much greater than all the room dimensions that the SPL tends to be come uniform. But then there are a number of significant other issues which effect the response such as leakage and energy storage in the structure. Such frequencies are typically below those produce by musical instruments and are in the realm of HT LFEs.
Last edited:
It's a first order rolloff, so they sum flat in quadrature.You mean you can get linear phase interconnects?