This study is a good example that takes into account many previous studies on phase audibility. The results would certainly suggest that phase is not a serious consideration for designing a crossover/loudspeaker.
Chapter 5
Just look at the results. Unless you plan on listening to test tones, it's not a good reason to design for first order filters. Of course you still have the superior hearing argument.
Dan
Chapter 5
Just look at the results. Unless you plan on listening to test tones, it's not a good reason to design for first order filters. Of course you still have the superior hearing argument.
Dan
Exactly. Here's a midrange driver (TD6M) in a closed box tuned to Q=.5 at 105 with 30 watts applied. Yellow is the SPL and cone excursion with no filter and green is with a 6dB electrical highpass at 500. Of course the green acoustic SPL curve is nothing like first order so all the supposed 'phase advantages' are somewhat moot but at least the simple filter is providing adequate driver protection if you don't get carried away with the volume control. 🙂
Attachments
The issue is more appropriately what occurs with an acoustic first order. That said, no matter how close to a first order acoustic slope is achieved, at some point due to the driver's response in a closed box, it will ultimately revert to second. As long as that point is far enough down in level (possibly the noise floor), it becomes a moot point.
The issue of power loss in the voice coil is valid, one reason why even with second order I've preferred to use an over-sized box and use the crossover to control the highpass, rather than size the box via T/S parameters. Tweeters are their own case, of course, but the issue is pretty much the same otherwise.
Now I'm trying dipoles, so...no box. I suppose that some would claim that dipoles are fraught with likely failure, too. 🙄 But then, drivers are quite often used with higher order crossovers done poorly and they sometimes exceed their thermal limits as well.
Dave
Unless you have listened to 1st order vs a higher order you may not know what the lack of time smear sounds like! Designing speakers is a process full of trade offs. I like using the minimum no. of components in the signal path of a loudspeaker.
Most 1st order speakers require a stepped baffle. This is a trade off! There are a lot of positives such as a much easier crossover with less itinerations to get it right.
Some people like first order, some have never tried it! It's a wonderful world. Excellent loudspeakers have been made in all sorts of configurations. I am not wrong liking 1st order and the people that like all the other orders are also not wrong. 🙂 We are different 🙂 We have choices 🙂
The only thing I object to is information that is clearly wrong 🙁
Terry
What are you saying is "clearly wrong"? If it has to do with first order systems and phase audibility due to the crossover, I'm not aware of any proof that anyone can "clearly hear it" in multi-way speaker systems in a room. Maybe you can provide a reference that does.
I do know what the lack of time smear sounds like. So far I'd say it's not audible in typical crossover regions, but I'm going to try some more. First order can be a great system, I've made them, stepped baffles with significant felt diffraction control and liked them. I'm now working with zero "time smear", fourth order flat phase using DSP, one step past a first order minimum-phase system I'd suggest. There is nothing in it that so far that makes it noticeably better than when I switch it to standard fourth order phase. That takes literally a few seconds to do. It is without doubt a subtle difference if it does prove audible, but so far it's not jumping out at me.
At least with this setup it will be possible to make a test with "all else equal". The one and only difference will be the system phase response. Comparisons of first order minimum-phase to any other systems is not a valid test, since differences other than the system phase response are issues.
At this point I agree with Linkwitz:
Linkwitz on phase distortion
One huge elephant in the room is that no matter what one does with a multi-way system, the area where it can actually be considered transient-perfect (that really is what's at issue) is at best a limited section of an arc that includes the design axis and only at the design distance, since moving closer or farther changes the excess-delay of one or more drivers. Driver directionality limits the arc.
If one prefers first order, I think it's fair to say that it's likely not because of the minimum-phase nature. There are other, more influential differences such as driver overlap and system power response.
Dave
Hello Terry,
(Sorry to hijack this thread again...)
Seems almost to simple. Just to clarify, .22mH in series with the woofer and 2.2uF in series with the tweeter produces a 6db per octave mirrored slope? And no notches or padding. I'm guessing but the component values suggest a crossover frequency around 4kHz -ish. Not doubting you, since you've heard the design, but how loud does the speaker play before the tweeter runs into distortion since it is only down 24db at 500Hz (below fs)?
I've always been fond of Green Mountain Audio speakers which utilize first order crossovers, so it must be possible.
I'm tempted to try this design....
Remember that that system uses 2 woofers for 4Ohm loading. Before you go to far I would want to actually have a look at my notes on that design. In the meantime have a look at http://www.diyaudio.com/forums/multi-way/68301-my-morel-mtm-projects-long-first-post.html
Terry
Last time I looked, some of the more expensive B&W were using 1st order crossovers.
Higher orders are used by a lot of manufacturers to control low quality drivers with poor behaviour. The diyer has the option to make better driver choices to allow the use of lower orders. It's been mentioned in this thread that driver choice is the key to success with 1st order crossovers.
FWIW, I've never had a driver damaged using a 1st order crossover even with high power usage as well as some stupid crossover point errors. I even had one crossing over to a tweeter below it's Fs and it survived.... didn't sound brilliant while playing but it lived and still plays to this day.
My philosophy when designing a speaker is firstly do minimal damage to the signal. That dictates a preference for a 1st order, linear phase design.
Whether you think you can hear it or not, it is irrefutable that conventional high order crossovers discard information.
Whether you think you can hear it or not, it is irrefutable that conventional high order crossovers discard information.
Last edited:
First order minimum-phase. They are not linear phase.
Only in a very small area. Outside of that, they all "discard" information.
I've got nothing against first order. I am simply pointing out that claims that they are superior due to their minimum-phase nature has no support in literature of any kind as far as I'm aware, despite efforts to determine such audibility.
Dave
My comments were not directed to preferences at all nor to whether a design measures well or not. I have preferred second order over time in passive systems, though I know that first order "measures better" in certain aspects though in a limited window. In the end I'm no different, I listen to what I prefer as well.
I would also say that logic may dictate only likelihood that I am correct, because someone may one day prove some level of audibility of phase above the bass region. DSP systems allow for reduction of variables in testing.
The thing is, I can listen to an LR4 system with its group delay and then listen to the same one with linear phase. All else is essentially equal. If minimum-phase or even linear phase made a large difference, I would expect it to be easily discerned. It should be obvious, but evidence suggests it is not. So far it is not to me, but as I said, I'm going to experiment more.
Dave
But thats like saying 'my mother, who isnt an audiophile cant hear the difference between one speaker and another so there isnt any difference'.
AES E-Library: On the Perception of Phase Distortion
There is literature and reserach available if you look for it. That paper above would also explain why you cant hear it.
That's not accurate at all and certainly not what I said. I know without question that there is a measurable difference. I also know that different types of crossovers can have quite audible differences. What is not addressed in most anecdotal claims is the fact that there are more variables at play than phase alone. There are reasons other than phase that are more influential. That it may take headphones and carefully contrived tests is simply more evidence that with most music signals it is not audible.
That reference appears to similar to what Linkwitz did here. The AES paper description includes this:
I may spend the $20 for that paper (I'm not a member). With an actual speaker system that may itself be an allpass (though one could use one of constant power), it would seem likely that adding more phase distortion to an existing one would be problematic.
Interestingly, Linkwitz says this:
I can't say why I cannot yet hear a difference, but I don't yet hear it. That is actually the point. Were it dramatic as is at times claimed, I should hear it easily, certainly when I'm making the attempt. The one thing that I can say is that the only difference in my testing can be reduced to phase alone.
Dave
I accept there are many variables. So the question is which ones have you changed and how can you claim that its been reduced to phase alone?
As I described earlier, I'm using DSP software, specifically SoundEasy v17 and its Ultimate Equalizer. Check item 5 of the DSP functions. This allows one to create a crossover and select whether one wants phase equalization or not.
Here's a link to measured responses using SE v17 by john k:
An externally hosted image should be here but it was not working when we last tested it.
He identifies the curves as:
You can find more info at this thread.
Dave
Last edited:
Many clearly do not understand that most of the benefits of first order xovers can be obtained with stopband performance that is functionally equivalent to conventional second or third order networks. It is true that a bit more is required of the driver, basically good transition band behavior for at least a half octave beyond its xover point for best results.
The 'secret' is to use an out of band notch, preferably for both LF and HF units. For a notch an octave beyond the xover frequency, a minimum stopband rejection of 20db is typical (this is a reduction in power of 100:1, incidentally). For a notch an octave and a half beyond the xover frequency, 30db minimum stopband rejection is usual, and so on and so forth. Two to three octaves beyond its xover point, most competent drivers contribute not much acoustically, so xover attenuation beyond 20-30db at these frequencies should concern only the terminally anally retentive.
So, for accepting a minor phase and amplitude effect in the passband and to somewhat into the transition band for each driver (most of which can be tweaked out with minor adjustments in xover component values) the best phase characteristics of first order xovers plus the best rejection characteristics of 2nd-3rd order filters can be obtained in the same xover network. And sometimes more.
The LF filter notch can usually be centered at the woofers's higher order breakup modes before its ultimate rolloff commences, which cannot be as effectively accomplished with a monotonic slope rolloff filter with anything short of a fourth order characteristic. Yet the result sounds like a phase coherent first order filter on steroids.
With my Iron Lawbreakers, I use an Altec 288-16G driving a 400hz exponential horn xovered at about 550hz. With a plain first order network, the sound noticeably starts falling apart well below the 1 Watt average input level due to horn unloading and signal appearing at and below the resonance of the HF driver's diaphragm. However, when I added a 160hz notch filter, this effect essentially vanished at any playback level my 60W/channel triode OTL could provide because the filter attenuation at any frequency below the horn cutoff was now was a minimum of 15 db and typically 40db or more, and the effect on phase response at the xover and above was minor and easily compensated for.
Wrt vertical lobing, an appropriate choice of driver spacing can help a great deal. With the center to center distance of the Iron Lawbreakers between the HF horn and the JBL 2220A LF driver being on the order of 13", the nulls are nearly directly above and below the speaker and serve to minimize the floor to ceiling room mode near that frequency. A case of making lemonade of out lemons, IMO. And the HF horn handles the setback alignment issue as a side benefit.
The 'secret' is to use an out of band notch, preferably for both LF and HF units. For a notch an octave beyond the xover frequency, a minimum stopband rejection of 20db is typical (this is a reduction in power of 100:1, incidentally). For a notch an octave and a half beyond the xover frequency, 30db minimum stopband rejection is usual, and so on and so forth. Two to three octaves beyond its xover point, most competent drivers contribute not much acoustically, so xover attenuation beyond 20-30db at these frequencies should concern only the terminally anally retentive.
So, for accepting a minor phase and amplitude effect in the passband and to somewhat into the transition band for each driver (most of which can be tweaked out with minor adjustments in xover component values) the best phase characteristics of first order xovers plus the best rejection characteristics of 2nd-3rd order filters can be obtained in the same xover network. And sometimes more.
The LF filter notch can usually be centered at the woofers's higher order breakup modes before its ultimate rolloff commences, which cannot be as effectively accomplished with a monotonic slope rolloff filter with anything short of a fourth order characteristic. Yet the result sounds like a phase coherent first order filter on steroids.
With my Iron Lawbreakers, I use an Altec 288-16G driving a 400hz exponential horn xovered at about 550hz. With a plain first order network, the sound noticeably starts falling apart well below the 1 Watt average input level due to horn unloading and signal appearing at and below the resonance of the HF driver's diaphragm. However, when I added a 160hz notch filter, this effect essentially vanished at any playback level my 60W/channel triode OTL could provide because the filter attenuation at any frequency below the horn cutoff was now was a minimum of 15 db and typically 40db or more, and the effect on phase response at the xover and above was minor and easily compensated for.
Wrt vertical lobing, an appropriate choice of driver spacing can help a great deal. With the center to center distance of the Iron Lawbreakers between the HF horn and the JBL 2220A LF driver being on the order of 13", the nulls are nearly directly above and below the speaker and serve to minimize the floor to ceiling room mode near that frequency. A case of making lemonade of out lemons, IMO. And the HF horn handles the setback alignment issue as a side benefit.
Last edited:
I agree whole-heartedly with thoriated's approach. A variation that I use successfully (and I'm far from the first) is to add a second pole at about one octave beyond the crossover, and then use the notch in the breakup region. As a rule of thumb, the breakup, depending on its severity, needs to be at least 35dB down. Listening to each driver and XO branch alone can be illuminating.
You also will need to use Zobels and conjugates to tame the drivers' impedance; the rolloff is mild enough that compromising it with a rising driver impedance is not at all helpful.
You also will need to use Zobels and conjugates to tame the drivers' impedance; the rolloff is mild enough that compromising it with a rising driver impedance is not at all helpful.
Ponce Seek!
What you call it does not make it what it is.
1) Through a first order high pass filter, displacement requirements of a MF or HF driver do not decreas as signal frequrncy is lowered. For this a higher order filter is required.
2) You may choose drivers using any criteria you like. That does not change the "abuse" that must be sustained by MF and HF drivers when fed signals with content below their operational frequency range.
3) The systems you are dealing with, multiple speakers mounted in baffles, even when operated through single pole filters, are multi-pole systems that exhibit incoherent phased outputs as a function of frequency.
4) "Time smear" is intrinsic to all speaker arrays. If it is not excessive, it is not audible
5) Phase coherence is achievable in higher order crossover filters, where one of the channels is derived from the other via a differential amplifier. In the digital domain, just about any signal conditioning is possible.
6) Full range drivers use a 2nd. order mechanical crossover (a compliance that decouples a "whizzer cone" from the larger LF diaphragm at higher frequencies).
In my view, based on the facts outlined here and in my previous posts, the pursuit of "1st. order loudspeaker systems", is like Ponce de León seeking the Fountain of Youth.
Regards,
WHG
What you call it does not make it what it is.
1) Through a first order high pass filter, displacement requirements of a MF or HF driver do not decreas as signal frequrncy is lowered. For this a higher order filter is required.
2) You may choose drivers using any criteria you like. That does not change the "abuse" that must be sustained by MF and HF drivers when fed signals with content below their operational frequency range.
3) The systems you are dealing with, multiple speakers mounted in baffles, even when operated through single pole filters, are multi-pole systems that exhibit incoherent phased outputs as a function of frequency.
4) "Time smear" is intrinsic to all speaker arrays. If it is not excessive, it is not audible
5) Phase coherence is achievable in higher order crossover filters, where one of the channels is derived from the other via a differential amplifier. In the digital domain, just about any signal conditioning is possible.
6) Full range drivers use a 2nd. order mechanical crossover (a compliance that decouples a "whizzer cone" from the larger LF diaphragm at higher frequencies).
In my view, based on the facts outlined here and in my previous posts, the pursuit of "1st. order loudspeaker systems", is like Ponce de León seeking the Fountain of Youth.
Regards,
WHG
- Status
- Not open for further replies.