No problem, my points (and those of John's re-stated) stand on their own. I'm simply looking for direct responses to direct comments and questions.you guys have to get a better grip of yourself
moderation
Dave
None of this is any different for the UE than it is for any other type of crossover. It's a basic issue of design. What sets the UE or any other DSP apart? Nothing, other than the ability to also equalize phase response.
Dave
That is exactly the point. It takes a lot of nonlinearity to get an audible effect, just as Klippel does. But it ius easy to get these nonlinearities below audibility, but then he does get to sell his measurement gear. The whole test is setup to show an effect that is a non-issue in the real world.
I think that the correctness of both Dave and John's points has been well established. I see no errors in them. I can't follow what "fntn" is actually saying, it appears to change, but what does seem clear is not correct. Loudspeakers are very nearly perfectly linear devices (with massive amounts of linear distortions). There deviations from nonlinearity, within their "intended" operating range of output, are extremely small for the most part. If they weren't then the sound would be almost unrecognizable (and yes, I have heard speakers do this when broken or driven too hard, but that should be the exception not the norm). In virtually all of my work I just design conservatively for signal level capability and simply ignore the non-linearities. Perception of nonlinear distortion is grossely over-estimated.
There is no confusion here at all. No one has suggested that these system are perfectly linear. No one has suggested that the bulk of the distortion is linear. However, since the goal of any system design is generally that it perform as a linear system over its intended operating range, it would be negligent to start with components in which nonlinearity was a dominant factor in that range. If you go to Mark k's or Zaph's web sites and look over the measurements they provide, the evidence is clear that for any quality driver linear distortion is the dominant problem. Choose any driver, define the bandwidth over which nonlinear distortion is reasonable and what you have left is the linear distortion. With todays technology designing a driver with low nonlinear distortion over some intended useful range is not the difficult task.
If it is intended to design a speaker than can play at 120 dB then it would be wise, to say the least, to start with components which behave linearly over the require frequency range at those levels. No one is suggesting that the design should start with drivers than exhibit 30% nonlinear distortion and expect anything other than crap.
What has been stated is that these system have linear and nonlinear aspects of their response to an input. And it has been stated and demonstrated that the linear aspects of the system can be corrected by a linear correction without affecting the nonlinearity adversely. It can be called equalization, response shaping, designing a crossover network, or what ever you like. But the idea of bringing the amplitude response into closer agreement with some desired target response using some type of filter between the source and the driver, active, passive, digital or analog, is applying a linear correction. It has never been suggested that such a correction does anything but addresses the linear aspects of the response. It has never been suggested that this is the end all to be all. What has been suggested is that with DSP the degree of accuracy with which this linear correction can be achieve is far superior to that which can be done with analog filter.
Making such a linear correction will not change the way the driver responses to a specific input because the driver has not been changed. Now if I do a 1 V RMS sine sweep on a raw driver and I can get an amplitude vs frequency and some crude measure of distortion such as THD. If I place some response shaping network in the signal path and do another 1 V RMS sweep I will get a different distortion measurement as well as a different amplitude. That is not indicative of a change in the nonlinearity of the driver, it is indicative of the change in the input just as the amplitude response reflects this change.
If you want to say that it is appropriate to measure the nonlinear distortion with the filter in place, I have no argument there. I have said for years I see little uses of distortion measurement made on raw drivers other than that they are informative of the raw materials going into a speaker.
The problem here is not the assertion that linear distortion is dominant but rather your continued implication that nonlinear distortion is. You seem to continue to imply that all the problems are in the nonlinear domain and the linear aspects are of little significance.
In any event, I have turned my attention to designing my own version of a ZDL system. I can assure you it will have both low nonlinear and low linear distortion.
Last edited:
I think if you tried yourself, you will find that phase is important. But I know of no other way to change configurations and test as fast as UE, so I really haven't tried. What other packages do is of no interest to me unless the basic method is described clearly enough for me to make a judgement whether I wish to try or not. Perhaps you have something particular in mind?
I have not compared or analyzed this, so I cannot comment. Klippel demonstration was just to explain what it would sound like, I don't think I'd ever select a driver like that. It's getting close to the point where I have to consider it more carefully though.
If you're referring to phase EQ, you're forgetting my early input wherein I showed the measured response of my new dipoles. This was simply to suggest an alternative EQ method and tool to Joachim back then. I surely had no idea it would cause someone to go viral.
In addition, I used the UE built into SE v17 back in the fall of last year in my primary 3-way passive system at the time, created the same crossover for Fc and slope to approximate the passive version and EQ'd the phase as well. Alternating between flat phase and minimum-phase I couldn't hear a difference. I'm not convinced at all that it's audible in a speaker system. But I always figure it's prudent to optimize as many aspects as possible. I can do it, so I do.
Of course I'm using that fully UE EQ'd dipole system now. I've tried it minimum-phase as well and have a lot more experimentation planned.
Dave
My experience between flat phase and minimum phase is that the difference is the preferred polarity I listen to each album. I also need to do more experiments to see what is optimal. One of the reasons I have not used open baffle designs is that the wave front is just to complicated, and introduces coloration to sound, also slightly smearing the focus. As learned in physics, waves just pass through each other, and there is no way in reality to perfectly combine multiple sources into a simple source wavefront. I do realize that others have different design tradeoff considerations.
Has a visitor a few days ago, we just setup a pair of speakers in our living room which had no special treatment to absorb reflections. I just flipped the polarity around and asked what he heard different. He just pointed out that one polarity seemed to be better focused, which was also my perferred polarity (I did not tell him though).
I think there's some sort of demo version of UE available one the SE web site to test sound card compatibility, not sure what it actually does.
Has a visitor a few days ago, we just setup a pair of speakers in our living room which had no special treatment to absorb reflections. I just flipped the polarity around and asked what he heard different. He just pointed out that one polarity seemed to be better focused, which was also my perferred polarity (I did not tell him though).
I think there's some sort of demo version of UE available one the SE web site to test sound card compatibility, not sure what it actually does.
Last edited:
BY this comment do you mean on the whole system or just on the individual filters, or each filter. I am not clear on that. I would agree that phase for a loudspeakers system isn't audible at all, but relative phase between drivers (which would be different for flat phase and minimum phase) has a substantial effect on directivity at the crossover. Sean Olives work implies that this is audible and my experience agrees.
I wasn't clear. I'm referring to phase of the system response. The phase is EQ'd for each driver to flat with the summed response phase for the system also being flat.
I would think that much would depend on the physical arrangement of the individual drivers with regard to directivity vis-a-vis phase. The off-axis phase rotation is going to be dependent on the change in rolloff of the drivers, whether the crossover is flat or minimum-phase. If the drivers are aligned vertically and for acoustic offset, it seems to me that the summed response at any axis should be the same, since the phase change due to rolloff of the drivers will result in the same delta between drivers regardless of the individual driver phase responses. This is going to be largely a power response issue. At least that's how it looks to me.
In my tests, the systems have all had aligned drivers with baffle diffraction control applied, so the off-axis response of the minimum-phase passive EQ was already fairly smooth.
If you start with non-aligned drivers and then optimize using phase correction for path length differences due to acoustic offset, then all bets are off. It would be interesting to see a model of both cases with the predicted off-axis responses.
Dave
System phase is certainly not audible as many studies have shown.
For vertically aligned drivers, in the horizontal plane what you say is true, the lobing is only vertical, but it can still have a significant effect on the power response. Olive has shown that the power response is audible even if the axial response is flat and smooth.
The difficulty I have in comparing these is that I do not have the capability to switch instantaneously, so differences have to be a bit more pronounced for me to be sure it's a real difference. The system can be switched from one phase EQ to the other in less than a minute, but for small differences this still makes it difficult to distinguish. Controlled studies may reveal it, but I can't say with certainly that I hear it.
Dave
Last edited:
I'm not taking sides on this but there have been studies that come down on both sides. One of the things I will say is that with a tool like the UE is is possible to go from a nonlinear phase to a linear phase system without altering the polar or power response. Thus an apples to apples comparison can be made. Additionally we put the phase linearization in the code because we could, so why not?
Here are a couple of older figures from my web site:
The top figure is the CSD of a Linkwitz/Riley 4th order crossover. The lower figure is an LR4 amplitude, linear phase crossover. The linear phase LR4 obviously has superior transient response. Audibility?
One of the things I have said for years is that we would not buy an amplifier if it had the phase response of a 2 or 3 way crossover, so why do we accept it in speakers? Granted, there are other issues with linear phase but having the ability to try it with a simple click is certainly worthwhile.
We could do the same and did the same 20 years ago with the Essex equalizer. We did a double blind test in a dark chamber. A digitally equalized speaker with LR4 and the phase linear version compared at a flip of a switch. The outcome was not statiscally significant different with music material of average complexity. With artificial signals like a dirac impulse there where small audible differences. The phase linear version sounded "tick" on that signal, the conventional LR4 sounded "tock". One listener out of maybe 20 was slightly better to identify the phase linear version. He picked the phase linear version significant more often as the one he prefered. In an open discussion afterwards he claimed that the phase linear version had better image focus. I concluded after that experiment that flattness of frequency- and power-response is much more important when the crossover is a LR4 type then the phase distortion that it introduces. Still it is "good to have" the phase linear option. Maybe some day comes a musically meaningfull signal along that needs a phase linear speaker to cary the message across.
Except that unless the inter driver spacing is much less than a wave length the LR type crossovers are not flat power response.
So the Essex Eq was FIR based?
I just want to point out that we are not just linearizing the phase of the crossover. We are bring the entire system to linear phase including the phase and GD associated with the woofer low frequency cut off. The system is acoustically linear phase at the design point from below audible frequencies to the upper limit of the UE DSP cut off.
Last edited:
It was a two way speaker with a rather low crossover point but i agree in terms of power response. So we set up the speaker in the nearfield. There was maybe 1.5m or a bit more from speaker to ear. The frequency response was equalized with an IIR filter and the phase was equalized with a FIR filter. I have published the papers about this earlier on this thread. You can find them on Prof. Hawksford own website. I just have to pick up my son from town and then i send you an ULR.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.