Some DSPs have limits for linear-phase crossover. For example designer has to use the same type and order for LP and HP though corner frequencies are adjustable independently. Also tap limit could prevent "optimal" result. It has happened at least once that I have not succeeded to create better sound with FIR compared to IIR. IIR was not too lame and weak, but had better sound balance and directivity features due to better tools for driver matching.
It's also possible that wide-band transients don't sound remarkably strong with minimum phase response. This has happened once. Side woofer concept is one possible reason. But usually simple 2-way is able to reproduce piano hammer properly no matter XO order while especially 3- and 4-ways with steep IIR slopes have never succeeded.
Latency requirement is one possible limit. For example Genelec has nowadays two latency options with better or worse timing. Not sure but I've probably listened the last one (before GLM was updated including both options). The first one looks typical achievement in DIY scene.
Last edited:
There is at least four Directiva threads, though not so much discussion anymore. Maybe initial enthusiasm is already gone.
I am still confused that time in music reproduction is so subject to controverse. Griesinger pointes out that wavefront detection is within 5-7 usec, yet as sort of echoing perhaps most talk about 1 msec or more.
My personal experience is of clarity in the soundstage, coherence in human voices, when attempting to improve timing. Preferably not steep filtering.
If one thinks of hearing as a sensory system required for us (as species) to survive, the usec finding of griesinger make sense.
Also JJ Johnston points into that direction, as does Manger.
Our hearing processing of pressure variations up to results in our concienceness in my view looks like the ERG (existence, relationship, growth) concept. Going up the stack takjes more processing time of our brains. But you cannot bypass a layer in the stack.
Having stated this, the "circle of confusion" is still truh. In other words "garbage in is garbage out" , a perfect reproduced garbage is still garbage.
So i periodically check if something new research is performer. ASR is not a meaningfull resource on this matter in my view.
^^^ yeah practical issues, something we have to live with. I should have phrased "anything we want" instead of "perfect" and with "ideal situation" like computer with DSP software running, perhaps same capabilities for crossovers as VituixCAD has 
Have you done thought experiments like so that anything is possible crossover wise? What would the construct consist of in this case, still a two way? What would the graphs look like, magnitude and phase / impulse? I mean is the ideal sound that you personally know and like and strive for characterizable in the graphs ( VituixCAD ) accurately or is something still missing, or is it loose concept? Is there possibility to reason backwards from some idealized (generated) graphs what kind of real system would actually measure like that and then sound as imagined? I mean its you and perhaps some handful of others who have enough experience on speaker systems that can perhaps correlate the graphs seen on screen into perceived sound in room and perhaps even backwards, imagine ideal graphs and then imagine possible systems that could produce such measurements?
There is naturally huge number of variables that can make speakers sound better or worse but it would be cool to get into bottom of it though, towards what to strive for. It is hard to know what that is when there is not enough experience, when best sound ever heard is not best in absolute terms and then trying to somehow correlate the graphs to that The CTA2034 graphs and numbers can be part of guidance while designing for good sounding speakers but this time domain aspect is floating in air for now, bit confusing what to target for, at least from my hobbyist perspective.
I've been using imagination to the timing issue from diffraction perspective, could be something to do with side firing woofers as well: When two physically different size sources share same baffle and there is crossover between them, they both have physically different acoustic environment (the outside structure of the speaker) and thus different diffraction "fingerprint" applied upon. At crossover these both play together perhaps making part of the problem. I mean how would one time align tweeter and woofer when baffle edge diffraction for both is coming at different times even if the direct sound was exactly right for both, phases aligned? Perhaps not biggie but still a phenomenon that might affect, so just minimize diffraction and then the time alignment becomes much better, reduces to that of the drivers. For example the side firing woofers you mentioned have first edges they counter and diffract at back and front edge of the speaker. These will arrive at very different time to listening position at crossover frequency than what comes from perhaps much narrower baffle woofer they are crossed over to. Perhaps three/four way speakers have problems with sound in this boring / exciting sound dilemma regardless of crossovers because of this as well, there is now three or more separate fingerprints on diffraction making too much of a mess in time domain even if the crossover had nice alignment.
I've been thinking diffraction stuff lately and wondered if diffraction really matters at all as for example mark100 / gnarly MEH has very abrupt termination at the MEH mouth which probably causes nasty diffraction and reflection, still the sound is supposedly better than ever and better than anything (its about 9th revision of MEH for him for example and vast variety of other speakers behind). This led me to think about the "fingerprint" concept, in a MEH this is the same for all the ways since it is a point source and relatively big one, all sound exists through the same physical environment, perhaps rendering diffraction at mouth non issue. Similar thing happens with any coaxial system but especially with big ones, perhaps wide baffle speaker are also liked because of this. Even though wide bandwidth of diffraction ensues it doesnt' perhaps matter as the signature is similar between all drivers as they are relatively small and relatively co-incident same place in relation to big baffle and edge far away from the transducers. Perhaps making the diffraction fingerprints more alike helps hearing system to integrate the system and sound as one.
Non-coincident driver arrangement on relatively small structure would not do this, but less ways and better diffraction mitigation treatment at the edge helps with it, reducing variance between different sources diffraction fingerprint and now even the inter driver timing doesn't have to be that good ( as you say two way speakers sound fine even with higher order filters). Would this make any sense?
Have you done thought experiments like so that anything is possible crossover wise? What would the construct consist of in this case, still a two way? What would the graphs look like, magnitude and phase / impulse? I mean is the ideal sound that you personally know and like and strive for characterizable in the graphs ( VituixCAD ) accurately or is something still missing, or is it loose concept? Is there possibility to reason backwards from some idealized (generated) graphs what kind of real system would actually measure like that and then sound as imagined?
I mean its you and perhaps some handful of others who have enough experience on speaker systems that can perhaps correlate the graphs seen on screen into perceived sound in room and perhaps even backwards, imagine ideal graphs and then imagine possible systems that could produce such measurements?There is naturally huge number of variables that can make speakers sound better or worse but it would be cool to get into bottom of it though, towards what to strive for. It is hard to know what that is when there is not enough experience, when best sound ever heard is not best in absolute terms and then trying to somehow correlate the graphs to that
The CTA2034 graphs and numbers can be part of guidance while designing for good sounding speakers but this time domain aspect is floating in air for now, bit confusing what to target for, at least from my hobbyist perspective.I've been using imagination to the timing issue from diffraction perspective, could be something to do with side firing woofers as well: When two physically different size sources share same baffle and there is crossover between them, they both have physically different acoustic environment (the outside structure of the speaker) and thus different diffraction "fingerprint" applied upon. At crossover these both play together perhaps making part of the problem. I mean how would one time align tweeter and woofer when baffle edge diffraction for both is coming at different times even if the direct sound was exactly right for both, phases aligned? Perhaps not biggie but still a phenomenon that might affect, so just minimize diffraction and then the time alignment becomes much better, reduces to that of the drivers. For example the side firing woofers you mentioned have first edges they counter and diffract at back and front edge of the speaker. These will arrive at very different time to listening position at crossover frequency than what comes from perhaps much narrower baffle woofer they are crossed over to. Perhaps three/four way speakers have problems with sound in this boring / exciting sound dilemma regardless of crossovers because of this as well, there is now three or more separate fingerprints on diffraction making too much of a mess in time domain even if the crossover had nice alignment.
I've been thinking diffraction stuff lately and wondered if diffraction really matters at all as for example mark100 / gnarly MEH has very abrupt termination at the MEH mouth which probably causes nasty diffraction and reflection, still the sound is supposedly better than ever and better than anything (its about 9th revision of MEH for him for example and vast variety of other speakers behind). This led me to think about the "fingerprint" concept, in a MEH this is the same for all the ways since it is a point source and relatively big one, all sound exists through the same physical environment, perhaps rendering diffraction at mouth non issue. Similar thing happens with any coaxial system but especially with big ones, perhaps wide baffle speaker are also liked because of this. Even though wide bandwidth of diffraction ensues it doesnt' perhaps matter as the signature is similar between all drivers as they are relatively small and relatively co-incident same place in relation to big baffle and edge far away from the transducers. Perhaps making the diffraction fingerprints more alike helps hearing system to integrate the system and sound as one.
Non-coincident driver arrangement on relatively small structure would not do this, but less ways and better diffraction mitigation treatment at the edge helps with it, reducing variance between different sources diffraction fingerprint and now even the inter driver timing doesn't have to be that good ( as you say two way speakers sound fine even with higher order filters). Would this make any sense?
Last edited:
As far as I am aware is that relative jumps in GD are far more noticeable.
Absolute GD is basically nothing more than just normal time delay.
It all depends on frequency of course.
But relative to higher frequencies you have quite a bit of it.
I do have to admit that I have to deep dive if you really want some references.
It has been awhile and ever since I just don't bother.
Mostly because GD just is a result of the design anyway. So there isn't much you can do about most of the time.
I also don't see much practicality in using even higher slopes than 8th order.
If one need such high slopes it means you're working on the edge already. Probably compromising something in the first place.
About ASR, what I find very strange and a bizarre way of discussing there, is a "new style " of discussing.
When I have a discussion about certain things, I expect that the people who are involved have a certain level of knowledge and/or experience. Or at least also take the time and effort to look things up. Of course use logic thinking and reasoning as well.
But instead people there very quickly yell "can you prove it?". Very often about just very basic and fundamental stuff. As a response they often refer to other forum posts or 3rd party news articles.
I am sorry, but that's NOT a discussion.
Let alone working out potential new concepts.
Showing that some people clearly have no clue what they are talking about. Let alone understand certain nuances or what they are saying/asking is totally strange to begin with.
This is not specific for ASR, I see this other (non audio) forums as well.
Some people (also on this forum) live under impression to be able to take shortcuts when it comes to knowledge.
Besides there is still something that's called logic thinking and deductive reasoning.
Anyway if you really want I can look up the references about GD, but it's basically just using Google as well as looking in the AES and IEEE library (and other organizations).
Last edited:
Not only that but the majority seems to live under the impression that engineers from commercial products are some kind of super humans. This also counts for some more known members there.
Totally not open for any kind of critique.
(Which for me is extra awkward since I worked for a couple of those speaker companies)
Which to me is a totally alien approach.
I am sceptical in nature even more so when commercial products (money) are involved.
It's sad, because I have seen many DIY projects that exceed most high-end products with ease.
(Compliments for those people btw!!).
The general amount of knowledge also seems to be much higher.
Yet many people even refuse that one can build such things, or even understand that one doesn't need a very expensive Klippel system (buzz word of the last two years)
Last edited:
About zero phase, this article came out a while back;
https://audioxpress.com/article/zero-phase-in-studio-monitors
Although the only reference is Flooyd Toole's book, it is in line with other research as well.
And also in line with my own experiences.
Another point about GD, is that it is always connected with the system alignment aka frequency response.
Something that seems to be totally omitted in Toole's book.
Here in DIY land most people seem to use proper alignments with Q=0.707 at most.
But if you have ever measured on some commercial products, some seem to have enormous peaks around Fb.
Something that will be audible for sure. With that also comes a nasty GD response.
Member
Joined 2003
Part of the struggle is isolation of the variables. It can be difficult to isolate "audibility of group delay" when changing from 2nd to 4th order slope for example, as group delay is not the only aspect affected by this change, of course amplitude response, power response, polar pattern, distortion, may all change as well. I think a lot of people have came to incorrect conclusions based on coincidental differences rather than hard evidence. An example would be "I heard this tweeter in so-and-so's speaker and I didn't like it, therefore this tweeter is not a good one". Of course this is jumping to conclusions but it happens all the time.
As far as research goes, it may have not been the goal to determine "why" something is audible, the question may simply be "if" something is audible, so the goal of the research is simply to determine priorities of various aspects of acoustic design and where efforts should be focused. In general, audio or not, research has often been biased or carried out with an agenda of proving what you already believe to be true, in which case the research can be skewed heavily in the direction that validates the pre-conceived ideas.
The problem I see at ASR is that many users are consumers without vast knowledge of acoustic design or electric theory etc., so they have found someone who appears to know what they're talking about and then proclaim "what that guy said" rather than complete their own research and form opinions of their own. Hence my "yes-men" comment above, as there may be a few select users carrying the knowledge of the group, with a handful of followers in blind agreement. It can make it rather difficult to have intelligent discussion with an open mind.
That happens 99.8% of the time unfortunately. Don't see why that would be jumping in conclusion, it's the reality.
People who actually did a proper double blind test with an open mind, I can count them literally on one hand.
What I find most ironic, is that many of those "conclusions" are totally not in line what one would expect from the science and theory. Therefor, many "conclude" that the science is incorrect.
Go figure.....
btw, the "why" can be a VERY important factor for answering the "if".
When research is done one a proper way, it's actually expected to answer those questions.
The biggest reason behind it, is because one has to be very aware for false positives.
Meaning that sometimes things can be "measured" but are actually wrong or influenced by other variables or things like "noise" (meaning of noise here is broad, just anything that can influence the results on a negative way).
When the "why" is not being answered, this can be very easily be overlooked and wrong conclusions can be drawn.
Last edited:
Member
Joined 2003
You can’t be serious. It is making assumptions without evidence , you have jumped to the conclusion that the tweeter is no good without considering any other aspect of the overall speaker design…what evidence do you have that the aspect you didn’t like about the speaker was the driver itself? the assumption being made is that good drivers can’t be implemented poorly. With that conclusion, all I should need is good drivers and a textbook butterworth calculator
If there is disagreement between the data and reality, the conclusion is that the data is bad or insufficient to reach a proper conclusion. If someone claims it measures good but sounds bad, it falls under poor interpretation of data, bad data or insufficient data. Over the years we have transitioned from single axis to full spatial acoustic design, I wonder why …
Book is superficial due to very large content, and already partly outdated in this matter. In addition, researchers are talking about audibility only which indicates that they don't know all methods to sense quality of timing. Using headphones for testing indicates the same. So I don't trust all claims and conclusions by scientists.
Indirectly yes, if we are talking about single minimum phase driver. I've been talking about multi-way with IIR crossover which disconnects group delay from individual minimum phase frequency responses. Quite badly in practice. For example symmetrical 8th order slopes could be "suicidal" for the sound. That compromise is for pro gear to maximize power handling and minimize excursion and lobing, but otherwise unnecessary and quite stupid in my opinion.
No, you're missing the point here.You can’t be serious. It is making assumptions without evidence , you have jumped to the conclusion that the tweeter is no good without considering any other aspect of the overall speaker design…what evidence do you have that the aspect you didn’t like about the speaker was the driver itself? the assumption being made is that good drivers can’t be implemented poorly. With that conclusion, all I should need is good drivers and a textbook butterworth calculator
If there is disagreement between the data and reality, the conclusion is that the data is bad or insufficient to reach a proper conclusion. If someone claims it measures good but sounds bad, it falls under poor interpretation of data, bad data or insufficient data. Over the years we have transitioned from single axis to full spatial acoustic design, I wonder why …
We can draw conclusions, and those conclusions are that we just don't know because the experiment that was being done wasn't adequate to prove anything. Meaning that the conclusions drawn from those experiments are false and shall not be used.
It's not like some kind of nor statement, that if one isn't true that it automatically means it is true.
That's not how science works.
How science does work, is that one can certainly expect certain things from logic and deductive reasoning.
Of course that is not as strong of and evidence and practical experiments certainly still have to be done to draw final conclusions.
But sometimes we can already tell how likely something will be or not.
Last edited:
Member
Joined 2003
I'm not missing the point, you've missed mine. Your argument above was the point I was making. We've come full circle, moving on.
Last edited:
In pro stuff I have never used any 8th order crossovers.
Usually just 4th order and if needed just a notch or param EQ when things are really bad.
This used to be all done by active analog filters back in the day, these days DSP obviously.
(although the system delay can sometimes be very tricky with live performances)
As for HP filters, and higher order actually gives more cone excursion around Fs.
Which often leads to more issues than a shorter peak.
This can easily be simulated as well.
So in practice a 2nd order filter is more than adequate enough.
When there is plenty of DSP resources, these days I use a LT-transform with HP filter for the high-side
(= 4th order in total, can be made 3rd order as well)
Just because it's convenient and reliable workflow.
I don't follow your idea of that Toole's book is superficial or outdated?
As for any other research, it's just only valid in its own context.
In this case the scope is mostly about constant directivity as well as room modes, which will be still true 30 years from now.
The rest I just see as filler stuff, to make the story complete.
But it's also simplifying and skipping over certain aspects to much, that's what I do agree on yes.
For example, one can easily draw conclusions that "therefor" other things don't matter.
Which can be true to some extend, but the nuance in there is extremely important.
Certain nuances are also more just "different", instead of having a negative impact on the listening experience.
Since our ears (read: brains) are very good in adapting and masking, that difference very quickly becomes non-detectable.
It's often a good thing to double check and acknowledge
Although to maybe explain a bit more about that last part in context of tweeters (or speakers).
If the frequency response as well as all other non-linearities (distortion, phase what have you) are the same, that means that from an acoustical point of view, it's literally the same thing.
Meaning it is extremely likely it will sound the same, unless there is something we can't measure yet (which I HIGHLY doubt)
Just to show that you don't need practical experiments to draw certain valid conclusions.
Last edited:
Yes I understand, but since it was not the scope of that whole research project to begin with (which I was explaining).
So I am confused why you mention the two together?
What we already can conclude is that if those timing issues are in the order of magnitude or below those phase shift issues, they won't be noticeable. (otherwise those phase shift would be audible as well)
But in the end, GD issues are things that are easy to do in a double blind experiment with headphones and some active EQ'ing on a computer. If they won't be noticeable there, they won't be noticeable with speakers, since the masking effect as well as reflections, room modes and background noise are much higher.
With timing (delay) I automatically think about room modes to begin with, since they will have a very dominant factor at the frequencies we are talking about (150-200Hz and below).
Which makes it automatically impossible to compare GD of two speakers in different rooms.
Testing in the same room and "if" there would be a difference, you might as well listen to room modes that are slightly shifting.
I do not agree. Using headphones is a mistake because all detectors are not in use. Ears are not very good for sensing pressure differences. They are very smart with timbre/tone but timing error does not change much that - subjectively.
Yeah I think I know what you mean, tactile feel and it is not the chest thumbing bass like in a club but for example plucked acoustic guitar or almost anything can have the feel and in relatively low listening levels as well. Its not just a sound but a sensation. This will not happen with headphones.
Exactly. It's really amazing how low volume is needed to sense wide-band transients if timing is not destroyed with high excess group delay at LF. Of course magnitude response has to be okay and music should contain those (unlimited) transients. For example all piano recordings don't have normal/strong hammer hit, and some - usually played with electric piano have very strong because there is no reflections. Also these differences are very easy to sense with low volume.