I had similar experience with my speakers (also active with Hypex FA253s) about LR2 vs LR4 at 300Hz. Maybe that 12" are not that good at higher frequencies than your 8"s.
It could also be related to group delay. A 300 Hz LR4 filter introduces a lot of group delay in the 100 - 400 Hz region. LR2 has much less.
My conclusion from a year ago was that my 12" woofer cabinets don't really like to go above 250 Hz... I was not sure if it was a woofer limitation or a cabinet limitation. I also thought that the MW16TX driver had an outstanding midrange as long as it was not asked to go below 150 Hz. So all in all, 200 Hz LR4 worked best.
Once again I am reminded that it is impossible to draw general conclusions about loudspeakers based on single event (anectdotal) evidence.
My conclusion from a year ago was that my 12" woofer cabinets don't really like to go above 250 Hz... I was not sure if it was a woofer limitation or a cabinet limitation. I also thought that the MW16TX driver had an outstanding midrange as long as it was not asked to go below 150 Hz. So all in all, 200 Hz LR4 worked best.
Once again I am reminded that it is impossible to draw general conclusions about loudspeakers based on single event (anectdotal) evidence.
^^^ Agreed.
The difficulty with drawing conclusions is making accurate measurements in 360 degrees in the all important 100-400Hz region is very difficult. Measuring In an anechoic environment is one thing but then in the room, or a particular room, changes all that.
I think this is one area where the ears might have to be the final arbiter... even thought I want to be able to reconcile what I'm hearing with what I'm measuring.
There must be some software tool that can split an audio track with IIR or FIR filters, using various filters eg. LR2 or LR4, and then re-combine them, so we can listen over headphones.
The difficulty with drawing conclusions is making accurate measurements in 360 degrees in the all important 100-400Hz region is very difficult. Measuring In an anechoic environment is one thing but then in the room, or a particular room, changes all that.
I think this is one area where the ears might have to be the final arbiter... even thought I want to be able to reconcile what I'm hearing with what I'm measuring.
There must be some software tool that can split an audio track with IIR or FIR filters, using various filters eg. LR2 or LR4, and then re-combine them, so we can listen over headphones.
here you go,
not big selection but something at least
https://audioexperiment.com/pages/crossover-filters-on-headphones/
not big selection but something at least
https://audioexperiment.com/pages/crossover-filters-on-headphones/
Hmmm interesting, but should be user selectable tracks. Songs I've listened to a hundreds of times on my headphones eg. New Favorite by Alison Kraus and Union Station or Rather Be by Clean Bandit, on HD600 headphones, But songs I know (to the point of boredom) are not the songs you may know, or use to audition sound systems.
Maybe I'll look for something that one could use his/her own familiar music to test. Some plugin that's usable on a plugin for the ancient Foobar or an up to date free DAW like REAPER.
Maybe I'll look for something that one could use his/her own familiar music to test. Some plugin that's usable on a plugin for the ancient Foobar or an up to date free DAW like REAPER.
Last edited:
I agree with acoustic LR2 on the MF/HF crossover. The polars are much better for sure.
I'm not sure about the LF/MF. The wavelengths might to large to make a difference. Have you got any graphs to illustrate the difference Juha?
Jack Oclee-Brown / KEF recently said that linear phase FIR filters seems to be audibly different to IIR filters in the LF/MF crossovers, whereas in the MF/HF he though it was a nice feature to have... but wasn't sure there was a benefit. I should probably find the reference, instead of typing stuff verbatim...
I'm not sure about the LF/MF. The wavelengths might to large to make a difference. Have you got any graphs to illustrate the difference Juha?
Jack Oclee-Brown / KEF recently said that linear phase FIR filters seems to be audibly different to IIR filters in the LF/MF crossovers, whereas in the MF/HF he though it was a nice feature to have... but wasn't sure there was a benefit. I should probably find the reference, instead of typing stuff verbatim...
Well, that depends on the own polar responses of the drivers, baffle dimensions/shape, drivers placement on the baffle and so on... This is especially true for the typical MF-HF crossing frequencies.
That's what hifijim exploites here with his unusual project with the strange looking MF-HF placement.
Last edited:
What I'm trying to get it that it might be more related to group delay, like Jim or Juha considered.
It remined me of when Jack @ KEF talked about the phase corrected LF/MF crossover they did for the LS60 wireless, removing the group delay.
Jump in from about 1 hour 10 mins for context...
but @1 hour 13 mins 30 secs:
"LS60 Wireless.. same experience it (phase correction) seems to be much more important. I believe that's because the starting point is worse. So on a 3 way speaker you have GD irregularity because of in 2 crossover points. One of them is much lower frequency which means the amount of GD in milliseconds is a lot more, so the one between the LF and MF, and so correcting it, you know, seems to be much more important thing in terms of audibility, than it was on a 2 way, with relative small amount of milliseconds of correction...
Maybe we're on a different page or paragraph...
It remined me of when Jack @ KEF talked about the phase corrected LF/MF crossover they did for the LS60 wireless, removing the group delay.
Jump in from about 1 hour 10 mins for context...
but @1 hour 13 mins 30 secs:
"LS60 Wireless.. same experience it (phase correction) seems to be much more important. I believe that's because the starting point is worse. So on a 3 way speaker you have GD irregularity because of in 2 crossover points. One of them is much lower frequency which means the amount of GD in milliseconds is a lot more, so the one between the LF and MF, and so correcting it, you know, seems to be much more important thing in terms of audibility, than it was on a 2 way, with relative small amount of milliseconds of correction...
Maybe we're on a different page or paragraph...
Last edited:
Kimmosto is talking about this (group delay) a lot and I get impression this is one of main reason(s) he seems to prefer 2 way speakers. Rule of thumb from his posts to keep excess group delay below 1ms at least to 100Hz or below. Audibility treshold of group delay is quite clear with multiple studies but its again a bit hard to isolate I think, with all other aspects that change with change in crossovers. Kimmosto describes group delay affecting initial pressure hit of transients and its magnitude, group delay makes it smaller in a way, less sensible. Its sensed more with skin and eyes for example than with ears and I think this is the reason he often says that tests on group delay with headphones are somewhat flawed as the critical sensation of tangibility does not come through as well as with loudspeakers. All this is moot with FIR processing as the group delay can be manipulated at will and is no issue.
Here is one such study where audibility thresholds of group delay are plotted https://www.researchgate.net/publication/352265058_Audibility_of_Group-Delay_Equalization/download
Here is hidden test https://audioexperiment.com/pages/group-delay-test-on-headphones/ which is work in progress as it was made copy pasting from the crossover test linked earlier and might contain errors. If you like you can try it though. You can inspect the group delay in VituixCAD, add active allpass blocks change its frequency and order to inspect what the group delay should be. The listening test uses web audio api biquad blocks and I've just copypasted biquads from VituixCAD to it.
For example two 125Hz 2nd order all pass filters make ~8ms at 100Hz. With only two cascaded allpasses its possible to make group delay that is barely audibility threshold and I should add few more stages so that the effect becomes clearly audible. But, if you don't hear any difference (assuming there are no errors how the listening test is implemented) then do not be affraid any sensible crossover won't make such crazy group delays and should not be audible.
Group delay is of course affected by diffraction for example, resonances, or bass alignment, so those could affect audibility irrespective of crossovers used I think.
Here is one such study where audibility thresholds of group delay are plotted https://www.researchgate.net/publication/352265058_Audibility_of_Group-Delay_Equalization/download
Here is hidden test https://audioexperiment.com/pages/group-delay-test-on-headphones/ which is work in progress as it was made copy pasting from the crossover test linked earlier and might contain errors. If you like you can try it though. You can inspect the group delay in VituixCAD, add active allpass blocks change its frequency and order to inspect what the group delay should be. The listening test uses web audio api biquad blocks and I've just copypasted biquads from VituixCAD to it.
For example two 125Hz 2nd order all pass filters make ~8ms at 100Hz. With only two cascaded allpasses its possible to make group delay that is barely audibility threshold and I should add few more stages so that the effect becomes clearly audible. But, if you don't hear any difference (assuming there are no errors how the listening test is implemented) then do not be affraid any sensible crossover won't make such crazy group delays and should not be audible.
Group delay is of course affected by diffraction for example, resonances, or bass alignment, so those could affect audibility irrespective of crossovers used I think.
Last edited:
Geddes says group delay is more audible with higher SPL levels. See this https://www.diyaudio.com/community/threads/geddes-on-waveguides.103872/post-2783843 and this https://www.diyaudio.com/community/threads/geddes-on-waveguides.103872/post-2783896 and if you read on he thinks crossovers are always a problem but unavoidable one and main concern regarding audibility of group delay is not in crossovers but diffraction and HOM for example. From his posts we can reason that problem of crossovers is not group delay but something else, like dip in power.
My opinion is that use active system with FIR if in doubt to eliminate group delay due to crossovers altogether. I've been able to get better sound just with IIR filters, cannot say its due to group delay (excess group delay is ~1ms around 100Hz in a three way system so kind of fullfills Kimmosto rule of thumb) or something else as every graph in VituixCAD six pack change if I try to affect group delay (with crossovers). What makes better sound is most probably combination of multiple things, system design. Optimize group delay, but don't sacrifice other aspects, perhaps more important ones, while doing so.
My opinion is that use active system with FIR if in doubt to eliminate group delay due to crossovers altogether. I've been able to get better sound just with IIR filters, cannot say its due to group delay (excess group delay is ~1ms around 100Hz in a three way system so kind of fullfills Kimmosto rule of thumb) or something else as every graph in VituixCAD six pack change if I try to affect group delay (with crossovers). What makes better sound is most probably combination of multiple things, system design. Optimize group delay, but don't sacrifice other aspects, perhaps more important ones, while doing so.
Last edited:
Regarding horns , last time i listned to zich a system was many years agoin Paris, an onken system. Beatifully made, but i continuesly heard a very slight whisteling sound. Jean Hiraga told me it
is a given with horns. If i understand correctly those were HOM's ?
At that time i read an article about a japanese engineer, he had designed the tunnel entry and exit of the japanese high speed train. And applied that to his hughe woorden horns. In short before the mouth he added slots in the horn walls of increasing width towards the end of the mouth. Such to create a smooth transition of horn to free air.
is a given with horns. If i understand correctly those were HOM's ?
At that time i read an article about a japanese engineer, he had designed the tunnel entry and exit of the japanese high speed train. And applied that to his hughe woorden horns. In short before the mouth he added slots in the horn walls of increasing width towards the end of the mouth. Such to create a smooth transition of horn to free air.
Horns, and waveguides, are complicated topic but yeah the profile is important including termination of the mouth. In general, deeper variety of horns more easily have resonances to them than just waveguides, which are usually much shallower as they are designed first and foremost to control coverage angle and not so much to provide acoustic loading for more SPL capability as there is plenty for home use without any. Both types of devices will benefit nice rounding at the mouth to have less diffraction, and reflection back to the throat, smooth transitions in general. Also phase plug in the driver and resulting exit wavefront matters, ideally the horn/waveguide is matched to the wavefront, be it flat or spherical or mixed soup. No wavefront is perfect so perhaps some HOM is expected, its quite hard to measure what the wavefront is so either flat or spherical is assumed, and flat can be converted to spherical. Direct radiating dome on a flat baffle is also a waveguide system of kind and subject to similar problems so its mostly about what kind of coverage angle, or directivity, benefits the application.
See ATH thread, I'm not sure any of the old stories regarding horn sound or horn honk or what ever has to be an issue anymore, its easy to make device without this kind of problems with todays software and manufacturing. Bad kind of "horn sound" is thing from the past. Similarly bad crossovers should be thing from the past, as we have VituixCAD and ideal perfect crossover can be assumed to be found to any speaker construct sooner or later. Only the speaker construct needs to be optimized, thats the real limitation of performance. In other words all we need is very good set of measurements, very fine acoustic radiation (what is radiated and to which directions) which only get better by making better construct. Also the quality what is radiated needs to be fine, use good drivers, sensible gain staging etc. And the whole concept needs to fit the application. There is no point doing perfect dipole system for example, if it cannot be positioned correctly in the room, perhaps boxed system would have sounded better after all. And most importantly listen to nice music all the time and don't mind if the sound is not perfect, and of course have fun
See ATH thread, I'm not sure any of the old stories regarding horn sound or horn honk or what ever has to be an issue anymore, its easy to make device without this kind of problems with todays software and manufacturing. Bad kind of "horn sound" is thing from the past. Similarly bad crossovers should be thing from the past, as we have VituixCAD and ideal perfect crossover can be assumed to be found to any speaker construct sooner or later. Only the speaker construct needs to be optimized, thats the real limitation of performance. In other words all we need is very good set of measurements, very fine acoustic radiation (what is radiated and to which directions) which only get better by making better construct. Also the quality what is radiated needs to be fine, use good drivers, sensible gain staging etc. And the whole concept needs to fit the application. There is no point doing perfect dipole system for example, if it cannot be positioned correctly in the room, perhaps boxed system would have sounded better after all. And most importantly listen to nice music all the time and don't mind if the sound is not perfect, and of course have fun
Last edited:
Some addendum items to complete the discussion of measurements:
In the horizontal plane, I made two sets of far field polar measurements. One set was measured on each driver’s own axis. A second set for the was measured on the system design axis, halfway between the tweeter and midrange. Of course, the system level measurement with the DSP crossover was made on the design axis. I do not know of any other way to make a system level measurement except from a single point microphone position.
Before I made vertical polar measurements, I compared my simulations (based on horizontal polar only) to measured reality (post 250). The simulation that most closely matched reality was the one which used design axis measurements of the midrange and tweeter, and on-axis measurements of the woofer. I was not surprised by this, as I found the same thing in three previous projects.
In the vertical plane, the test setup was far more cumbersome, so I chose to make all my measurements on the design axis. This allowed me to avoid repositioning the speaker on the rotating table. I made system level vertical polars, as well as mid and tweeter vertical polars. I did not do vertical polar measurements of the woofers.
My full-up Vcad simulations (starting in post 253) now use horizontal and polar responses for the mid and tweeter. The woofer responses are horizontal polar only, and both woofers were measured together.
Because of this, my full-up Vcad simulations have the driver coordinates for the midrange and tweeter as 0,0,0. The woofers have a y coordinate dimension, but the mid and tweeter are set to “0”. Why? As the mid and tweeter were rotated about a fixed axis for vertical polar measurements, their relative spacing was incorporated into the phase/timing of their response at each angle.
I point this out because I do not want to mislead or confuse anyone who might use my work as a template.
j.
In the horizontal plane, I made two sets of far field polar measurements. One set was measured on each driver’s own axis. A second set for the was measured on the system design axis, halfway between the tweeter and midrange. Of course, the system level measurement with the DSP crossover was made on the design axis. I do not know of any other way to make a system level measurement except from a single point microphone position.
Before I made vertical polar measurements, I compared my simulations (based on horizontal polar only) to measured reality (post 250). The simulation that most closely matched reality was the one which used design axis measurements of the midrange and tweeter, and on-axis measurements of the woofer. I was not surprised by this, as I found the same thing in three previous projects.
In the vertical plane, the test setup was far more cumbersome, so I chose to make all my measurements on the design axis. This allowed me to avoid repositioning the speaker on the rotating table. I made system level vertical polars, as well as mid and tweeter vertical polars. I did not do vertical polar measurements of the woofers.
My full-up Vcad simulations (starting in post 253) now use horizontal and polar responses for the mid and tweeter. The woofer responses are horizontal polar only, and both woofers were measured together.
Because of this, my full-up Vcad simulations have the driver coordinates for the midrange and tweeter as 0,0,0. The woofers have a y coordinate dimension, but the mid and tweeter are set to “0”. Why? As the mid and tweeter were rotated about a fixed axis for vertical polar measurements, their relative spacing was incorporated into the phase/timing of their response at each angle.
I point this out because I do not want to mislead or confuse anyone who might use my work as a template.
j.
I've been reading this thread with great interest and much appreciate the wealth of information and insight into thought processes, especially since I am in the process of building an active TMWW system with dome tweeter at 1050mm (seated head) height, and 125mm midrange adjacent. Our initial objectives are dissimilar, and I have not modelled the response of the system, but am now intrigued by VituixCad to guide the design especially with regards to baffle profile and diffraction.
I am curious about the audibility of the vertical lobing that we see on the system vertical polar plot. The on-axis is extremely flat, which I imagine sounds great in the intended listening position. You explained in post #106 that you wanted to focus this project on having the smoothest possible DI and for this reason spaced the tweeter ~210mm from the midrange (centers) based on 1.2 x 2KHz wavelength. I would have to say that you achieved near-perfect DI from 1K to 5K and well controlled elsewhere. What I'm curious about is when you stand up to fix yourself a peanut butter sandwich do you notice the change in character, or is the 2K dip too subtle to notice while you are standing up / moving around? Or is the central lobe wide enough at your listening position that you don't "see" the dip?
Cheers,
Ben
I am curious about the audibility of the vertical lobing that we see on the system vertical polar plot. The on-axis is extremely flat, which I imagine sounds great in the intended listening position. You explained in post #106 that you wanted to focus this project on having the smoothest possible DI and for this reason spaced the tweeter ~210mm from the midrange (centers) based on 1.2 x 2KHz wavelength. I would have to say that you achieved near-perfect DI from 1K to 5K and well controlled elsewhere. What I'm curious about is when you stand up to fix yourself a peanut butter sandwich do you notice the change in character, or is the 2K dip too subtle to notice while you are standing up / moving around? Or is the central lobe wide enough at your listening position that you don't "see" the dip?
Cheers,
Ben
First of all, I recommend VituixCad. It is a very well put together package of software. Easy to use, mostly intuitive. The documentation is concise, but it always seems to contain the answer to my question. There are some very helpful youtube videos.
I don't notice any audible effects of vertical lobing with the filter I am using in post 260. The character of the sound stays very stable as I move around the room. I think our ear/brain can listen through the room effects and learn to ignore the ceiling/floor reflection to some extent. In any case, the PIR (predicted in room) curve and the power curve represent the timbre of what we hear from our speakers... in theory anyway...
The only way to avoid vertical lobing is with a coaxial speaker, or a woofer-assisted full range driver perhaps... or a line array. In any case, any speaker I have ever owned has had vertical polar response that showed deep nulls at certain frequencies and certain angles. This one is no worse, and is better than some of them. The big CTC distance does not make the nulls worse, it just moves them to different frequencies and angles.
What is more problematic than a null is an off-axis peak. I have found this can be a problem with 3rd-order filters. I am hoping to experiment with 3rd order filters in the future and see if I can work around it.
I am pleased that you are finding my thread useful... Good luck with your project !
j.
I don't notice any audible effects of vertical lobing with the filter I am using in post 260. The character of the sound stays very stable as I move around the room. I think our ear/brain can listen through the room effects and learn to ignore the ceiling/floor reflection to some extent. In any case, the PIR (predicted in room) curve and the power curve represent the timbre of what we hear from our speakers... in theory anyway...
The only way to avoid vertical lobing is with a coaxial speaker, or a woofer-assisted full range driver perhaps... or a line array. In any case, any speaker I have ever owned has had vertical polar response that showed deep nulls at certain frequencies and certain angles. This one is no worse, and is better than some of them. The big CTC distance does not make the nulls worse, it just moves them to different frequencies and angles.
What is more problematic than a null is an off-axis peak. I have found this can be a problem with 3rd-order filters. I am hoping to experiment with 3rd order filters in the future and see if I can work around it.
I am pleased that you are finding my thread useful... Good luck with your project !
j.
Last week I was my local pub/tavern- what’s the North American parlance Jim? ( I was using “bottle shop” when I was over there and got told off “what’s that?! Buddy it’s called it’s a liquor store!”)
“Nah mate, it’s a bottle shop” I retorted.
Anyway there was some live music and the performer was using one of those woofer assisted line arrays? I must say I was impressed. I think they have a whole line of 2.75-3” drivers in a skinny tower attached to with a floor mounted woofer system and it managed to project quite a large/wide sound tens of meters away.
Have you seen/hear the type? I think Turbosound, Bose and variety of other brands have been doing them for awhile now.
My cousin with deep pockets bought one for his home theatre and it sounded terrible with all the reflections/reverb, but for outdoors,
At least for the single single/guitarist it’s quite good and very portable. Perhaps not a full band.
Just goes to show me that small room acoustic’s matter! 51% IMHO
“Nah mate, it’s a bottle shop” I retorted.
Anyway there was some live music and the performer was using one of those woofer assisted line arrays? I must say I was impressed. I think they have a whole line of 2.75-3” drivers in a skinny tower attached to with a floor mounted woofer system and it managed to project quite a large/wide sound tens of meters away.
Have you seen/hear the type? I think Turbosound, Bose and variety of other brands have been doing them for awhile now.
My cousin with deep pockets bought one for his home theatre and it sounded terrible with all the reflections/reverb, but for outdoors,
At least for the single single/guitarist it’s quite good and very portable. Perhaps not a full band.
Just goes to show me that small room acoustic’s matter! 51% IMHO
I heard one of those systems once. It was Bose. A tall very narrow array, on top of a woofer box. I was impressed that it sounded very clean, natural, and dispersed the sound very widely and evenly. It pretty much sounded the same in the front row as the back. This was a small cafe with about a dozen tables and maybe 40 people.
I was confused by "bottle shop" when in Melbourne the first time...
Names for bars in North America:
Pub - a bar where people go to talk, usually specializing in beer and whisky, lots of craft brews and imported beers on tap, specialty bourbons and scotches. Limited or no TV, music level is kept low, rarely have live music, but sometimes they do. Usually good food. Often decorated as an Irish or English pub, but I have been in a cowboy pub in South Dakota that was distinctly decorated in Dakota/Wyoming style. I was in an excellent pub in Miami that was 100% Cuban themed.
Sports bar - self explanatory. Food can range from excellent to terrible, beer selection can range from excellent to boring. LOTS of TVs showing LOTS of SPORTS.
Wine bar - specializing in wine, but most have very good mixed drinks as well, and a few good craft beers. Very similar to a pub in that the food is usually good, no TV, but the music is often louder. People tend to dress snappier.
Club - a trendy place where young people (or those pretending to be young) pay too much for bad liquor, bad music played very loud, sometimes excruciatingly loud. Everyone over dresses, many women wear almost nothing, some people attempt to dance. Usually smells like a blend of perfume, spilled liquor, and sweat. Later in the evening it often smells like pot and vomit.
Every city has specialty venues which focus on live music. There are usually a lot of blues bars (nice), but the occasional jazz club (yes!) can be found. They can go by any name. Well worth seeking out.
I was confused by "bottle shop" when in Melbourne the first time...
Names for bars in North America:
Pub - a bar where people go to talk, usually specializing in beer and whisky, lots of craft brews and imported beers on tap, specialty bourbons and scotches. Limited or no TV, music level is kept low, rarely have live music, but sometimes they do. Usually good food. Often decorated as an Irish or English pub, but I have been in a cowboy pub in South Dakota that was distinctly decorated in Dakota/Wyoming style. I was in an excellent pub in Miami that was 100% Cuban themed.
Sports bar - self explanatory. Food can range from excellent to terrible, beer selection can range from excellent to boring. LOTS of TVs showing LOTS of SPORTS.
Wine bar - specializing in wine, but most have very good mixed drinks as well, and a few good craft beers. Very similar to a pub in that the food is usually good, no TV, but the music is often louder. People tend to dress snappier.
Club - a trendy place where young people (or those pretending to be young) pay too much for bad liquor, bad music played very loud, sometimes excruciatingly loud. Everyone over dresses, many women wear almost nothing, some people attempt to dance. Usually smells like a blend of perfume, spilled liquor, and sweat. Later in the evening it often smells like pot and vomit.
Every city has specialty venues which focus on live music. There are usually a lot of blues bars (nice), but the occasional jazz club (yes!) can be found. They can go by any name. Well worth seeking out.