So... what *disadvantages* can waveguides have?

[andreasmaaan]

Supposed horn colouration seems to be one of the last bastions of supersition for those who otherwise take a scientific approach to audio.

I suspect the reasons for this are (a) that horn loudspeakers' functioning is relatively complex and poorly understood, and (b) that most people rarely (if ever) hear horn loudspeakers for themselves, so the myth that horns inherently colour the sound is never dispelled by practical experience.

Also - like all the most compelling audiophile myths - they have a shred of basis in fact.

However, the kinds of distortions that may be produced by horns are not mysterious bogeymen that could result in audible colouration even if difficult or impossible to detect in measurements. If your horn has an audible problem resulting from HOMs or any other cause, you will see that problem in the relevant measurements.

If a horn (or waveguide) speaker measures well, it will be - just like any speaker that measures well - an excellent sounding speaker (which you may or may not like, depending on your preferences, listening habits, and room, etc.).

 

[oivavoi]

However, the kinds of distortions that may be produced by horns are not mysterious bogeymen that could result in audible colouration even if difficult or impossible to detect in measurements. If your horn has an audible problem resulting from HOMs or any other cause, you will see that problem in the relevant measurements.

If a horn (or waveguide) speaker measures well, it will be - just like any speaker that measures well - an excellent sounding speaker (which you may or may not like, depending on your preferences, listening habits, and room, etc.).

I have absolutely no technical insight here, so I have to rely on others who have But from what I have understood from Geddes, for example, it's not completely straightforward to measure HOMs, even though it obviously can be done. Here's a post he wrote at the DIYaudio forum:
https://www.diyaudio.com/forums/multi-way/145096-hom-measured-46.html#post3791186

I understand probably less than half of what he wrote, but it seems to me like he claims that it's difficult to measure HOMs well? Could the process he describes be done on the Klippel?

In another post in the same thread he also clarifies the concept generally. As he says there, there haven't been done many investigations on the audibility of HOMs / diffraction in horns, but according to him they are always present in a horn or a waveguide, there's no getting around it.
https://www.diyaudio.com/forums/multi-way/145096-hom-measured-46.html#post3791154

As with everything in loudspeaker design it's a question of weighing different things against each other. Horns and waveguides obviously have many advantages. Some of my all-time favorite listening experiences with loudspeakers have been with horns!

---------------------------------
"Earl,
I realize you aren't going to do it but, if you set out to measure HOMs, what process/procedure would you use?

Perhaps measurement could become a community effort."

Hi Paul

Makarski notes that the directivity is not strongly affected by the HOMs. Still there has to be an effect. This is where I would first look, but you would need extremely accurate measurements of the polar response because you are looking for a needle in a haystack. None of the polar responses that I have seen are anywhere near accurate enough and even those that I do have not shown to be accurate enough either (and mine are the most accurate that I have seen.)

But lets say that accuracy of the polar response can be solved and I am sure it could by someone with the resources. Then you would need to decompose the measured polar response into its "mode" noting what was due to the main mode and what was due to the HOMs. That requires that you have a coordinate system that allows for this separation. I tried this some years back with a spherical system, but it didn't work out too well and it was difficult to get a giant sphere with a horn embedded in it. This is how I do my polar maps now so it was not a total loss and I can propagate any wave front back to a sphere and look at what is happening but the resolution is not high enough to resolve the details.
[...]
"

 

[richard12511]

I have absolutely no technical insight here, so I have to rely on others who have But from what I have understood from Geddes, for example, it's not completely straightforward to measure HOMs, even though it obviously can be done. Here's a post he wrote at the DIYaudio forum:
https://www.diyaudio.com/forums/multi-way/145096-hom-measured-46.html#post3791186

I understand probably less than half of what he wrote, but it seems to me like he claims that it's difficult to measure HOMs well? Could the process he describes be done on the Klippel?

In another post in the same thread he also clarifies the concept generally. As he says there, there haven't been done many investigations on the audibility of HOMs / diffraction in horns, but according to him they are always present in a horn or a waveguide, there's no getting around it.
https://www.diyaudio.com/forums/multi-way/145096-hom-measured-46.html#post3791154

As with everything in loudspeaker design it's a question of weighing different things against each other. Horns and waveguides obviously have many advantages. Some of my all-time favorite listening experiences with loudspeakers have been with horns!

---------------------------------
"Earl,
I realize you aren't going to do it but, if you set out to measure HOMs, what process/procedure would you use?

Perhaps measurement could become a community effort."

Hi Paul

Makarski notes that the directivity is not strongly affected by the HOMs. Still there has to be an effect. This is where I would first look, but you would need extremely accurate measurements of the polar response because you are looking for a needle in a haystack. None of the polar responses that I have seen are anywhere near accurate enough and even those that I do have not shown to be accurate enough either (and mine are the most accurate that I have seen.)

But lets say that accuracy of the polar response can be solved and I am sure it could by someone with the resources. Then you would need to decompose the measured polar response into its "mode" noting what was due to the main mode and what was due to the HOMs. That requires that you have a coordinate system that allows for this separation. I tried this some years back with a spherical system, but it didn't work out too well and it was difficult to get a giant sphere with a horn embedded in it. This is how I do my polar maps now so it was not a total loss and I can propagate any wave front back to a sphere and look at what is happening but the resolution is not high enough to resolve the details.
[...]
"

I would think that if the HOMs were difficult to measure, then they would be similarly difficult to hear.

 

[617]

I would think that if the HOMs were difficult to measure, then they would be similarly difficult to hear.

Well my memory is that Earl and his wife Lydia did psychoacoustic tests which identified them as a mechanism of distortion, more prominent as levels get higher. It's an interesting issue. I will say my Abbey's don't any horn honk but they sound strange for other reasons.

 

[andreasmaaan]

I have absolutely no technical insight here, so I have to rely on others who have But from what I have understood from Geddes, for example, it's not completely straightforward to measure HOMs, even though it obviously can be done. Here's a post he wrote at the DIYaudio forum:
https://www.diyaudio.com/forums/multi-way/145096-hom-measured-46.html#post3791186

I understand probably less than half of what he wrote, but it seems to me like he claims that it's difficult to measure HOMs well? Could the process he describes be done on the Klippel?

As @richard12511 said, if it's so difficult to measure, there is no chance of it being heard.

But a Klippel is not required for such measurements. In fact, I'm not sure exactly why Geddes believes that a standard measurement rig might not be capable. Given that he seems to suggest that a very large baffle or a spherical enclosure would be necessary, I can only infer that he believes that any measurable effects of HOMs are swamped by baffle edge diffraction under normal circumstances. If that's the case, then I think we have some idea of the audibility of the HOMs he has in mind relative to other inevitable diffraction effects present in all loudspeakers...

In another post in the same thread he also clarifies the concept generally. As he says there, there haven't been done many investigations on the audibility of HOMs / diffraction in horns, but according to him they are always present in a horn or a waveguide, there's no getting around it.
https://www.diyaudio.com/forums/multi-way/145096-hom-measured-46.html#post3791154

As with everything in loudspeaker design it's a question of weighing different things against each other. Horns and waveguides obviously have many advantages. Some of my all-time favorite listening experiences with loudspeakers have been with horns!

I agree that there's no getting around the existence of HOMs. But the same is true of every possible kind of distortion produced by a transducer or loudspeaker: none can ever be fully eliminated.

Geddes sums up the limits of his own argument when he says:

It is my opinion that only the main non-diffracted or reflected wave is desirable. All other aspects are aberrations and undesirable. How do we weight the relative audible effects of them? That is completely unknown and would be a massively complicated problem to take on. Hence my approach is and always has been to reduce all of these effects as I simply do not know which is which and how audible any of them are.

I agree that these effects are aberrations. What I disagree with is that their audibility is completely unknown. We have a very good understanding of temporal masking, such that we are able to infer at what amplitude and delay a reflection or diffraction effect may become audible.

According to Zwicker and Fastl, who have experimentally determined post-masking thresholds:

A Gaussian-shaped condensation impulse with a duration of only 20 µs produces a spectral shape that corresponds to that of white noise. It is similar to that of a Dirac impulse. Postmasking produced by a white-noise masker can therefore be measured without spectral influences using such a brief Gaussian impulse as a test sound. The peak value of this Gaussian impulse expressed in level is plotted as the ordinate in Fig. 4.22, which shows the level necessary to just reach threshold in postmasking as a function of delay time td from the end of the masker. The parameter in the figure is the overall level of the white-noise masker. The solid lines indicate results that show almost no decay for the first 5 ms after the masker is switched off. There, the values correspond to those measured in simultaneous masking. After about 5ms delay time, the threshold in postmasking decreases and, at about 200 ms delay time, reaches the threshold in quiet...

Postmasking depends on the duration of the masker. Figure 4.23 shows a typical result measured using, as a test sound in this case, a 2-kHz tone burst of 5-ms duration. Again, the time at which the test-tone burst is presented after the end of the masker is plotted as the abscissa. The level of the test-tone burst is the ordinate. For a masker duration of 200 ms, the solid curve indicates postmasking comparable to that displayed in Fig. 4.22. Quite different from that is the postmasking produced by a masker burst, which lasts only 5 ms, as indicated by the dotted line in Fig. 4.23. In this case, the decay is initially much steeper. This means that postmasking depends strongly on the duration of the masker and therefore is a highly nonlinear effect.

Without plunging into the details of all these experiments, what these results demonstrate is that the resolution of our hearing is orders of magnitude less precise than the resolution of our measurement devices. You can take measurements in your living room with a 70€ USB measurement mic that will have more than adequate resolution for these purposes.

As I argued before, if HOMs or other similar diffraction/reflection effects are audible (i.e. if they are so high in level and/or late that they rise above the post-masking threshold), they will absolutely certainly be measurable using standard measurement techniques.

 

[oivavoi]

As @richard12511 said, if it's so difficult to measure, there is no chance of it being heard.

But a Klippel is not required for such measurements. In fact, I'm not sure exactly why Geddes believes that a standard measurement rig might not be capable. Given that he seems to suggest that a very large baffle or a spherical enclosure would be necessary, I can only infer that he believes that any measurable effects of HOMs are swamped by baffle edge diffraction under normal circumstances. If that's the case, then I think we have some idea of HOMs' audibility relative to other inevitable diffraction effects present in all loudspeakers...



I agree that there's no getting around the existence of HOMs. But the same is true of every possible kind of distortion produced by a transducer or loudspeaker: none can ever be fully eliminated.

Geddes sums up the limits of his own argument when he says:



I agree that these effects are aberrations. What I disagree with is that their audibility is completely unknown. We have a very good understanding of temporal masking, such that we are able to infer at what amplitude and delay a reflection may become audible.

According to Zwicker and Fastl, who have experimentally determined post-masking thresholds:



View attachment 108321

View attachment 108322

Without plunging into the details of all these experiments, what these results demonstrate is that the resolution of our hearing is orders of magnitude less precise than the resolution of our measurement devices. You can take measurements in your living room with a 70€ USB measurement mic that will tell you if you have audible HOMs or mouth reflections etc.

As I argued before, if HOMs or other similar diffraction/reflection effects are audible (i.e. if they are so high in level and/or late that they rise above the post-masking threshold), they will absolutely certainly be measurable using standard measurement techniques.

Thanks, excellent and interesting response!

 

[Zvu]

I guess that audibility of HOM can be put to the test by simple AB testing.

Gedlee used foam plug to fight the HOM. One could make waveguide loaded loudspeaker with foam plug and without plug (corrected for the plug influence in frequency response for tweeter). If multiple AB tests show no difference/preference one over the another, that is how much HOM is detrimental to the sound.

This would require pretty precise measurements (quasi-anechoic with longer gate frequency, at least >8ms) since very small differences in level for waveguide loaded tweeter working from 1-20kHz can become audible and tip the scale to one over the other.

 

[andreasmaaan]

I guess that audibility of HOM can be put to the test by simple AB testing.

Gedlee used foam plug to fight the HOM. One could make waveguide loaded loudspeaker with foam plug and without plug (corrected for the plug influence in frequency response for tweeter). If multiple AB tests show no difference/preference one over the another, that is how much HOM is detrimental to the sound.

This would require pretty precise measurements (quasi-anechoic with longer gate frequency, at least >8ms) since very small differences in level for waveguide loaded tweeter working from 1-20kHz can become audible and tip the scale to one over the other.

You'd also have to verify that the foam plug had no effect on directivity. I don't think this would be very significant, but it may have some effect e.g. because the length of foam from throat to mouth would vary depending on angle (i.e. it may attenuate higher frequencies more the further one moves off-axis).

 

[oivavoi]

I guess that audibility of HOM can be put to the test by simple AB testing.

Gedlee used foam plug to fight the HOM. One could make waveguide loaded loudspeaker with foam plug and without plug (corrected for the plug influence in frequency response for tweeter). If multiple AB tests show no difference/preference one over the another, that is how much HOM is detrimental to the sound.

This would require pretty precise measurements (quasi-anechoic with longer gate frequency, at least >8ms) since very small differences in level for waveguide loaded tweeter working from 1-20kHz can become audible and tip the scale to one over the other.

Another possibility would be to record the outpout in an anechoic chamber of a tweeter mounted on a flat baffle, and the same tweeter mounted inside a well-measuring waveguide. Would one be able to hear the difference between these two recordings when reproduced over headphones? If not, then the waveguide does not create an audible difference beyond affecting dispersion.

 

[andreasmaaan]

Another possibility would be to record the outpout in an anechoic chamber of a tweeter mounted on a flat baffle, and the same tweeter mounted inside a well-measuring waveguide. Would one be able to hear the difference between these two recordings when reproduced over headphones? If not, then the waveguide does not create an audible difference beyond affecting dispersion.

This would be a particularly interesting experiment given that a dome tweeter would be the worse-case scenario for these issues compared with a compression driver.

 

[thewas]

Another possibility would be to record the outpout in an anechoic chamber of a tweeter mounted on a flat baffle, and the same tweeter mounted inside a well-measuring waveguide. Would one be able to hear the difference between these two recordings when reproduced over headphones? If not, then the waveguide does not create an audible difference beyond affecting dispersion.

They should be equalised though to the same recording axis frequency response though as the waveguide can alter it.

 

[ctrl]

I would think that if the HOMs were difficult to measure, then they would be similarly difficult to hear.

Well my memory is that Earl and his wife Lydia did psychoacoustic tests which identified them as a mechanism of distortion, more prominent as levels get higher. It's an interesting issue. I will say my Abbey's don't any horn honk but they sound strange for other reasons.

What does Dr. Geddes have to say about this:

How audible are they? I personally have never made any claims on this either way. I showed how HOM like signals have some very unique characteristics that meld very well with subjective aspects of horns. This is not proof, I have never claimed it was (physics can never prove something is true, it can only prove something is not true. No one has proven that what many hear as poor sound in a horn, and many many do, is NOT HOMs.)

Link

So Dr. Geddes denies ever having claimed that HOM are tonally relevant, but no one has shown him otherwise either.
This makes you wonder why so many people (also here in the forum) claim the opposite and present HOMs as the explanation for "horn sound".

If there are any studies on the audible effects of HOMs, let me know. I am not aware of any paper that provides evidence of audibility of HOMs.

But there is a paper by GedLee from 2006 about "Audibility of Linear Distortion with Variations in Sound Pressure Level and Group Delay" from which I assume the whole argument about the audibility of HOM's starts. Simply because people don't really read through the studies and only perceive what supports their belief.

There GedLee have attempted to provide psychoacoustic evidence for sound degradation by GD and linear distortion. Many probably take this study as the basis of proof that sound reflected back into the WG/Horn (or from from the side walls, small steps,...) will degrade the tonal experience due to the propagation delay and linear distortion that occurs (depending on the sound level) - but GedLee does not claim this in the study!

Can this study be used to justify the audibility of HOMs? For this, one should look at how the experiments were conducted.

(Can someone check the following assumptions to make sure I have not misunderstood anything)

So an HP was set at 2kHz and this signal was fed back to the original, thereby creating 2, 4, 6 dB shelving levels. I have recreated this in VituixCAD.
The blue curve represents the unchanged music signal and in red the HP is shown.The black curve represents the overall result - here a +4dB shelving with 0ms delay:

The decisive step is now coming. Now the added back signal was provided with a delay (0.2, 0.4,...1ms). That is, the entire high pass was delayed by a fixed amount of time. This looks like this for the variant with 4dB shelving and 0.4ms delay:

I think it is clear to everyone that this has nothing at all to do with the reality of a WG or horn, since there the reflections are frequency-dependent and also their delay. At most, one could take the results of the study as a very rough guide.


What does the study say about a shelving filter with 2dB at any delay (0.2 - 1ms)? Does this lead to a definite sound degradation?:

Under these conditions, no deterioration of the sound is detectable.

This would therefore correspond to a WG or horn with a frequency response similar to the filter used:

From here on you can stop arguing with this study, because all other cases with proven influence on the sound used a shelving of 4 or 6 dB.
All WG or horns not designed by kindergarten children, don't show such massive comb filter effects.

Except perhaps with Oblate Spheroid Waveguides there are sound pressure fluctuations/comb filter effects of 8dB on axis due to the interference from "propagation time-delayed" back reflections (GedLee Nathan):

I'm kidding
.

 

[Biblob]

https://www.diyaudio.com/forums/multi-way/338806-acoustic-horn-design-easy-ath4.html

This is such a gem of software, Marcel (Mabat) updates it and it's documentation regularly. Very recently a great update has been released that simplifys the process even further. Some people already validated the modeled response translated into a 3d-printed horn.

I often read the discussions in the thread. Although it often goes way above my head, I feel like the software itself is a big step forward in being able to design a SOTA small or big horn-speaker in all kind of shapes and sizes. The discussion often comes down to the preference of the directivity, because technically any smooth radiating horn can be made.

I encourage every speaker-designer/builder to take a look at it. (if you're interested in using a WG/horns.)

 

[tuga]

Le Cléac’h talks about High Order Modes in his Acoustical Horns and Waveguides 2010 ETF lecture from page 101 onwards:

http://www.rintelen.ch/download/JMMLC_horns_lecture_etf10.pdf

 

[oivavoi]

So Dr. Geddes denies ever having claimed that HOM are tonally relevant, but no one has shown him otherwise either.
This makes you wonder why so many people (also here in the forum) claim the opposite and present HOMs as the explanation for "horn sound".
.

I think he's being a bit coy. I'm pretty sure I've read him say on Diyaudio that HOMs can probably be audible. It's the self-protective way a scientist often has of making claims, without making them too strong so one doesn't become vulnerable for counter-attacks. In this quote he seems to say that lots of people perceive "horn sound", that HOMs is something which we know to exist, and that HOMs can be an explanation for that.

But the thing which matters is still not so much what Geddes has or hasn't said, but to what degree those things are audible. As you say, we don't know much about that.

EDIT: But the lack of audibility studies doesn't apply only to HOMs. It also applies to lots of other things on horn and waveguide loudspeakers: Do listeners have any pro-horn preference when listening blindfolded? Are horns more dynamic, as many horn lovers (including me) seem to think? Does it matter that compression drivers and horns perhaps emit soundwaves with higher velocity and "slam"? Etc. There are no audibility or preference studies on this

 

[Frank Dernie]

I would think that if the HOMs were difficult to measure, then they would be similarly difficult to hear.

Maybe, maybe not.
It could be like cabinet talk, it is there in the measurement but impossible to separate from the driver radiation.
AFAIK those that work on it predict it from FE calculation and separately measure cabinet vibration to validate the FE model.

 

[Ron Texas]

Waveguides like a lot of things in audio are a matter of design and implementation. If there's a downside, it's that the designer can screw it up. Around here the Klippel will reveal all...

 

[q3cpma]

Maybe, maybe not.
It could be like cabinet talk, it is there in the measurement but impossible to separate from the driver radiation.
AFAIK those that work on it predict it from FE calculation and separately measure cabinet vibration to validate the FE model.

Wouldn't the NFS show the source of such cabinet talk (balloon plots)?

 

[617]

What does Dr. Geddes have to say about this:

Link

So Dr. Geddes denies ever having claimed that HOM are tonally relevant, but no one has shown him otherwise either.
This makes you wonder why so many people (also here in the forum) claim the opposite and present HOMs as the explanation for "horn sound".

If there are any studies on the audible effects of HOMs, let me know. I am not aware of any paper that provides evidence of audibility of HOMs.

But there is a paper by GedLee from 2006 about "Audibility of Linear Distortion with Variations in Sound Pressure Level and Group Delay" from which I assume the whole argument about the audibility of HOM's starts. Simply because people don't really read through the studies and only perceive what supports their belief.

There GedLee have attempted to provide psychoacoustic evidence for sound degradation by GD and linear distortion. Many probably take this study as the basis of proof that sound reflected back into the WG/Horn (or from from the side walls, small steps,...) will degrade the tonal experience due to the propagation delay and linear distortion that occurs (depending on the sound level) - but GedLee does not claim this in the study!

Can this study be used to justify the audibility of HOMs? For this, one should look at how the experiments were conducted.

View attachment 108332

View attachment 108333

(Can someone check the following assumptions to make sure I have not misunderstood anything)

So an HP was set at 2kHz and this signal was fed back to the original, thereby creating 2, 4, 6 dB shelving levels. I have recreated this in VituixCAD.
The blue curve represents the unchanged music signal and in red the HP is shown.The black curve represents the overall result - here a +4dB shelving with 0ms delay:
View attachment 108334

The decisive step is now coming. Now the added back signal was provided with a delay (0.2, 0.4,...1ms). That is, the entire high pass was delayed by a fixed amount of time. This looks like this for the variant with 4dB shelving and 0.4ms delay:
View attachment 108339
I think it is clear to everyone that this has nothing at all to do with the reality of a WG or horn, since there the reflections are frequency-dependent and also their delay. At most, one could take the results of the study as a very rough guide.


What does the study say about a shelving filter with 2dB at any delay (0.2 - 1ms)? Does this lead to a definite sound degradation?:
View attachment 108342
Under these conditions, no deterioration of the sound is detectable.

This would therefore correspond to a WG or horn with a frequency response similar to the filter used:
View attachment 108344

From here on you can stop arguing with this study, because all other cases with proven influence on the sound used a shelving of 4 or 6 dB.
All WG or horns not designed by kindergarten children, don't show such massive comb filter effects.

Except perhaps with Oblate Spheroid Waveguides there are sound pressure fluctuations/comb filter effects of 8dB on axis due to the interference from "propagation time-delayed" back reflections (GedLee Nathan):
View attachment 108345
I'm kidding
.

Thanks for clarifying the parameters of the distortion study.

 

[Chromatischism]

They should be equalised though to the same recording axis frequency response though as the waveguide can alter it.

I think that EQ would put the nail in the coffin on any "coloration".