How is HOM measured?

[mige0]

Being interested and doing testing myself in diffraction outside of horns/waveguides, it is obvious that this is the beneficial aspect. Were this not the case, the foam would be completely ineffective. I don't see the point of the comment. One desires to filter out the HOM, i.e., reduce their level in relation to the main wave. They cannot be separated in a specific measurement, but it seems obvious to me at least that they can be treated separately in that they can be reduced to greater degree than the main wave by the simple fact that they do not move in the same path as the main wave, being diffraction effects. The simple fact that they pass through a longer path means that they are filtered differently, more highly damped, than the main wave.


http://www.diyaudio.com/forums/showthread.php?postid=1776031#post1776031


HOM can be measured. In underwater acoustics they call it "matched field processing". Under water sound is propagating in a waveguide. Measuring the acoustic field with several geophones at different depths makes it possible to sort out the vertical wavenumber spectrum of the sound. This can be done by taking a Fourier transform over the array and thereby sorting out the different modes. For a horn it may be more difficult. You need to measure the spatial variation at the mouth of the horn by moving the microphone in steps across the mouth. Then a Fourier transform of there measurements at a given frequency will give you the wavenumber spectrum, kx and the different HOM's. Not quite sure how the near field evanescent waves will influence the results though.

seh

That is quite correct. I worked in underwater sound at Penn State where the torpedo sonar heads were developed. I know HOW to do it, I just don't have the capability. And in the end we already know that we don't want them and measureing them won't change that. At B&C I showed them how to measure the HOM from the drivers using microphones along a plane wave tube. The HOM in the horn could be done the same way. Since its axisymmetric, you only need a line of point measurements and you can sort out everything.

Theoretically you can measure the far field and calculate the mouth velocities - acoustic holography - and from that you could calculate the modes. But this method is highly prone to errors and singular or near singular matrices in the inversion process - we tried this several year ago. WIth enough computer power however the matrices could be analyzed with SVD, but basically it all requires a tremendous amount of effort, which in the end doesn't change what you want to do.

http://www.diyaudio.com/forums/showthread.php?postid=1776210#post1776210





----------------

Can I conclude from the above (min / non-min phase) behaviour that I could - most easily - detect HOM doing on axis measurements at different distances and compare for different decay in CSD for example?

Michael

I don't see what distance has to do with anything.

http://www.diyaudio.com/forums/showthread.php?postid=1841943#post1841943


It seems to make sense for Earl to not g into the depth of HOM measurements for one reason or an other.

Maybe he will join at on time here for quantifying HOM

Michael

 

[mige0]

Thanks for the conformation, Earl. One other thing, agreed that at the origin the potential surface is flat. However, The thing that has me wondering is that a surface of constant velocity potential is not (necessarily) a surface of constant velocity. This is certainly the case here since the potential surfaces are elliptical. Similarly, a surface of constant velocity would not be orthogonal to the stream lines, thus the velocity vector would not be perpendicular a surface of constant velocity. The point being that it would seem (nearly) impossible to actually construct a diaphragm which had pistonic motion and had the velocity vectors correctly aligned with the flow lines. Is this the origin of at least some of the HOMs? Also, are there other sources of HOMs, for example, arising at higher frequency due to relative comparison of wave length to WG dimensions, sort of like the HOM that are present in a constant area duct went the wave length is on the order of the duct cross sectional dimensions or smaller?

John

The wave equation that I use is in terms of pressure not velocity potential, hence the two variables are pressure and pressure gradient which is velocity. I suppose that it could be done in a velocity potential form (that is more common in CFD than acoustics), but that's not the way I did it nor have I seen that done. The fact that the pressure gradient at the origin is a function of the distance from the axis is described in my past writings which shows how a uniform velocity will still generate HOMs. This function also changes with frequency and so it is highly unlikely that BOTH the velocity contour and the change in frequency could be achieved simultaneously. However, I do show how one could surpress any desired mode by velocity shaping, and one of my patents describes how to do this.

The nonorthogonal nature of the velocity to the pressure is the reason for the HOMs that are generated within the device. And all devices where the wavefront shape is changing will have these. The HOMs that are created by a mismatch of the source to the duct are another issue. Those HOMs are created by the boundary conditions at the throat.

The HOM do have a cut-in phenomina exacty like waves in a duct and they exist, just like in the duct, depending on how the source is aligned with the duct. A duct fed with a flat plane wave will not have HOMs even if it is possible for them to exist. Everything depends on how the source fits onto the duct.

This is why I don't believe Jean-Michels simulations. The throat does not look right to me and its very easy to get that wrong. Even easier if thats what you want to do. Comparing a single frequency from one device to another is also meaningless as other frequencies may be completely different.

Thanks Earl.

Is is also ture that these HOM propagate at velocities higher than the fundamental wave, as they do in a duct, so that below some cut off there can be no HOMs?

Below some cutoff the HOM are "evanescent" in any duct straight or flared. They exist and propagate but with a complex wavenumber which makes them attenuate with distance exponentially. BUT in a short device these waves COULD make it to the mouth and still be part of the radiation. Its a complex problem.

The HOM probably do in fact travel at different speeds just like the unflared duct (everything else is similar) but it would be a complex mathematical problem to calculate that speed. Maybe not so complex, just not something that I have done.

http://www.diyaudio.com/forums/showthread.php?postid=1777655#post1777655

 

[mige0]

Sounds like semantics to me, but your words are fine, its the same thing for alol practical purposes.



We know that HOM are not good, but we don't know precisely at what level they become audible or objectional. But we also know that their audibility is SPL level dependent (our AES paper). Hence the "best guess" at this point is that the lower the HOM are the louder the device will play before they are objectionable. That appears to be the hypothesis that is best supported by the data at this point. Hence it is anything but MOOT for a loudspeaker whose design intent is to achieve very high SPLs. It is MOOT for low level SPLs.

As anyone who has heard my speakers will tell you their ability to play extremely high SPLs without audible distortions is second to none. This is strong support for the hypothesis given above.

http://www.diyaudio.com/forums/showthread.php?postid=1788762#post1788762

 

[Paul W]

What if you used a sine chirp of a high frequency only one cycle long? Maybe then you could see the reflected sines coming in just after the first? Anyone know if SE can do something like this? I've never used the analog measurements section.

Hi Brandon,
I sincerely agree we should try to quantify HOMs 'cause otherwise we are groping in the dark rather than evaluating the merits of different solutions though measurement.

In early 2007, I used SE V11 to generate 1-2 cycle bursts of single frequencies in search of HOMs. I did get some "interesting" output signals, at various frequencies, but nothing I can say were HOMs. Potential problems with those "measurements" included the earshattering levels I used (possibly introducing severe distortion or outright clipping in the microphone/preamp/compression driver). Based on what Earl has said more recently, those extreme levels were probably unnecessary. I no longer have those "measurements", but Dennis H found some of my screen shots from HTguide and posted them here in, I think, Earl's waveguide thread.

I may take another stab at it in a couple of months. This time with a rectangular window CSD plot expanded to about 50db range, and with a timescale including any "glitches" in the impulse response. To achieve a very quiet environment, outdoors, baffle flush with the ground surface...driver and horn in a "pit" with mic suspended above. Measurements will step across the WG radius, in small increments, looking for late arriving signals from the impulse.

Might work, might not
Paul

 

[ZilchLab]

What the foam does, in a gross sense, at least, is visible by clicking between Pages 33 and 34 here:

http://www.gedlee.com/downloads/Sound Quality Improvements in Compression Driver Systems.pdf

There may be clues in the behavior; clearly, it is frequency selective. What else do we see there?

 

[soongsc]

Hi Brandon,
I sincerely agree we should try to quantify HOMs 'cause otherwise we are groping in the dark rather than evaluating the merits of different solutions though measurement.

In early 2007, I used SE V11 to generate 1-2 cycle bursts of single frequencies in search of HOMs. I did get some "interesting" output signals, at various frequencies, but nothing I can say were HOMs. Potential problems with those "measurements" included the earshattering levels I used (possibly introducing severe distortion or outright clipping in the microphone/preamp/compression driver). Based on what Earl has said more recently, those extreme levels were probably unnecessary. I no longer have those "measurements", but Dennis H found some of my screen shots from HTguide and posted them here in, I think, Earl's waveguide thread.

I may take another stab at it in a couple of months. This time with a rectangular window CSD plot expanded to about 50db range, and with a timescale including any "glitches" in the impulse response. To achieve a very quiet environment, outdoors, baffle flush with the ground surface...driver and horn in a "pit" with mic suspended above. Measurements will step across the WG radius, in small increments, looking for late arriving signals from the impulse.

Might work, might not
Paul

I would probably be interesting to just use the scope function to look at the sine waves. This also provides easy indication whether anything is clipping. I have looked at some woofers this way. I suspect that in the wave guide it would be easy to saturate either the mic or the preamp, so it's convenient to use a preamp with switchable gain like used in the Wallin preamp.

 

[catapult]

I've been thinking about the foam thing. I don't really think it's fixing HOMs (reflections between the horn walls) in a Geddes waveguide. The reason is once you get out an inch or so in the throat, the walls of his circular waveguides are at nearly 90 degrees and a reflection from one wall can't hit the other wall. The path length of an HOM is barely longer than the direct path and the foam attenuation would be about the same. More likely, the foam is improving the sound because reflections from the mouth to the throat and back to the mouth do have to pass through a lot more foam.

Now it would be a different story with something like a 90x50 waveguide or horn. In that case, you could get reflections back and forth between the walls on the 50 degree axis (HOMs) and the foam would help absorb them. That's an argument why a 90x90 (or some number bigger than 90) might be better than a 90x50 without the foam. But the foam changes things and allows the 90x50 to sound about as good as the 90x90 if both have the foam.

 

[Pano]

More likely, the foam is improving the sound because reflections from the mouth to the throat and back to the mouth do have to pass through a lot more foam.

That's exactly what I've been thinking too. Strange minds think alike!

But if it works for HOMs, then it should work for reflections, and visa-versa. As you say, the shape of the horn throat should make a big difference here. As would the horn mouth.

 

[Paul W]

I have an 18Sound XT1464 (60x50) here for testing so it may make HOMs easier to identify.

 

[mige0]

Anybody here that could describe the "sound" of HOM ?
I haven't found this yet.

Thanks for the PowerPoint in PDF. Much is said about HOM but nothing about its characteristic / sonic pattern

Also I'm confused about if its non linear or linear distortion as there are different statemets from Earl on this (or as least as I interpret it).

By the way the soundfield plots at page 35 / 36 are an excellent match to Bjorm Kolbreks simulation of Earls OS

http://www.diyaudio.com/forums/showthread.php?postid=1791730#post1791730

Thanks
Michael

 

[jzagaja]

What about "Round The Horn" by Philip Newell and Keith Holland, (Speaker Builder, 8/94). Cepstral analysis was used to compare honky and non-honky horns with Quad ESL57, some of the horns sounded like electrostatic panels (short axisymmetric).

 

[KSTR]

It went it the wrong thread, but people wanting to measure HOMs might want to take note of this:
http://www.diyaudio.com/forums/showthread.php?postid=1843932#post1843932
-Klaus

 

[JoshK]

I've been thinking about the foam thing. I don't really think it's fixing HOMs (reflections between the horn walls) in a Geddes waveguide. The reason is once you get out an inch or so in the throat, the walls of his circular waveguides are at nearly 90 degrees and a reflection from one wall can't hit the other wall. The path length of an HOM is barely longer than the direct path and the foam attenuation would be about the same. More likely, the foam is improving the sound because reflections from the mouth to the throat and back to the mouth do have to pass through a lot more foam.

But the throat starts out at 12º and then rounds to 90º tangential. The transition curve between these two angle causes some diffraction/reflection doesn't it? Isn't that HOM?

 

[KSTR]

I've been thinking about the foam thing. I don't really think it's fixing HOMs (reflections between the horn walls) in a Geddes waveguide. The reason is once you get out an inch or so in the throat, the walls of his circular waveguides are at nearly 90 degrees and a reflection from one wall can't hit the other wall.

That's not the aspect of HOM (resulting from diffraction due to non-smooth changes in acoustic impedance) we are dicussing here, as I understand it. Your example seems to be an additional effect, which may or may not occur depending on geometry. With a non planar wavefront feeding a plane wave tube this effect is well known (see ie. the Celestion AES-Paper on their driver design).

Patrick's drawing has visualized the effect of the foam pretty well, the HOMs travel back to the throat or any other kink in the profile and get reflected there: (at least) 3 times through the foam the sound must travel, instead on only once, that's why there is attenuation. Earl has written about this very clearly and I think it is pretty intuitive also.

- Klaus

 

[doug20]

HOMs are probably part of the problem. But I don't think the whole horn honk thing. The mouth termination makes a big difference - I know - I've done tests. IIRC, Geddes talks about HOM mostly in the horn thoat. I'm sure both contribute. How each contributes should be dependant on the design of the horn.

So if we solve the HOM issue we can still have a "Honk" in our horns?

 

[EarlK]

Measuring HOMs ?

To Search for HOMs ?

- Seems to me that FR waves going sideways for a while before finally emerging ( from a horn bell ) will be delayed to the axial wave . This is measurable .

- I'd first look at the Group Delays of swept sines ( all of which ARTA executes quite nicely ) on horns ( or WGs ) with & without the necessary internal foam ( HOM sponges ) / at levels one would normally use / as well as some exaggerated levels ( to really get the "bounce-effect" going ) .

- Group Delay anamolies ( above & below the median value ) just about always correlate to some measurable FR irregularity ( when compared to the same driver on a Plane Wave tube ) .

- Of course, determining what contribution to FR is from the horn and what is from the driver, is itself a hairpulling experience .

- Response irregularities that grow more ragged as drive levels are increased ( excluding all other possible factors ) would be my guess as to what should develop.

- Unfortunately in my enviroment, it's the controlling of all the mitigating factors that would keep me busy. Until they're minimized I wouldn't offer a conclusion of any consequence .

- Just a couple of thoughts on the matter .

>< cheers

 

[catapult]

But the throat starts out at 12º and then rounds to 90º tangential. The transition curve between these two angle causes some diffraction/reflection doesn't it? Isn't that HOM?

Right but, with a 1" throat, it's nearly at 90 degrees by the time it moves out 1". Once it gets out there, it can't reflect any more. So reflections might travel through 2" more foam than the direct sound as they bounce around in the throat. 8" of foam for the direct, 10" of foam for the HOM -- not a heckuva lot of difference.

The mouth-throat reflection, diffraction or whatever you want to call it has to travel from throat-mouth-throat-mouth. 24" of foam for the reflection, 8" for the direct. Much bigger difference.

 

[ZilchLab]

Look here:

http://www.diyaudio.com/forums/showthread.php?postid=1843578#post1843578

What's happening at 2 kHz?

And at 15 kHz?

How come?

 

[catapult]

Look here:

http://www.diyaudio.com/forums/showthread.php?postid=1843578#post1843578

What's happening at 2 kHz?

And at 15 kHz?

How come?

The foam is clearly acting as a lowpass filter absorbing at bit more at 15K than at 2K so you'd need to account for that in the crossover. To see it, I pasted one over the other as a transparent selection in a paint program and it was fairly obvious as long as I kept the top one moving a bit. When I stopped moving it around it just collapsed into a meaningless jumble of lines. I didn't see any particular effect on the ripples with or without the foam.

 

[ZilchLab]

The foam is clearly acting as a lowpass filter absorbing at bit more at 15K than at 2K so you'd need to account for that in the crossover. To see it, I pasted one over the other as a transparent selection in a paint program and it was fairly obvious as long as I kept the top one moving a bit. When I stopped moving it around it just collapsed into a meaningless jumble of lines. I didn't see any particular effect on the ripples with or without the foam.

Yes. Nothing is happening at 2 kHz, all of the curves are attenuated ~3 - 5 dB at 15 kHz, and it is linear. There's a bit more rolloff in the last octave far off-axis, but other than that, I don't see any changes in the smoothness of the response.

To see it, I downloaded the PDF, and then one click of the scroll wheel between pages 33 and 34 flips them instantly. You have to watch it a while to see the respective curves, as they're not in color, just greyscale, and he apparently did not use the same colors for the two measurements. Use 4 - 5 kHz as a visual reference for any individual curve. The black one at -5 dB is the same shade on both slides. All of the others are doing essentially the same dance.

Are we seeing HOM removal, or merely the foam behaving as a lowpass...?