Waveguides and constant-directivity horns usually have drooping high-frequency response because of mass-rolloff. That's an issue in the compression driver, actually. But if there isn't collapsing directivity or increasing efficiency, there's nothing to counter the rolloff from mass. So equalization is usually provided in the crossover to counter that.
In some cases, that's all you need. But in many cases, there are internal resonances that cause peaks and dips in the passband. Those must be equalized out too. One of the ones I think is most responsible for "horn honk" is the peak right at the beginning of the horn's passband. Some have two or three peaks like that, and some even ripple all the way through their passbands.
By the way, even a smooth flare profile can create this. It isn't just caused by transitions. But transitions are usually the worst offenders, in my opinion. So if a flare adapter is used, be sure to match it with the compression driver exit angle and the horn flare entrance angle. You really want a smooth transition, and a profile that supports both the desired directivity and acoustic load. Those are the two main goals of the horn - Directivity and acoustic loading.
In some cases, that's all you need. But in many cases, there are internal resonances that cause peaks and dips in the passband. Those must be equalized out too. One of the ones I think is most responsible for "horn honk" is the peak right at the beginning of the horn's passband. Some have two or three peaks like that, and some even ripple all the way through their passbands.
By the way, even a smooth flare profile can create this. It isn't just caused by transitions. But transitions are usually the worst offenders, in my opinion. So if a flare adapter is used, be sure to match it with the compression driver exit angle and the horn flare entrance angle. You really want a smooth transition, and a profile that supports both the desired directivity and acoustic load. Those are the two main goals of the horn - Directivity and acoustic loading.
When I put a waveguide on my XT25 it didnt drop the high frequencies, it loads the low end. The waveguide isnt rolling off the highs they are just lower in relation to the bottom end loading.
I'm aware of some horns having issues with incorrect shape or a throat that isnt a smooth transition. This isnt what I'm commenting on. I'm aware that the low end of a waveguide is a higher spl than the high end, but that isnt a change in the tweeter's high end. It's a rise in the low end. In most cases its better to use a cap to roll off the low end to lower distortion than to boost the high end which would boost unwanted noise with it.
Seriously? You see this as different? I mean it is the lows that have gain, while the highs do not (relative to a direct radiator), but isn't it all just a matter of what you call the reference.
I tend to agree, we can describe it either way: Six one way and a half-dozen the other.
But it's kind of a flip side of the same coin, a different way to express the same thing. So with that in mind, I agree. The horn loading is greatest up to about 4kHz for most 1" compression drivers. Above that, it starts to rolloff. The rolloff is due to diaphragm mass, so measurement on a plane wave tube shows it. Most constant directivity horns and waveguides show it too. Then over and above this, some horn/waveguides have peaks and dips that have to be equalized. So all that stuff is usually dealt with in the crossover.
I also agree that attenuation in the lower octave or two will decrease distortion. If the top-end is boosted, it tends to raise distortion. But really, no matter how you slice it, if you have 10dB more at 1.5kHz than you do at 15kHz, then you still need to provide a conjugate transfer function in the crossover that equalizes this. So I guess it kind of doesn't matter if you say you're cutting the low end or boosting the high end, if the transfer function required has this kind of curve, then the distortion/noise signature will be the same. So again, we can describe it either way: Six one way or a half-dozen the other.
But in addition to the mass-rolloff issue, I think it is important to re-emphasize the matter of the peaking and ripple in some devices. It is not uncommon, many devices have it. I think this is what you're talking about, and I definitely agree with you there.
Many horn/waveguides peak right at the bottom end of their range, right where crossover is. So some of these things are even more boosted at the bottom end than just what happens from mass rolloff - They are peaking down low, more than what you would see in a plane wave tube. That really sounds nasty to me, and is the epitome of "horn honk." So that must be dealt with in one way or another, usually with additonal damping or notch filters in the crossover.
Here are a few relevant links that talk about this stuff in more detail:
But it's kind of a flip side of the same coin, a different way to express the same thing. So with that in mind, I agree. The horn loading is greatest up to about 4kHz for most 1" compression drivers. Above that, it starts to rolloff. The rolloff is due to diaphragm mass, so measurement on a plane wave tube shows it. Most constant directivity horns and waveguides show it too. Then over and above this, some horn/waveguides have peaks and dips that have to be equalized. So all that stuff is usually dealt with in the crossover.
I also agree that attenuation in the lower octave or two will decrease distortion. If the top-end is boosted, it tends to raise distortion. But really, no matter how you slice it, if you have 10dB more at 1.5kHz than you do at 15kHz, then you still need to provide a conjugate transfer function in the crossover that equalizes this. So I guess it kind of doesn't matter if you say you're cutting the low end or boosting the high end, if the transfer function required has this kind of curve, then the distortion/noise signature will be the same. So again, we can describe it either way: Six one way or a half-dozen the other.
But in addition to the mass-rolloff issue, I think it is important to re-emphasize the matter of the peaking and ripple in some devices. It is not uncommon, many devices have it. I think this is what you're talking about, and I definitely agree with you there.
Many horn/waveguides peak right at the bottom end of their range, right where crossover is. So some of these things are even more boosted at the bottom end than just what happens from mass rolloff - They are peaking down low, more than what you would see in a plane wave tube. That really sounds nasty to me, and is the epitome of "horn honk." So that must be dealt with in one way or another, usually with additonal damping or notch filters in the crossover.
Here are a few relevant links that talk about this stuff in more detail:
Last edited:
A cap lowers distortion and excursion beyond the crossover frequency. Yes its different if you're boosting background noise in a recording with EQ up top. Boosting the top end is not the same as filtering the low end. Kind of a reverse Dolby encoding if listening to FM radio. Some guys have old recordings on open reel. Not everyone is digital.
Last edited:
I tried the bwaslo's crossover on the SEOS-12. It deals with the 2kHz and 4kHz resonances. I also tried many different crossovers of my own. I use the emulator function of LspCad so it's very easy.
I got a sound a little different for each crossover, each waveguide I tried. What was constant, was the little harshness, the little pain in the ears, the little lack of delicacy. That's the reason why I suspect the DE250 to be the reason of this. I know that some skilled designers (far more than me) are using it, that many people are very happy with it, so I must be wrong (and I hope to). But I don't know where. I need to investigate more.
I got a sound a little different for each crossover, each waveguide I tried. What was constant, was the little harshness, the little pain in the ears, the little lack of delicacy. That's the reason why I suspect the DE250 to be the reason of this. I know that some skilled designers (far more than me) are using it, that many people are very happy with it, so I must be wrong (and I hope to). But I don't know where. I need to investigate more.
gtb, harshness in your context is probably just a matter of target curve: constant directivity devices throw a huge quantity of HF energy into the room when equalized flat on-axis in anechoic (gated) situation, far more than direct radiator tweeters and beaming horns.
As what you are hearing at your listening position is a mix of the direct and power responses, I would suggest that you arrange for a slow rolloff above about 6kHz, resulting in an approx 6dB drop at 20kHz at your listening position. This is probably what a "conventional" direct radiator speaker will give you when EQed for flat response at 1m on-axis (gated or anechoic)...
Look at the target cuves and X curve litterature
As what you are hearing at your listening position is a mix of the direct and power responses, I would suggest that you arrange for a slow rolloff above about 6kHz, resulting in an approx 6dB drop at 20kHz at your listening position. This is probably what a "conventional" direct radiator speaker will give you when EQed for flat response at 1m on-axis (gated or anechoic)...
Look at the target cuves and X curve litterature
Last edited:
These can be made to work. I like to cross above that point to avoid the issue, in exchange for larger horns. But then compromise is what multiway often comes down to.
When you have an acoustic problem, your crossover won't fix it and this is one of the ways that changing a crossover can give misleading results. The problems can be pushed down to make them less obvious but it won't be perfect. I'd suggest you have an EQ within reach when subjectively testing to speed up the process. It can sometimes give clues to whether an issue will actually go away.
It rained here this weekend and I decided to dust off my 'scope. Must be all this speaker building
Returned form Vacation ...
... so will respond now:
WHG: So much for the O.S.W.G. arguments in your papers, you just equalized them “all” away. (Rhetorical Assertion)
WHG: A horn is a band bass device that changes the frequency response of the drivers output. To some degree these changes may be mitigated by modifying (equalizing) the drive signal. Only on a PWT does the drivers output remain unchanged. Here is where most driver measurements take place so as not to be confused by horn artifacts. The dispersion pattern of course cannot be equalized by modifying the drive signal to a single driver, in the case of multiple drivers, that is another story.
WHG: Of course this philosophy works well in a setting where it is claimed that the distortion produced cannot be heard even though it most certainly exists and is measurable.
WHG: Statistical Inference is not a “Proof”. Some of your peers do not agree with your claims as well. For the findings to get anywhere near to a “proof”, they must be replicated by other independent researchers. My remarks here do not refute the fact that I have read and understand the content of all of your papers as well as many more by others.
WHG: Diffraction, or bending of the wave path, is accompanied by changes in frequency response. The resulting distortions may be mitigated through equalization of the drive signal. This is regularly done in a PA setting where CD horns are used.
WHG: If Be is used to make the diaphragm of a given size it will be acoustically smaller than its plastic, aluminum or titanium counterparts; and because of this, it will certainly perform differently as the onset of breakup modes and mass roll-off will occur at higher frequencies whether we choose to hear them or not.
WHG: Attendance at too many Rock Concerts, sans ear plugs, could be the answer as well. In any event the sizzle of a good Zildjian or Paiste Symbol extends to 20KHz and beyond and the differences between them are audible [1] & [2]
WHG: For those that may be interested in the spectral content of musical instruments, reference to the following article is provided.
There's life above 20 kilohertz! A survey of musical instrument spectra to 102.4 kHz
WHG: Here is an on-line listening test for those that might be interested in determining their distortion detection threshold:
Listening Test
... so will respond now:
WHG: So much for the O.S.W.G. arguments in your papers, you just equalized them “all” away. (Rhetorical Assertion)
WHG: A horn is a band bass device that changes the frequency response of the drivers output. To some degree these changes may be mitigated by modifying (equalizing) the drive signal. Only on a PWT does the drivers output remain unchanged. Here is where most driver measurements take place so as not to be confused by horn artifacts. The dispersion pattern of course cannot be equalized by modifying the drive signal to a single driver, in the case of multiple drivers, that is another story.
WHG: Of course this philosophy works well in a setting where it is claimed that the distortion produced cannot be heard even though it most certainly exists and is measurable.
WHG: Statistical Inference is not a “Proof”. Some of your peers do not agree with your claims as well. For the findings to get anywhere near to a “proof”, they must be replicated by other independent researchers. My remarks here do not refute the fact that I have read and understand the content of all of your papers as well as many more by others.
WHG: Diffraction, or bending of the wave path, is accompanied by changes in frequency response. The resulting distortions may be mitigated through equalization of the drive signal. This is regularly done in a PA setting where CD horns are used.
WHG: If Be is used to make the diaphragm of a given size it will be acoustically smaller than its plastic, aluminum or titanium counterparts; and because of this, it will certainly perform differently as the onset of breakup modes and mass roll-off will occur at higher frequencies whether we choose to hear them or not.
WHG: Attendance at too many Rock Concerts, sans ear plugs, could be the answer as well. In any event the sizzle of a good Zildjian or Paiste Symbol extends to 20KHz and beyond and the differences between them are audible [1] & [2]
WHG: For those that may be interested in the spectral content of musical instruments, reference to the following article is provided.
There's life above 20 kilohertz! A survey of musical instrument spectra to 102.4 kHz
WHG: Here is an on-line listening test for those that might be interested in determining their distortion detection threshold:
Listening Test
Attachments
Last edited:
Just like evolution has not been "proven". Some people just refuse to accept the truth. Our results have been replicated by others - Voishvillo at JBL for example.
Since diffraction is a 3D effect it could only be corrected at a single point, hence it cannot be "mitigated through equalization of the drive signal".
And you seriously believe that this is what makes them sound different. That's a very weak argument.
This is true with existing technology. With enough resolution in the model of a diaphragm/horn, in the (far?) future, one might be able to control in 3d, with an attendant modification to the input signal- (freq X diffracts in XYZ fashion, so in the presence of freq X we overlay Freq A,B,C as their breakup artifacts sum and cancel the artifact of freq X)
Just a silly and unrealistic thought, for all intents and purposes, nowadays, but an interesting mental excercise. There would be all manner of hurdles to this method including a necessarily axisymmetric horn/guide, so long as we're using round CDs.
Sure, if you could control the shape of the diaphragm as a function of frequency you could fix everything. I won't hold my breath waiting for that to happen however.
A set of concentric rings, each individually controlled would be a start. Using a plethora of small transducers is currently vogue as well. There are things that you can do that way as well.
A set of concentric rings, each individually controlled would be a start. Using a plethora of small transducers is currently vogue as well. There are things that you can do that way as well.
That device is a tough one to work with, in my opinion. I had planned to use it at one time, but after learning more about it, decided against. I do think that Bill Waslo is right to use a series of notch filters to try and tame it though.
The thing is, I've never been a fan of notch filters for response shaping. If a driver or horn has a peak within the passband, I don't use it. The driver's electro-mechanical parameters shift at various power points, so using tank circuits to tame resonances is sort of like trying to hold baloons underwater with chopsticks. Any electro-mechanical shifts change the shape of the peaks, and the tank circuits can't shift with them. So since the driver shifts constantly at various power points, the response shape moves too. At higher power levels, the ripple is greater than it is at low power. So which power level do you set the tank circuits for? You have to choose. If it's right at low power, it's not at high power, and vice-versa. That whole approach just isn't for me.
The characteristic response curve of most constant directivity horns and waveguides roughly tracks the plane wave response of the driver. But since they are close to conical shapes, they usually have some ripple, and yet the best ones just have a peak down low and any higher modes well damped, so less ripple. Efficiency increases as frequency rises, so the combined effect of all these things is usually a peak at cutoff followed by a plateau and then a gradual rolloff. I used to say they had a "peak at cutoff and then negative slope", a phrase from an early mentor.
Some waveguide/horns don't just have the bottom peak though, they stay reactive throughout the passband. Such a device has repeated peaks and dips, and they sound unnatural to me as a result. If the ripple isn't too severe and you don't compare it side-by-side with a smoother source, it may not be too objectionable. But do an A/B comparison with a smoother speaker, and the roller coaster response becomes obvious.
As an aside, Robert Moog did a lot of work on circuits that created this kind of roller coaster response, caused by resonances or comb filtering. He found it to be an interesting and pleasant effect. Moog included comb filters which he called "phasers" in his synthesizers and even in single-purpose effect units. His most famous patent was on his "ladder filter", which created a series of ripples in the response curve, with resonances that could be varied in frequency and Q.
So again, some might find this kind of sound "pleasant", even if inaccurate. Others, like you and I, would not. I think Moog's ladder filter is excellent as a musicians tool, used for sound production, part of an artists tool for creating the sound he wants. But for accurate sound reproduction, the roller-coaster response curve is artificial and undesireable, in my opinion.
I think that's really important, when you're dealing with waveguide/horns that introduce anomalies in the passband, you have to try and equalize them out. But can you really? I mean, surely, you can use tank circuits to limit the peaks, setting the response curve closer to flat. And if the on-axis response tracks the power response, all would agree that the equalization is appropriate. Still, I think I would prefer a smoother device, one that didn't need this kind of equalization. It seems like a better approach, and is definitely possible.
Every time I've used a horn that introduced a lot of ripple, the sound was harsh no matter how I equalized it. When equalized flat, it started to sound more "polite" but still sounded spitty and unnatural to me. For the longest time, that's what I though attracted many horn lovers to tractrix and LeCleach horns. They didn't want that harsh unnatural sound that some constant-directivity horns made. Of course, they had to trade polar response for clarity.
So to me, the promise of some of the so-called waveguide/horns is that they provide the smoothness and clarity of the tractrix and LeCleach horns with the polar uniformity of a constant-directivity horn. They're sort of a hybrid between the two. Some older radial horns had this quality too, they sounded very smooth but still had uniform horizontal coverage.
As for me, that's the goal. I want the smoothness of the best old-school horns with the coverage of the constant-directivity devices. I've found it in many radial horns, those with smooth features and throats that aren't too long. And I've found it in certain waveguide/horn profiles too.
Last edited:
Fleas vs Elephants
Darwin's study spanned years over a sample size in the thousands; not a week or two over a sample size of less than 50.
Here is a presentation of Voishvillo's work on the physics of distortion and its perception. At issue here is the metrics of thresholds, not that a particular distortion cannot be heard. It depends on who is doing the listening and the nature of the signal that contains the distortion.
But the resulting changes in frequency response can. (Addressed by the redacted part of my comment)
You will have to convince a whole lot of musicians who buy and use them about the "truth" of this matter as you hear it.
Darwin's study spanned years over a sample size in the thousands; not a week or two over a sample size of less than 50.
Here is a presentation of Voishvillo's work on the physics of distortion and its perception. At issue here is the metrics of thresholds, not that a particular distortion cannot be heard. It depends on who is doing the listening and the nature of the signal that contains the distortion.
But the resulting changes in frequency response can. (Addressed by the redacted part of my comment)
You will have to convince a whole lot of musicians who buy and use them about the "truth" of this matter as you hear it.
Bill
Your attempts to obscure the issues astound me. Voishvillo's conclusions and ours are virtually identical and even his process is identical to what we did. And yet you seem to be trying to present it as a rebuttal to what I have said when in fact it completely supports it. Discussing anything with you just becomes an attempt to dig through your muddling of the issues.
Comparing my study to Darwin's is so absurd as to be laughable - you completely missed the point.
Go ahead and believe what you believe, with no data to support it, and I will follow the data.
Your attempts to obscure the issues astound me. Voishvillo's conclusions and ours are virtually identical and even his process is identical to what we did. And yet you seem to be trying to present it as a rebuttal to what I have said when in fact it completely supports it. Discussing anything with you just becomes an attempt to dig through your muddling of the issues.
Comparing my study to Darwin's is so absurd as to be laughable - you completely missed the point.
Go ahead and believe what you believe, with no data to support it, and I will follow the data.
Wayne - all horn/driver devices have peaks.
Any "electro-mechanical parameters shift " is going to be insignificant. Tank circuits work just fine - you used to suggest them! At any rate, your argument is extremely weak and not at all in line with what I have found. You would have to support what you are saying with some data as I do not see it as reasonable.
Well, still, if possible, I'd trade a small amount of CD performance for a few less notch filters. I agree that for a home system, the tanks are likely to see pretty much the same thing all the time. In the pro world, where drivers are run to thermal limits sometimes, it might be a bigger concern.
I think it might be reasonable to consider that Wayne treats thermal handling from a pro perspective, and thus what might be less of an issue in home use, might be a consideration in his system design- after all, it's one heck of a home that needs the cooling plates on his 12pi hornsub.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.