Yes, as I said, some nonlinearities can be WRT the velocity - like viscous loses which depend on v^2. That's not an amplitude dependence.
An effect whose character would change with level is clearly crossover distortion. For a large signal, this effect is minimal, sometimes .001% but for a small signal its effect can be 50%, if the signal is small enough. That's what makes it so insidious is that its effect is opposite what we think of as nonlinearity. Its level increases just as our hearings masking of it decreases - what could be worse?
In case I didn't make this point clear. We are talking about extremely small delays when we talk about diffraction inside of a horn. This takes them well outside of the normal range of "other delayed reflections from other causes". You should also keep in mind that diffraction is not "white" in that it is not uniform in frequency response as a reflection is. It will, at the very least be high passed, and most likely colored. Make this signal very short in delay and I don't see it as comparable to what we normally think of as a reflection.
David,
I'm not Earl, but I thought I'd mention one difference I can think of (between HOM and more usual delayed reflections) which is the variation in frequency of the refections and their delays. Reflections such as from off a boundary would (I think) be rather broadband (at least within the passband of an involved driver). Reflections from HOM would (probably) vary a lot as frequency dependent modes develop and die out across the frequency range, unlike from simple propagation. Why that would be more audible with level, I don't know. Maybe those weirdly processed sounds are just below threshold of audibility when the SPL levels are low?
I'm not Earl, but I thought I'd mention one difference I can think of (between HOM and more usual delayed reflections) which is the variation in frequency of the refections and their delays. Reflections such as from off a boundary would (I think) be rather broadband (at least within the passband of an involved driver). Reflections from HOM would (probably) vary a lot as frequency dependent modes develop and die out across the frequency range, unlike from simple propagation. Why that would be more audible with level, I don't know. Maybe those weirdly processed sounds are just below threshold of audibility when the SPL levels are low?
Crossover distortion is especially vile, and it has friends and neighbors to help it out. Amplifiers rely on feedback to minimize crossover distortion, but HF stability of the amplifier requires an internal low-pass filter to decrease forward gain at (ultrasonic) high frequencies. This means there is less feedback available at the highest frequencies, where crossover distortion is most troublesome.
Not done yet. The phase margin (stability factor) of a feedback amplifier is degraded by reactive loads. When the phase margin is degraded, slewing events (and clipping) take longer to settle to zero, and effectiveness of feedback is decreased (feedback moves closer to the positive-feedback, unstable region).
Not done yet. Another artifact of Class AB operation are current pulses from the +/- supply rails magnetically radiated into the input sections of the amplifier. This is not theoretical; look at the distortion residue of the amplifier on a scope connected to a distortion analyzer, and move the wires for the plus and minus supplies around. You'll see the shape of the distortion residue change; that's magnetic induction of the Class AB current pulses into the input sections of the amplifier.
I should add that most op-amps have a Class AB output section; this is done to keep the silicon cool, and still deliver the rated voltage swing into a 600-ohm load. They can be externally force-biased into Class A (with a current source), but this can cause the op-amp to enter unusual high-distortion regions.
These electronic distortions, which are worst at low levels, are very different than anything a loudspeaker driver does.
Lynne
Agreed that amplifier distortions have a very different character than loudspeakers do and as such the higher levels in the loudspeaker do not mask those in the amplifier. That said, I do believe that these problems are well understood and can be brought under control. Not that they always are of course.
Agreed that amplifier distortions have a very different character than loudspeakers do and as such the higher levels in the loudspeaker do not mask those in the amplifier. That said, I do believe that these problems are well understood and can be brought under control. Not that they always are of course.
So, how does an OS sound different from a round tractrix ?
To me, both would have not many edges causing reflections, just the expansion compared to a rounded throat to a conical horn...............
To me, both would have not many edges causing reflections, just the expansion compared to a rounded throat to a conical horn...............
David,
Earls response in post #199 to your questions explains his take on why diffraction in horn throats is "audible beyond its measure".
"Take your diffraction slot question as a good point to start. When the wave front reaches this slot two things happen, but you only acknowledge one. First there is a reflection of some of the wave (as you state), but there is also transmission and part of this transmission are waves diffracted from this aperture. The diffracted waves do not travel along the axis of the device, but travel at angles to the axis, bouncing off of the walls of the device. These diffracted waves arrive at the listening, any listener anywhere, delayed in time from the direct wave at delays that depend on location. It is this delayed diffraction signal that is audible beyond its measure."
I agree with Earl that what he suggests may be why horns with smoother transitions (less diffraction artifacts) may subjectively sound less different with changing levels, but have another explanation as well.
We all agree harmonic distortion is non-linear, it changes with voltage.
Even order harmonic distortion always results in the same musical note octaves higher, and is therefore fairly benign, odd order harmonic distortion results in notes (or timbre) not in the original recorded composition, which may be interesting, but sounds quite different (generally worse) as more is added.
Compression drivers have increasingly higher percentages of odd order (objectionable) distortion as voltage is increased.
With a diffraction horn where the response ripple is uniform with axis (like the JBL 2380 family) and can therefore be equalized out, the equalization will not correspond to the non linear distortion changes encountered as voltage (and distortion) increases.
This results in a greater change in sound character as voltage increases compared to a horn that does not require a number of narrow band filters to correct response.
Earl's "Subjective Testing of Compression Drivers" tests with his wife, Lidia W. Lee and Roberto Magalotti of B&C Speakers found 27 college students could not reliably tell the difference between a 6 dB difference in level (14, 20, 28 volts) in three drivers tested at high levels of distortion.
The lowest test voltage used was higher than any of the three driver manufacturers ever publish distortion figures at, the subjects found they could identify the drivers, but not the 6 dB differences in voltage resulting in a narrow envelope of high distortion (5-10% at 1000 Hz ranging up to 10-35% in the 4-9 kHz range) .
From this they concluded that non-linear distortion (THD) is not audible, but linear (frequency response) distortion is.
Since the three driver's distortion characteristics are as unique as their frequency response, I find Earl's conclusion in that study a bit inconclusive 😉.
Diffraction throat horns or driver to horn couplers are a required for the typical line array type of sound reinforcement that have gained popularity over the past two decades, obviously diffraction throats are not going away anytime soon.
That said, for small scale and home use, SPL and dispersion requirements don't require diffraction throats for most applications.
Subjectively, do you not find any difference in low /high level character between the smoother throat JBL horns compared to their diffraction horn counterparts with similar dispersion patterns?
Art
This is where things get complicated. That's because, as it turns out all horns and waveguides have HOMs, just that some have more than others. There does not have to be a slope discontinuity for the presence of HOMs, its just that when there is one then the amount of HOMs are substantially higher. HOMs can be generated within the horn, but they can also be present from the compression driver itself if the phase plug is not properly done, for example. There are lots of ways that HOMs can exist, the idea is to minimize them.
Art - I still don't follow this claim. The subjects could tell the drivers apart - granted, but this was unaffected by the level. Hence the presence or absence of nonlinear distortion was not a factor in the listeners ability to detect the driver. How can it be any other way?
Originally posted by Weltersys:
Earl's "Subjective Testing of Compression Drivers" tests with his wife, Lidia W. Lee and Roberto Magalotti of B&C Speakers found 27 college students could not reliably tell the difference between a 6 dB difference in level (14, 20, 28 volts) in three drivers tested at high levels of distortion.
The lowest test voltage used was higher than any of the three driver manufacturers ever publish distortion figures at, the subjects found they could identify the drivers, but not the 6 dB differences in voltage resulting in a narrow envelope of high distortion (5-10% at 1000 Hz ranging up to 10-35% in the 4-9 kHz range) .
From this they concluded that non-linear distortion (THD) is not audible, but linear (frequency response) distortion is.
Since the three driver's distortion characteristics are as unique as their frequency response, I find Earl's conclusion in that study a bit inconclusive 😉
Your assumption was the driver's frequency response difference is what the listeners detected.
There was no absence of nonlinear distortion in any of the tests at all three levels tested, non-linear distortion in high percentages were present in all the recordings so the individual character of each driver's non-linear as well as linear response is what the subjects heard, and could identify.
Not surprising they could tell the drivers apart at the levels of distortion they were running at, I'd expect the differences between drivers would have been harder to detect if they had been operated in their linear range (say 1-10 watts) rather than a narrow range of high distortion as they were.
Art
Earl's "Subjective Testing of Compression Drivers" tests with his wife, Lidia W. Lee and Roberto Magalotti of B&C Speakers found 27 college students could not reliably tell the difference between a 6 dB difference in level (14, 20, 28 volts) in three drivers tested at high levels of distortion.
The lowest test voltage used was higher than any of the three driver manufacturers ever publish distortion figures at, the subjects found they could identify the drivers, but not the 6 dB differences in voltage resulting in a narrow envelope of high distortion (5-10% at 1000 Hz ranging up to 10-35% in the 4-9 kHz range) .
From this they concluded that non-linear distortion (THD) is not audible, but linear (frequency response) distortion is.
Since the three driver's distortion characteristics are as unique as their frequency response, I find Earl's conclusion in that study a bit inconclusive 😉
Earl,
Your assumption was the driver's frequency response difference is what the listeners detected.
There was no absence of nonlinear distortion in any of the tests at all three levels tested, non-linear distortion in high percentages were present in all the recordings so the individual character of each driver's non-linear as well as linear response is what the subjects heard, and could identify.
Not surprising they could tell the drivers apart at the levels of distortion they were running at, I'd expect the differences between drivers would have been harder to detect if they had been operated in their linear range (say 1-10 watts) rather than a narrow range of high distortion as they were.
Art
Sorry guys but I'm still not seeing it.
You nod to each other knowingly about the evils of diffraction. My image of diffraction at the slot is that it allows the wavefront to spread around the bend and maintain a beamwidth defined by the wall contours. Isn't that exactly what we want? If this particular pernicious form of diffraction exists, can we see a response curve with the offending feature highlighted? Certainly mouth reflections are always there and we see the frequency response effects due to them.
This just sounds too much like an audiophile arguement, an appealing notion without proof. I am not doubting the existence of HOMs or diffraction effects. I'm not even sceptical about the claims of heightened audibility (okay, a little sceptical) I'm just looking for something I would understand, like a reponse curve or a time waveform that shows something beyond the expected mouth reflections,
David
You nod to each other knowingly about the evils of diffraction. My image of diffraction at the slot is that it allows the wavefront to spread around the bend and maintain a beamwidth defined by the wall contours. Isn't that exactly what we want? If this particular pernicious form of diffraction exists, can we see a response curve with the offending feature highlighted? Certainly mouth reflections are always there and we see the frequency response effects due to them.
This just sounds too much like an audiophile arguement, an appealing notion without proof. I am not doubting the existence of HOMs or diffraction effects. I'm not even sceptical about the claims of heightened audibility (okay, a little sceptical) I'm just looking for something I would understand, like a reponse curve or a time waveform that shows something beyond the expected mouth reflections,
David
David,
Yeah, a response curve or something proving heightened audibility of diffraction would be nice, but we probably won't see that.
As much as I am a fan of hard evidence, I still care about subjective impressions from informed listeners such as you (and Earl) with experience in objective analysis and speaker design.
Personally, I don't think diffraction artifacts are nearly as bad as the double digit harmonic distortion we hear in any compression drivers driven above 50% of their rated thermal limits.
So I'll ask the question again: subjectively, do you not find any difference in low/high SPL character between the smoother throat JBL horns compared to their diffraction horn counterparts with similar dispersion patterns?
Come on, talk now before you are silenced by non-disclosure and anti-libel contract clauses with your upcoming employer 🙂!
Art
Last edited:
I wish we, or anyone had all the answers, all the data, a complete picture. We don't. We all act on a limited set of data, add some experience, try some things and move on. I have said before, and I will say it again, and, most importantly, it is the same thing that everyone says (I don't mean about my own but about their own), and that is the speakers that I have now sound better than any speakers that I have ever heard anywhere. Maybe that's just dumb luck, I think not - not after 45+ years of doing this. I am trying to convey why I think that is and the data that I have.
That is the problem. How do you separate the little suckers from the rest of the impulse response?
A copy of the paper can be found here:
CiteSeerX — Audibility of Linear Distortion with Variations in Sound Pressure Level and Group Delay
It's worth reading.
eh, all this difraction gives the music a "live" sound, actually reverby.
It adds well on reverby vocals, but incorrectly it is adding all the time on everything.
That's why the smooth horns sound better.
Waterfall plots can show these reflections (showing the frequencies ring longer than they should), as does wavelet analysis.
see post 1052 and the effect of different screens bouncing super highs back at the horn.
The semi clone of JBL 3731's build thread - Page 36
post 406
Hey guys...we need a little rallying here... - Page 14
There is a big paper measuring 16 different horns (can't find link now). The ones that sounded and measured the best had similarities, such as no diffraction slot, no corners, and a smooth transition in the horn wall, instead of little breaks so common in cd-horns. How well the throat transitions to the start of the horn is another factor. You don't want parallel walls, nor a round opening on a compression driver going into a square throat of the horn. Oh yea, ideally the mouth fold back flush to the baffle, or an exagerated lip over found in le Cleach horns.
Norman
It adds well on reverby vocals, but incorrectly it is adding all the time on everything.
That's why the smooth horns sound better.
Waterfall plots can show these reflections (showing the frequencies ring longer than they should), as does wavelet analysis.
see post 1052 and the effect of different screens bouncing super highs back at the horn.
The semi clone of JBL 3731's build thread - Page 36
post 406
Hey guys...we need a little rallying here... - Page 14
There is a big paper measuring 16 different horns (can't find link now). The ones that sounded and measured the best had similarities, such as no diffraction slot, no corners, and a smooth transition in the horn wall, instead of little breaks so common in cd-horns. How well the throat transitions to the start of the horn is another factor. You don't want parallel walls, nor a round opening on a compression driver going into a square throat of the horn. Oh yea, ideally the mouth fold back flush to the baffle, or an exagerated lip over found in le Cleach horns.
Norman
Yep, found the receipt. I laugh when my guys call me "'The O.G." but When Art speaks, the O.G listens. I will do just that.
found it.
a 3meg pdf on the test on horns and their echos.
Included are le cleach, avante garde, stereo lab, and some older horns including sectoral.
http://forums.melaudia.net/attachment.php?aid=1760
You also have a flat freq response, and now get into dispersion and acoustics of how it interacts with your room.
Less difraction, less reflection, = better sound.
Norman
a 3meg pdf on the test on horns and their echos.
Included are le cleach, avante garde, stereo lab, and some older horns including sectoral.
http://forums.melaudia.net/attachment.php?aid=1760
You also have a flat freq response, and now get into dispersion and acoustics of how it interacts with your room.
Less difraction, less reflection, = better sound.
Norman
First of all that comment is in plain English. This is simply inflammatory. Secondly what you are saying is spot on and employs a great deal of common sense as well. That's just asking for it man, be careful!
Hello Art
FWIW I agree with Earl that the first generation Bi Radial horns could sound harsh depending on crossover points and drive level. That said the latter generation does not suffer from the same issues at least IMHO.
This of course has no meaning as it is a subjective evaluation with no data to back it up.
I would really like to see a measurement for HOM's that we could use to evaluate horns/waveguides.
Rob🙂
Thoughts:
Please correct me if I'm off course. Except for the first thought, which is more of a statement, as I have issues with cable people and hairy spiders etc.
I figure if its there, it can be measured in in the fr or distortion plot. If it can't be measure its not there. Moving on..
I'm thinking that diffraction is a real vs. straw issue when it occurs x distance from the throat/driver and results in true "more speakers than one" filter/phasey combing etc., other than that we have a Ray Bolger in Wizard of Oz impression.
Regarding EQ for simple radial horns et al:
Fat man mirror horn is easily fixed by transfer thin man mirror (EQ). Broken mirror all over space is hard to invert/transfer. Most diffraction horns are far far from the broken mirror analogy, but it may be convenient to say otherwise.
But to put it all to rest do we have audio software that can visually display a model of horn and emitter diffraction thusly, and if not, why not? Not talking auto EQ stuff here but in depth modeling and a visual representation of changes in a horn model wavefront, if not in real in depth in RT, time than as rough soon deeper later as model progresses.
Wavefront Analyzer - Types of LASIK Eye Surgery - TLC LASIK
Please correct me if I'm off course. Except for the first thought, which is more of a statement, as I have issues with cable people and hairy spiders etc.
I figure if its there, it can be measured in in the fr or distortion plot. If it can't be measure its not there. Moving on..
I'm thinking that diffraction is a real vs. straw issue when it occurs x distance from the throat/driver and results in true "more speakers than one" filter/phasey combing etc., other than that we have a Ray Bolger in Wizard of Oz impression.
Regarding EQ for simple radial horns et al:
Fat man mirror horn is easily fixed by transfer thin man mirror (EQ). Broken mirror all over space is hard to invert/transfer. Most diffraction horns are far far from the broken mirror analogy, but it may be convenient to say otherwise.
But to put it all to rest do we have audio software that can visually display a model of horn and emitter diffraction thusly, and if not, why not? Not talking auto EQ stuff here but in depth modeling and a visual representation of changes in a horn model wavefront, if not in real in depth in RT, time than as rough soon deeper later as model progresses.
Wavefront Analyzer - Types of LASIK Eye Surgery - TLC LASIK
Last edited:
- Status
- Not open for further replies.