Re: Re: Re: Re: Comments on data
Rather than invest the time modeling things in Matlab based on a hypothesis, why don't you simply go to the garage and make some waveguides loaded with open cell foam? If you cut a few corners, you can knock something out in less than the time you would spend in Matlab.
If I am not mistaken, there are only two people on this thread who've heard the difference it can make. (If anyone else has heard the Summa, or built your own waveguides with a HOM reducing plug, please speak up!)
I can't recommend this mod more highly.
Rybaudio said:
Rather than invest the time modeling things in Matlab based on a hypothesis, why don't you simply go to the garage and make some waveguides loaded with open cell foam? If you cut a few corners, you can knock something out in less than the time you would spend in Matlab.
If I am not mistaken, there are only two people on this thread who've heard the difference it can make. (If anyone else has heard the Summa, or built your own waveguides with a HOM reducing plug, please speak up!)
I can't recommend this mod more highly.
Re: Re: Re: Re: Re: Re: Re: Comments on data
Yeah, it would would help! But somehwere in this thread Geddes published the FR with foam. IIRC, the roll off was in the top octave, 10K and above. Is that right?
Sure not what I heard with my "HOM Foam". Using the right type of foam in the right way is going to be essenial.
Will start looking for some.
("E.T., Phone Home" turns into "E.G., Foam HOM').
Patrick Bateman said:
Yeah, it would would help! But somehwere in this thread Geddes published the FR with foam. IIRC, the roll off was in the top octave, 10K and above. Is that right?
Sure not what I heard with my "HOM Foam". Using the right type of foam in the right way is going to be essenial.
Will start looking for some.
("E.T., Phone Home" turns into "E.G., Foam HOM').
I tried four or five densities of foam, different shapes, etc.
Only open cell foam would ever work and even then only a fairly narrow range of densities. In my system the response was not affected at all at 1 kHz. and is down by 3 dB at 10 kHz. The response drop at about 15 kHz is in the compression driver.
the 3 dB drop is easily recovered in the crossover, but it has to be done. Just putting foam in a horn without making crossover changes is a sure disaster.
And I agree with Patrick. If you have not heard the use of foam properly done then you need to. Its not a small change.
Only open cell foam would ever work and even then only a fairly narrow range of densities. In my system the response was not affected at all at 1 kHz. and is down by 3 dB at 10 kHz. The response drop at about 15 kHz is in the compression driver.
the 3 dB drop is easily recovered in the crossover, but it has to be done. Just putting foam in a horn without making crossover changes is a sure disaster.
And I agree with Patrick. If you have not heard the use of foam properly done then you need to. Its not a small change.
Re: Re: Re: Re: Re: Comments on data
I've heard the Summa. The Summa had no "horn" colorations that I could discern at all, and sounded very much like the frequency response and impulse measurements shown here.
People who love the old-school horns may not notice what Dr. Geddes' waveguide/foam does, since I suspect they aren't bothered by HOM's as much as other listeners. I don't know if that's a matter of individual perception or self-imposed training to ignore the HOM coloration. If old-school "horn" colorations bother you, though, you owe it to yourself to audition the Geddes waveguide/foam system.
Patrick Bateman said:
I've heard the Summa. The Summa had no "horn" colorations that I could discern at all, and sounded very much like the frequency response and impulse measurements shown here.
People who love the old-school horns may not notice what Dr. Geddes' waveguide/foam does, since I suspect they aren't bothered by HOM's as much as other listeners. I don't know if that's a matter of individual perception or self-imposed training to ignore the HOM coloration. If old-school "horn" colorations bother you, though, you owe it to yourself to audition the Geddes waveguide/foam system.
I started with a full range driver used dipole, and then built a waveguide to control directivity. I went through several iterations of WG, and the resulting HOM's were quite obvious and offensive. Lining the WG's with open cell foam got me back to the natural sound of the driver run dipole, but with the directivity control that was my original goal. Once you can really hear the creation of HOM's and make them go away, then you'll become sold on Dr. Geddes approach.
If anyone who has some HOM foam in use lives within a 4 hour or so drive of central Pennsylvania and will give me an earfull let me know. I'm still waiting to hear this magical difference. I'm definately one of the people that is bothered by "old-school horn colorations" (almost every horn/CD system) but every time I measure or look at measurements of these systems there are things going on a lot less subtle than HOM. I usually see some combination of these problems:
- lack of high end extension (sourced in the driver)
- driver diaphram resonances in the 10-20k region
- improper EQ (the set of transfer functions on and off axis has not been optimized)
- horn's radiation pattern is nonuniform in frequency (when the full set of on and off axis transfer functions is optimized together they are individually far away from optimal)
- the frequency dependence of the horn's radiation pattern is different in the horizontal and vertical planes (and in between)- this (usually conveniently unmentioned) trait makes it very hard to optimize the set of on and off-axis transfer functions for the reason mentioned above.
It seems that there are so few systems out there that get these items right it is hard to make a fair assessment. Most pro audio speakers completely ignore the last point all-together because they're primarily concerned with the horizontal radiation pattern close to 0 degrees vertically, but in a home where the entire soundfield produced by the speaker influences the percieved tonality this cannot be ignored. How many horn speakers out there (1) use a compression driver that has solid high end extension and no major problems below 20k (2) use an axisymmetric (or similarly horiz/vert behaving) horn whose directivity is close to uniform in frequency and (3) are truely optimized EQ-wise (decisions were not limited by means (not enough time, need to use passive, limited processing power, limited measurement resolution, limited measurement completeness(ie on axis only)) or designer skill)? It is within this set of speakers that we should be evaluating subtle differences.
A general question to those of you that have worked with or heard speakers using this foam:
Comparing the before and after (adding foam) family of transfer functions of the speaker (after adjusting EQ) exactly how close were they? I'm mainly interested in isolating the foam's effect and trying to elminate the variable of the ease of EQing the system In other words, I want to know if without the foam you really did have the system's transfer function optimized (may require several high Q bands of EQ) or does the foam version sound better because it was easier to EQ (damped some of the acoustical resonances making the system easier to EQ) and hence ended up with a more optimized set of transfer functions?
A few specific questions:
- Did the foam effect the radiation pattern of the horn?
- Do your compression drivers have diaphram resonances in the 10-20k region, and were the radiated effects of this EQed to the same level before and after the foam? Here I want to make sure that the "horn sound" wasn't removed simply because the peak in the response around these resonances was attenuated to a level lower relative to the nominal level of the response than it was without the foam.
I don't doubt that when the foam is used it results in a better sounding speaker a lot of the time, but if I can (with some effort) do the same thing with EQ and not lose efficiency, I'd prefer to do that. I want to look at the system's behavior modulo any EQ applied to it.
PS- if anyone is organizing the group buy put me down for a pair.
- lack of high end extension (sourced in the driver)
- driver diaphram resonances in the 10-20k region
- improper EQ (the set of transfer functions on and off axis has not been optimized)
- horn's radiation pattern is nonuniform in frequency (when the full set of on and off axis transfer functions is optimized together they are individually far away from optimal)
- the frequency dependence of the horn's radiation pattern is different in the horizontal and vertical planes (and in between)- this (usually conveniently unmentioned) trait makes it very hard to optimize the set of on and off-axis transfer functions for the reason mentioned above.
It seems that there are so few systems out there that get these items right it is hard to make a fair assessment. Most pro audio speakers completely ignore the last point all-together because they're primarily concerned with the horizontal radiation pattern close to 0 degrees vertically, but in a home where the entire soundfield produced by the speaker influences the percieved tonality this cannot be ignored. How many horn speakers out there (1) use a compression driver that has solid high end extension and no major problems below 20k (2) use an axisymmetric (or similarly horiz/vert behaving) horn whose directivity is close to uniform in frequency and (3) are truely optimized EQ-wise (decisions were not limited by means (not enough time, need to use passive, limited processing power, limited measurement resolution, limited measurement completeness(ie on axis only)) or designer skill)? It is within this set of speakers that we should be evaluating subtle differences.
A general question to those of you that have worked with or heard speakers using this foam:
Comparing the before and after (adding foam) family of transfer functions of the speaker (after adjusting EQ) exactly how close were they? I'm mainly interested in isolating the foam's effect and trying to elminate the variable of the ease of EQing the system In other words, I want to know if without the foam you really did have the system's transfer function optimized (may require several high Q bands of EQ) or does the foam version sound better because it was easier to EQ (damped some of the acoustical resonances making the system easier to EQ) and hence ended up with a more optimized set of transfer functions?
A few specific questions:
- Did the foam effect the radiation pattern of the horn?
- Do your compression drivers have diaphram resonances in the 10-20k region, and were the radiated effects of this EQed to the same level before and after the foam? Here I want to make sure that the "horn sound" wasn't removed simply because the peak in the response around these resonances was attenuated to a level lower relative to the nominal level of the response than it was without the foam.
I don't doubt that when the foam is used it results in a better sounding speaker a lot of the time, but if I can (with some effort) do the same thing with EQ and not lose efficiency, I'd prefer to do that. I want to look at the system's behavior modulo any EQ applied to it.
PS- if anyone is organizing the group buy put me down for a pair.
How much foam do you need? I bought a half a sheet of 1.5" foam and cut the plugs for my ID horns out of that. I would be glad to hook an old Termpro buddy up with some. I got the foam from a place on another link, foamdirect or something like that- 30ppi urethane foam.
I'm using Beyma CP755-NDs on my straight ID horns...only had to bump up 16 and 20k by a dB or 2. My RTA is broken and I've never got my copy of ELF to work on any computer I have, so I can't take measurements of the effects for you.
Midbasses are the new ID xS69s, and have 2 IDW15s in the back. Amps are HSS Fidelity class A tube on the horns, and Linear Powers on the midbasses and subs. Processing is up in the air right now.
Everything is good, going back to school and found a wonderful woman (again, she was my old high school sweetie).
I'm on IM every now and again, same screen name- IDAccord.
I'm using Beyma CP755-NDs on my straight ID horns...only had to bump up 16 and 20k by a dB or 2. My RTA is broken and I've never got my copy of ELF to work on any computer I have, so I can't take measurements of the effects for you.
Midbasses are the new ID xS69s, and have 2 IDW15s in the back. Amps are HSS Fidelity class A tube on the horns, and Linear Powers on the midbasses and subs. Processing is up in the air right now.
Everything is good, going back to school and found a wonderful woman (again, she was my old high school sweetie).
I'm on IM every now and again, same screen name- IDAccord.
winslow,
I'll jump on AOL sometime and shoot you a message. Those are some pretty heavy duty compression drivers for car use!
Dr. Geddes,
I have a question about some research of yours. I read your paper "Subjective Testing of Compression Drivers" when it first came out in the AES journal and found it interesting. When writing the above post I considered listing nonlinear distortion as a common problem because I have noticed that it can vary widely (an order or two of magnitude) between compression drivers and on several occasions I have had good reason to believe it was the culprit (other factors weren't present, measured nonlinear distortion was "high", sounded like a lot of low order nonlinear distortion usually does). I didn't add it because I remembered your article threw some doubt on its significance. Instead, I reread the article and noticed something I overlooked the first time around: all of the drive levels in the comparison were very high (~10% distortion). Is this not unacceptably high distortion in the first place? My main question is if there was an audible difference between low level playback (at a level of distortion known to be negligible) and the three higher levels of playback used in the comparison? It seems to me that playback at 8% and 20% distortion might not sound all that different (both crappy) but that playback at 1% and 10% distortion could sound very different. BTW, excuse my rough use of terminology here- by playback at a given level of distortion I mean playback at an RMS drive voltage that corresponds to the drive voltage for that level of THD when driven with a sine wave. If my question is not clear please ask me for a clarification.
I'll jump on AOL sometime and shoot you a message. Those are some pretty heavy duty compression drivers for car use!
Dr. Geddes,
I have a question about some research of yours. I read your paper "Subjective Testing of Compression Drivers" when it first came out in the AES journal and found it interesting. When writing the above post I considered listing nonlinear distortion as a common problem because I have noticed that it can vary widely (an order or two of magnitude) between compression drivers and on several occasions I have had good reason to believe it was the culprit (other factors weren't present, measured nonlinear distortion was "high", sounded like a lot of low order nonlinear distortion usually does). I didn't add it because I remembered your article threw some doubt on its significance. Instead, I reread the article and noticed something I overlooked the first time around: all of the drive levels in the comparison were very high (~10% distortion). Is this not unacceptably high distortion in the first place? My main question is if there was an audible difference between low level playback (at a level of distortion known to be negligible) and the three higher levels of playback used in the comparison? It seems to me that playback at 8% and 20% distortion might not sound all that different (both crappy) but that playback at 1% and 10% distortion could sound very different. BTW, excuse my rough use of terminology here- by playback at a given level of distortion I mean playback at an RMS drive voltage that corresponds to the drive voltage for that level of THD when driven with a sine wave. If my question is not clear please ask me for a clarification.
Rybaudio said:
Does everyone understand why the waveguides are loaded with foam?
It's to deal with REFLECTIONS in the waveguide.
Picture the following sequence of events. This is a drastic simplification, but you get the general idea:
1: A soundwave is transmitted down the length of a waveguide.
2: A fraction of the soundwave's energy is reflected back down the throat.
3: A few milliseconds later the reflected wave exits the waveguide.
4: The aforementioned sound wave is now DELAYED in time, and it's frequency response is now dramatically altered.
The foam simply attenuates these reflections.
Does that make sense?
Can you see how no amount of EQ can completely fix a problem with reflections?
Yes, I understand that EQ can correct frequency response aberrations caused by reflections in the waveguide. But EQ CANNOT fix time problems. Most importantly, why not reduce the HOMs alone?
Amplifier power is plentiful and affordable. Even if you're into tube amps, compression drivers have PLENTY of headroom.
:: PB ::
I think that that picture isn't entirely faithful to reality. As a counterexample consider a more common and well understood situation: if you place a subwoofer into a room the reflections from the walls do the same thing; to get a nice picture of the modal behavior you can look at the image sources for example. It would seem that the time lag from the reflections would not be invertible, but it in fact is! If you fix the frequency domain the time domain snaps into place as well! If HOM are what I think they are (the spatial parts of the solution to the wave equation that correspond to separation constants above the lowest one) then the time/frequency domain problems may be amenable to inversion via upstream processing.
a little less simplified: some of the wave front coming out of the CD goes right at the throat of the guide and ricochets until it escapes. How many reflections is up for grabs. The frequency relationship to the number of reflections is out there as well. The foam attenuates those reflections more due to their longer reflected path out of the guide...those waves that are most disruptive to the "fidelity" of the signal are attenuated most.
If a CD provided a planar wave front this would not be the problem it is.
If a CD provided a planar wave front this would not be the problem it is.
Rybaudio said:
Huh?
That makes no sense.
If you have a late reflection INSIDE the waveguide, and you attenuate it with EQ, the signal is still "smeared" in time.
I live in Washington. (The NY location in my profile is a joke.)
Come to my house, listen to my foam loaded waveguides, you'll "hear" what I am talking about.
This isn't a subtle difference by any means. In my opinion, it's one of the most AUDIBLE improvements you could make to a horn or waveguide.
For example, I've used $80 compression drivers and I've used $400 compression drivers on the same waveguide. The improvement rendered by the foam mod is THAT significant - it's money well spent, believe me.
Keep in mind you WILL need EQ and a bit of know-how to fix the change in frequency response.
We're fixing the REFLECTION problem, not the frequency response problem. This is mostly a problem in the time-domain.
:: PB ::
Comments
A lot of discussion - a lot of it wrong.
First, Summa's or ESP's as they are now called, do meet all of the requirements asked for above except maybe a diaphragm resonance at 16 kHz. and a falloff above that. but here I would dispute the need or even the audibility of anything above 10 kHz. But otherwise the ESP line meets all of the requirements stated.
The "invertibility" argument is flawed since it is one-dimensional. The time-frequency can be corrected, exactly, with EQ, but only at one spatial position and only for a single ray of sound. No global sound field correction of an HOM, a horn reflection or cabinet diffraction is possible with electronic EQ. Thus acoustic correction via foam and correction via EQ are completely different - the acoustic solution being global and the EQ being limited to a single point in space. The only solutions of loudspeaker problems that truely work are acoustic ones.
RybAudio - you did not read my compression driver paper very carefully or not carefully enough. Levels of input were used from low level (low distortion) to high level (high distortion) and the subjects could not hear the distortion (statistically) when compared to an undistorted signal - not compared to each other. The reference was always an undistorted signal. Thus your comment is not relavent.
A lot of discussion - a lot of it wrong.
First, Summa's or ESP's as they are now called, do meet all of the requirements asked for above except maybe a diaphragm resonance at 16 kHz. and a falloff above that. but here I would dispute the need or even the audibility of anything above 10 kHz. But otherwise the ESP line meets all of the requirements stated.
The "invertibility" argument is flawed since it is one-dimensional. The time-frequency can be corrected, exactly, with EQ, but only at one spatial position and only for a single ray of sound. No global sound field correction of an HOM, a horn reflection or cabinet diffraction is possible with electronic EQ. Thus acoustic correction via foam and correction via EQ are completely different - the acoustic solution being global and the EQ being limited to a single point in space. The only solutions of loudspeaker problems that truely work are acoustic ones.
RybAudio - you did not read my compression driver paper very carefully or not carefully enough. Levels of input were used from low level (low distortion) to high level (high distortion) and the subjects could not hear the distortion (statistically) when compared to an undistorted signal - not compared to each other. The reference was always an undistorted signal. Thus your comment is not relavent.
More comments
I should also point out that HOM and diffraction are both non-minimum phase and as such to EQ them in BOTH the time and frequency domains will require a non-minimum phase EQ, which cannot really be done analog - only DSP. You cannot simply EQ a waveguide with a minimum phase EQ and correct the time domain. Doesn't work that way.
One guy had a comment "Too much "mere verbal dispute" going on in this thread."
I'd like to point out that I doubt that more scientific work has gone into any loudspeaker design than what is in mine. I am a scientist - I don't do anything without scientific support. I'm not simply waving my hands here - I have the data and the math to back up everything that I say and do. It is simply not possible (or advisable) to get into extensive detail in posts like this. Read my published work and you will see that I do my homework.
I should also point out that HOM and diffraction are both non-minimum phase and as such to EQ them in BOTH the time and frequency domains will require a non-minimum phase EQ, which cannot really be done analog - only DSP. You cannot simply EQ a waveguide with a minimum phase EQ and correct the time domain. Doesn't work that way.
One guy had a comment "Too much "mere verbal dispute" going on in this thread."
I'd like to point out that I doubt that more scientific work has gone into any loudspeaker design than what is in mine. I am a scientist - I don't do anything without scientific support. I'm not simply waving my hands here - I have the data and the math to back up everything that I say and do. It is simply not possible (or advisable) to get into extensive detail in posts like this. Read my published work and you will see that I do my homework.