catapult said:
I'm sorry but I didn't see any discussions in that article of the type that we have been discussing here.
Robert - you incorrectly atribute the conical waveguide requirement for a spherical wave at the throat to Putland. He did make this observation, but only after I did so in my paper, which preceded his. He seems to have led people to believe that he figured this out (you are not the first to make this error), since he makes no mention of the fact that I had also pointed this out in my paper. An error on his part that seems to be getting propagated.
Earl,
Haven't read your waveguide papers in a couple of years. I forget what the issue is about planar wavefronts entering a conical waveguide, and why spherical wavefronts are preferable for such waveguides.
Would you refresh my memory about why planar into conical is not a desirable condition, in non-mathematical terms, and why it is presumably relatively worse than any possible HOM creation in an OS waveguide.
Since planar wavefront compression drivers are a practical given, and conicals are so much easier and cheaper to make, what are the tradeoffs where neither is the perfect ideal.
Haven't read your waveguide papers in a couple of years. I forget what the issue is about planar wavefronts entering a conical waveguide, and why spherical wavefronts are preferable for such waveguides.
Would you refresh my memory about why planar into conical is not a desirable condition, in non-mathematical terms, and why it is presumably relatively worse than any possible HOM creation in an OS waveguide.
Since planar wavefront compression drivers are a practical given, and conicals are so much easier and cheaper to make, what are the tradeoffs where neither is the perfect ideal.
pooge said:
Your asking for "brain surgery made easy", or "waveguide theory for dummy's" and I'm not so sure how successful I can be.
For any coordinate system there is a shape to the coordinates at any location. For a spherical waveguide this shape is always a sphere, but its radius varies. For OS, for example, it is flat at the origin and very nearly spherical at larger values of the "radial" coordinate.
Now if the wavefront does not match the shpe of the coordinate system at its "entrance" then it has to be "fit" to the boundary conditions by taking a sum of terms, or modes, such that this sum fits the boundary condition. When there is a perfect fit, this is the classical 1P concept. For a spherical waveguide only a spherical wavefront at the throat will be composed of a single "mode", namely a radially propagating spherical wave. Any other shape will require HOM (that loathsome term defining non-ideal wave propagation conditions) created right at the entrance and these HOM will then propagate to the mouth as waves that bounce off of the walls and travel a longer path length, thus delaying them in time.
Now a perfectly flat wavefront at the throat of an OS waveguide will not require any HOM to "fit" the boundary condition and a "nearly" pure 1P wave will propagate ending up as a spherical wave at the mouth. However, the math of the OS coordinate system requires what is usually called "leakage". By this I mean that the main mode continuously leaks into HOM as it propagates. This is very small effect at LFs, but gets more pronounced as the frequencies go up. At any rate the actual effects are quite complicated and mathematically very hard to compute.
A flat wavefront at the throat of a spherical waveguide however, will have HOM right from the start, at all frequencies. There will be HOMs in evidence at very LFs and they will increase in level all the way up to the HFs. A dome will be a "better" fit to a shperical waveguide than a flat wavefront, but still not ideal. There will be HOM present at all frequencies.
The least HOM will be generated by a flat wavefront - from a compression driver - into a OS waveguide. Regardless of what others here may say, this is the situation - like it or not.
I KNEW you couldn't keep the math out of it! 
So the bottom line is more HOMs starting at a lower frequency (harder to filter with foam) and getting higher in amplitude with frequency.
What does equalizing the high frequencies do; i.e., does boosting the high frequencies due to their loss in the foam, and for the nature of the waveguide itself, create a higher level of HOMs that offset their loss in the foam (or is it a situation in which equalization is done by attenuating the lower frequencies to match woofer sensitivity)?
So the bottom line is more HOMs starting at a lower frequency (harder to filter with foam) and getting higher in amplitude with frequency.
What does equalizing the high frequencies do; i.e., does boosting the high frequencies due to their loss in the foam, and for the nature of the waveguide itself, create a higher level of HOMs that offset their loss in the foam (or is it a situation in which equalization is done by attenuating the lower frequencies to match woofer sensitivity)?
pooge said:I KNEW you couldn't keep the math out of it!
So the bottom line is more HOMs starting at a lower frequency (harder to filter with foam) and getting higher in amplitude with frequency.
What does equalizing the high frequencies do; i.e., does boosting the high frequencies due to their loss in the foam, and for the nature of the waveguide itself, create a higher level of HOMs that offset their loss in the foam (or is it a situation in which equalization is done by attenuating the lower frequencies to match woofer sensitivity)?
Well there weren't any equations!
The LFs in a waveguide are attenuated - passively thats all you can do. Since the system is linear, boosting or cutting does not affect the ratios of the HOM.
I looked at the Laplace equation in two dimensions to get a conceptual handle on how near field waveguides might work whilst preparing my waveguide article.
I found this useful because the complex Laplace equation is satisfied by a pair of conjugate harmonic functions with which you can form a coordinate system that consists of streamlines that follow the duct wall and a set of orthogonal equipotential surfaces across the duct.
For an axis symmetrical duct you can specify as the “r” of the kr number of an acoustic field in the duct, by the length of a tangent from a point on the wall to where it crosses the duct axis.
When at kr=1 the phase angle of the specific impedance is 45 degrees as the kr number increases the phase angle reduces.
Since the solutions are complex conjugates the imaginary part disappears under convolution and the wave front travelling down the duct can be seen as moving these solutions into phase and convolving them, increasingly cancelling the imaginary part.
What happens now is that the field starts to look more like a real potential field and the stream function that defines the stream lines becomes rapidly smaller and the wavefront is now more aptly described as an equipotential surface between a pair of tangents to the duct wall.
In this region the wavefront begins to leave the wall if the wall is still curving away because the field is now predominantly a radially symmetrical scalar field, and all of the vectors in its associated vector field point to the origin.
What I then did was to look at what was needed for constant directivity given this beaming behaviour.
Noting that a piston radiator needs to have a diameter that reduces at the same rate the frequency increases to have constant directivity, if the “r” value of the acoustic field reduced at this rate the piston that may be said to be equivalent to the wavefront would result in decreasing this notional diameter at a suitable rate to give constant directivity.
So be doing some calculations I determined that a circular arc starting at the throat at the required angle and meeting the baffle at right angles would be a good candidate for this.
Rcw.
I found this useful because the complex Laplace equation is satisfied by a pair of conjugate harmonic functions with which you can form a coordinate system that consists of streamlines that follow the duct wall and a set of orthogonal equipotential surfaces across the duct.
For an axis symmetrical duct you can specify as the “r” of the kr number of an acoustic field in the duct, by the length of a tangent from a point on the wall to where it crosses the duct axis.
When at kr=1 the phase angle of the specific impedance is 45 degrees as the kr number increases the phase angle reduces.
Since the solutions are complex conjugates the imaginary part disappears under convolution and the wave front travelling down the duct can be seen as moving these solutions into phase and convolving them, increasingly cancelling the imaginary part.
What happens now is that the field starts to look more like a real potential field and the stream function that defines the stream lines becomes rapidly smaller and the wavefront is now more aptly described as an equipotential surface between a pair of tangents to the duct wall.
In this region the wavefront begins to leave the wall if the wall is still curving away because the field is now predominantly a radially symmetrical scalar field, and all of the vectors in its associated vector field point to the origin.
What I then did was to look at what was needed for constant directivity given this beaming behaviour.
Noting that a piston radiator needs to have a diameter that reduces at the same rate the frequency increases to have constant directivity, if the “r” value of the acoustic field reduced at this rate the piston that may be said to be equivalent to the wavefront would result in decreasing this notional diameter at a suitable rate to give constant directivity.
So be doing some calculations I determined that a circular arc starting at the throat at the required angle and meeting the baffle at right angles would be a good candidate for this.
Rcw.
Sound card
I am posting here as continuing in the holm impulse thread would be horribly off topic.
Please do not think this is an attack, I consider myself a ‘disciple’ of your schools of thought on sound reproduction, however I feel something must be said. You constantly talk of hard evidence, referencing papers, actual results etc... however the ‘evidence’ you give for your own loudspeakers is data acquired using the ONBOARD sound on some cheap oem motherboard in a package HP computer. As a scientist, how can you consider this acceptable? I have little doubt that the onboard sound in question will be a ‘realtek AC97 codec’ type that will resample everything going through it and be of (at best) questionable accuracy. Not to mention the noise issues that usually plague such integrated devices.
May I ask what kind of microphone you use? As your choice of soundcard is the equivalent of ripping the electret insert out of a $2 skype headset and calling it a highly accurate measurement mic....
gedlee said:
Theo404 said:
gedlee said:
I am posting here as continuing in the holm impulse thread would be horribly off topic.
Please do not think this is an attack, I consider myself a ‘disciple’ of your schools of thought on sound reproduction, however I feel something must be said. You constantly talk of hard evidence, referencing papers, actual results etc... however the ‘evidence’ you give for your own loudspeakers is data acquired using the ONBOARD sound on some cheap oem motherboard in a package HP computer. As a scientist, how can you consider this acceptable? I have little doubt that the onboard sound in question will be a ‘realtek AC97 codec’ type that will resample everything going through it and be of (at best) questionable accuracy. Not to mention the noise issues that usually plague such integrated devices.
May I ask what kind of microphone you use? As your choice of soundcard is the equivalent of ripping the electret insert out of a $2 skype headset and calling it a highly accurate measurement mic....
There are all sorts of “tricks of the trade” which enable you to improve accuracy of measurements with quite simple and modestly specified equipment.
A very useful one is based upon the “root n” effect.
This is a statistical effect that the square root of the simple average of a series of measurements is this much more accurate than any single measurement alone, provided the errors in measurement are randomly distributed about the exact value.
This has diminishing returns because you need to take 100 measurements to get a ten times improvement, but taking nine for instance has a useful three times improvement.
There are various signal processing techniques in speaker measurement software such as Fourier analysis that enable you to such things as auto correlation.
This is based upon the statistical principle that a signal correlated with a phase shifted version of itself will have no correlation if the signal is random.
There are many other techniques that you can use and these are just two of the easier ones.
Rcw.
A very useful one is based upon the “root n” effect.
This is a statistical effect that the square root of the simple average of a series of measurements is this much more accurate than any single measurement alone, provided the errors in measurement are randomly distributed about the exact value.
This has diminishing returns because you need to take 100 measurements to get a ten times improvement, but taking nine for instance has a useful three times improvement.
There are various signal processing techniques in speaker measurement software such as Fourier analysis that enable you to such things as auto correlation.
This is based upon the statistical principle that a signal correlated with a phase shifted version of itself will have no correlation if the signal is random.
There are many other techniques that you can use and these are just two of the easier ones.
Rcw.
Re: Sound card
Dr. Gedddes uses an Earthworks mic for his critical measurements, which is highly accurate.
As for his soundcard choice, I've compared results with my m-audio outboard card and the onboard hp laptop realtek soundcard, there is no difference in my measurements what so ever.
Dr. Geddes has also mentioned in the past something consistent with what I had found, even the cheapest of microphones, the freebies that come with computers, seem to be able to do perfectly acceptable LF meausrements for room acoustics.
I also have B&K measurement mics as well as what looks like a very cheap mic that came with my ATB PC Pro software, and it's every bit as accurate, within it's respective range, as the B&K mic. Both were tested in a lab here against a reference mic and both were give back to me with reference files indicating no adjustments within the audible range.
I think people over-state the importance of high quality equipment for getting accurate measurements. As was said, a lot of tricks are used in the software that compensates for the problems within the equipment. This allows for a cheap and noisy soundcard to suddenly become a precision measuring device.
Theo404 said:
Dr. Gedddes uses an Earthworks mic for his critical measurements, which is highly accurate.
As for his soundcard choice, I've compared results with my m-audio outboard card and the onboard hp laptop realtek soundcard, there is no difference in my measurements what so ever.
Dr. Geddes has also mentioned in the past something consistent with what I had found, even the cheapest of microphones, the freebies that come with computers, seem to be able to do perfectly acceptable LF meausrements for room acoustics.
I also have B&K measurement mics as well as what looks like a very cheap mic that came with my ATB PC Pro software, and it's every bit as accurate, within it's respective range, as the B&K mic. Both were tested in a lab here against a reference mic and both were give back to me with reference files indicating no adjustments within the audible range.
I think people over-state the importance of high quality equipment for getting accurate measurements. As was said, a lot of tricks are used in the software that compensates for the problems within the equipment. This allows for a cheap and noisy soundcard to suddenly become a precision measuring device.
Re: Re: Sound card
This is exactly my point, it seems very odd to me to get a mic of that value (presumably hire) and attach it to dollars worth of noisy onboard soundcard...
I wasnt aware comparisons had been made, I can see that with the convolution methods usually employed in measurements that noise isnt so much of an issue, but surely it still is to an extent, especially when considering the rigor which Dr. Geddes employs in his testing methodology.
pjpoes said:
This is exactly my point, it seems very odd to me to get a mic of that value (presumably hire) and attach it to dollars worth of noisy onboard soundcard...
I wasnt aware comparisons had been made, I can see that with the convolution methods usually employed in measurements that noise isnt so much of an issue, but surely it still is to an extent, especially when considering the rigor which Dr. Geddes employs in his testing methodology.
My measurements are accurate - period!
It is absurd to worry about the sound card as it is more than sufficiently accurate for the task. Even the Realtek sound card is more accurate than the Earthworks microphone. This whole discussion is ridiculous as anyone who understand acoustic measurements will tell you. Any errors that could exist in the sound card are orders of magnitude below the errors that occur in the microphone, its preamp and, last but not least, the measurement technique itself.
I started my career using HP spectrum analyzers that had BAD 12 bit A/D's. These analyzers cost over $20,000 in their day and they were excellent for the task. Today a sound card in a PC is about 100 times better than that HP analyzer. Get real!
It is absurd to worry about the sound card as it is more than sufficiently accurate for the task. Even the Realtek sound card is more accurate than the Earthworks microphone. This whole discussion is ridiculous as anyone who understand acoustic measurements will tell you. Any errors that could exist in the sound card are orders of magnitude below the errors that occur in the microphone, its preamp and, last but not least, the measurement technique itself.
I started my career using HP spectrum analyzers that had BAD 12 bit A/D's. These analyzers cost over $20,000 in their day and they were excellent for the task. Today a sound card in a PC is about 100 times better than that HP analyzer. Get real!
rcw said:
I continue to not understand your claims and this kind of writup does not help. For me to accept what you are claiming I would have to see a mathematical description and derivation of what you are doing, the kind of thing that can't be done in a internet forum. I had hoped to find that on your web site, but I didn't see anything there that was applicable.
I don't even understand what you mean by the "complex" Laplace equation is since this is simple single term a second order Diff EQ. The equation itself is not complex, but could be applied to complex solutions, although all real physics have to have real solutions. At any rate the Laplace equation is only a rough approximation of the wave equation at very low frequencies (exact when k=0). Basically, what you are saying just does not seem valid to me and you would have to prove the validity rather just just claiming the results.
Re: Re: Sound card
This is completely correct, but its not "tricks", it's simply better chip A/D's, that cost almost nothing. Modern A/D's are so good compared to ones of a decade or two ago that they are simply not a factor. The A/D in even the cheapest sound card is better than all the analog circuitry leading up to it. Things like the mic preamp and the mic itself. These things are far noisier and higher in distortion that the A/D on the sound card. If there is a "trick", it's the incredible job that has been done on A/Ds in chips. No other trick is required.
I gained a huge improvment with my measurements lately from the use of HOLMImpulse. This is because of the "technique" and nothing else. I used to use noise as it was the only thing that could get me the data that I wanted to see. And yes, I used to use external high cost sound cards. They made no difference since the resolution of the measurements was limited by the technique of using noise and cross correlation. I could not get data clean enough to allow for less than 1/3 octave averaging. With the new HOLM sotware, I get very good data at 1/6 octave, but it appears to be uncertain at 1/12. And again, the sound card makes no difference what-so-ever, its all in the technique.
With HOLM, I can now see that I have a pertty big problem with my "fixturing". I believe that there are some resonances in it that are affecting the data. These are the kinds of things that influence good measurements, NOT the sound card. To worry about the sound card simply means that you don't understand the accuracy limitations of real acoustic measurements.
The sound card MAY not be good enough for electronic measurments of amps and such, but I really don't know as I haven't tried it. It may bell be good enough even there, I would not doubt it for a minute.
pjpoes said:
This is completely correct, but its not "tricks", it's simply better chip A/D's, that cost almost nothing. Modern A/D's are so good compared to ones of a decade or two ago that they are simply not a factor. The A/D in even the cheapest sound card is better than all the analog circuitry leading up to it. Things like the mic preamp and the mic itself. These things are far noisier and higher in distortion that the A/D on the sound card. If there is a "trick", it's the incredible job that has been done on A/Ds in chips. No other trick is required.
I gained a huge improvment with my measurements lately from the use of HOLMImpulse. This is because of the "technique" and nothing else. I used to use noise as it was the only thing that could get me the data that I wanted to see. And yes, I used to use external high cost sound cards. They made no difference since the resolution of the measurements was limited by the technique of using noise and cross correlation. I could not get data clean enough to allow for less than 1/3 octave averaging. With the new HOLM sotware, I get very good data at 1/6 octave, but it appears to be uncertain at 1/12. And again, the sound card makes no difference what-so-ever, its all in the technique.
With HOLM, I can now see that I have a pertty big problem with my "fixturing". I believe that there are some resonances in it that are affecting the data. These are the kinds of things that influence good measurements, NOT the sound card. To worry about the sound card simply means that you don't understand the accuracy limitations of real acoustic measurements.
The sound card MAY not be good enough for electronic measurments of amps and such, but I really don't know as I haven't tried it. It may bell be good enough even there, I would not doubt it for a minute.
I have a 15 year old HP digital storage oscilliscope, I get less noisy measurements with my Realtek soundcard and oscilliscope software than I do on that 20,000 dollar piece of test gear. And yes, it's the A/D converters that seem to make the difference. With an appropriate amplifier to lift the signal above the noise floor I find I can get very precise measurements of amplifier noise such as distortion. The noise floor issue exists whether I use an oscilliscope, Agilent Distortion meter, or soundcard. The difference between the M-Audio and Realtek97 onboard for this purpose is so minor, and the "signal booster" I use brings it up with enough clean gain to overcome the noise floor of either. Yes the M-Audio is cleaner, but I need the signal well above the noise floor of either for accurate amplifier measurements anyway. In other words, they can get really good measurements of amplifiers and such too.
As for the noise you are having with the "fixturing", what do you mean by "fixturing". Is this the "fixture" that holds the mic, the speakers, etc? Is this noise in the system? I'm curious.
As for the noise you are having with the "fixturing", what do you mean by "fixturing". Is this the "fixture" that holds the mic, the speakers, etc? Is this noise in the system? I'm curious.
Yes for amps, peamps and that sort of thing a very low noise floor sure helps. I've seen very nice custom front ends built for prosumer soudcards that made for a wonderful measuring tool.
But for loudspeaker measurements? There is so much more noise ahead of the soundcard that it's not hard to find plenty good hardware.
But for loudspeaker measurements? There is so much more noise ahead of the soundcard that it's not hard to find plenty good hardware.
pjpoes said:
I believe that my test stand has resonances in it. I use a stand on which I place the systems and the top rotates. But I get some resonant "holes" in every test that I do regardless of the speaker and other don't seem to get this on the exact same drivers. To Wit Brendans recent testing. For some reason his data on the same woofers that I have data for looks a lot cleaner than mine. Now that could be two things, his data lacks resolution to detect what I detect, or my measurements have these resonances built into them. I am begining to assume the later might be the problem.
However, some years ago we did a correlation study between my data and some B&C woofers. My data was correct as there measurements, done the same way as I do them, showed the same problems. There data sheets don't show this however.
That was in Thailand, different setup, but similar. And my data correlated with that setup.
Anyone will tell you that correlating data can be a real problem.
You will not find anything of that sort on Rod Elliots website as that is not its purpose.
It would frighten away many people who want to do some diy audio, and its only function would be to demonstrate how clever the author is, Rod already knows that, that’s why he published it.
You did however make condescending remarks about the article and I just thought I would let you know that it has a very rigorous scientific and mathematical background and if you care to test the veracity of this I am glad to oblige.
Rcw.
It would frighten away many people who want to do some diy audio, and its only function would be to demonstrate how clever the author is, Rod already knows that, that’s why he published it.
You did however make condescending remarks about the article and I just thought I would let you know that it has a very rigorous scientific and mathematical background and if you care to test the veracity of this I am glad to oblige.
Rcw.
I don't recall making any condescending remarks about your paper, I don't recall making any remarks about it at all. In fact I have specifically avoided doing so.
I think that I have been asking for the "very rigorous scientific and mathematical" support for your statements. What you have stated, thus far, just does not make sense to me. I said that I do not see how such a complex mathematical development can be discussed in this kind of forum, so I completely understand your disinclination to do it here. However, it does not seem unreasonable for me to ask that you post, somewhere, this rigorous mathematic treatess so that we can all try and follow what it is you are saying.
I think that I have been asking for the "very rigorous scientific and mathematical" support for your statements. What you have stated, thus far, just does not make sense to me. I said that I do not see how such a complex mathematical development can be discussed in this kind of forum, so I completely understand your disinclination to do it here. However, it does not seem unreasonable for me to ask that you post, somewhere, this rigorous mathematic treatess so that we can all try and follow what it is you are saying.
pooge said:
No – its not about cone's or any other form of contours – its about "smoothness" of wave front shape transitions.
*If* you bend a wave front – this is what is called diffraction.
*If* your transition is rapid – like with OS of small throat dimension this "transition zone" of the OS converges towards a scattering junction.
Easy to see.
pooge said:
no risk - no fun – and no development in better understanding either...
Typical response from you. No matter what anyone says, you answer "no", even when what that person says is correct.
You continually argue about the wavefront when that is not the topic of conversation. That is no way to reach an "understanding".
You demand that everyone else proove their point with data. Talk about living in a glass house and throwing stones...
Thank you moderators for providing an ignore function.
You continually argue about the wavefront when that is not the topic of conversation. That is no way to reach an "understanding".
You demand that everyone else proove their point with data. Talk about living in a glass house and throwing stones...
Thank you moderators for providing an ignore function.