Software would be great, Earl. I am trying to understand the math behind this to the best of my ability (though I minored in Physics in college, some 30+ years ago, I am just a medical doc with this as a hobby and my obsessionality as my driver). I would love to see your software, when you make it available, Earl, thanks. And thanks again, for your input and guidance. Jay
I'm curious to know which ARTA measurements you have an issue with, I'm presuming it's the directivity sonogram, since directivity is your hobbyhorse ?
Are you saying there is something bogus or incorrect about the way ARTA calculates some of its results, or just that they don't seem to match up with what your own software generates ?
ARTA is a pretty complex and powerful piece of software which does take a bit of skill and knowledge to use correctly, and like similar programs provides plenty of rope to hang yourself with if not well versed in its correct use.
I've seen plenty of measurements made with ARTA posted on this forum that are clearly incorrect or inaccurate due to incorrect settings such as excessive smoothing, (I normally stick to 1/24th octave for frequency response, although for polar responses you can only go up to a maximum resolution of 1/12th octave) poor choice of start/end markers on the impulse letting reflections into the measurement, poor choice of time or amplitude scale for CSD and so on.
If its measurements taken by others that look "too good" to you then it could well be misuse of the software. On the other hand if you have used ARTA yourself in direct comparision to your own software and you see serious divergence in the results there could well be a problem that deserves investigating.
Do you have any specific examples ?
Last edited:
Hi Simon
No, I have never used ARTA and yes, it is other peoples examples that I take issue with. I have tried to get a set of common measurements displayed with both systems to compare, but have been unable to do that. I would still love to compare identical systems measured at the same time but shown in the two different pieces of software.
Several problems ensue of course, like the fact that my smoothing is done in critical bands not octaves so naturally there is going to be a difference there (Zwicker or Moore is an option, Moore's CB are much narrower in general, I can also do fractional octaves, but critical bands is the better way). Prof. Farina showed results indicating a strong preference for systems smoothed with critical bands over ones smoothed in fractional octaves. There are other proprietary things that I do but none of them as far as I can tell account for the large differences that I see in what I do and what is see posted. But of course, if the systems are different then there are going to be large differences, which is what makes it hard to come to any conclusions.
No, I have never used ARTA and yes, it is other peoples examples that I take issue with. I have tried to get a set of common measurements displayed with both systems to compare, but have been unable to do that. I would still love to compare identical systems measured at the same time but shown in the two different pieces of software.
Several problems ensue of course, like the fact that my smoothing is done in critical bands not octaves so naturally there is going to be a difference there (Zwicker or Moore is an option, Moore's CB are much narrower in general, I can also do fractional octaves, but critical bands is the better way). Prof. Farina showed results indicating a strong preference for systems smoothed with critical bands over ones smoothed in fractional octaves. There are other proprietary things that I do but none of them as far as I can tell account for the large differences that I see in what I do and what is see posted. But of course, if the systems are different then there are going to be large differences, which is what makes it hard to come to any conclusions.
Are you familiar with sound radiation from a sphere? Its not usually stuff taught at the undergrad level, but perhaps. It involves spherical harmonics, which are also used in quantum mechanics. Basically the entire shell concept in atoms are the same as the "modes" in sound radiation. They are both eigenfunctions of the spherical wave problem. It would make no sense to post any of this here as it would neither be easy (equation and the like) or very easy to follow. I can certainly refer you to the pertinent texts (my own or others).
I never studied acoustical engineering in school (that I can recall) but I have been attempting to teach myself little bits at a time. Not certain if I would be able to follow the math but I am certainly willing to try. With many things in nature, there are estimates that we can use to approximate things within reason and then there are ever improving formulae that get us closer to accuracy. I am thinking that there is some way of coming up with a better estimate of farfield than just a multiple of a dimension that anyone would be able to understand.
I guess what I am saying is that I am certainly willing to try to dig through the math and learn it but my true goal here is to understand some basic concepts here that could be practically used in setting up measurements.
As an aside, if you are looking for volunteers to measure drivers or speakers with your software and to compare it to ARTA or Soundeasy, I volunteer. I use both and typically get measurements with both that are in good agreement when I use them in similar environments.
Jay
I guess what I am saying is that I am certainly willing to try to dig through the math and learn it but my true goal here is to understand some basic concepts here that could be practically used in setting up measurements.
As an aside, if you are looking for volunteers to measure drivers or speakers with your software and to compare it to ARTA or Soundeasy, I volunteer. I use both and typically get measurements with both that are in good agreement when I use them in similar environments.
Jay
Hi Jay
I would say that you should not list errors for locations such that r is not much greater than the cone radius. Thats because the errors could be much higher than what the simplified formula predicts. To me "much greater than" would be 4 or 5 times, although usually "much greater than" is assumed to mean 10 times.
Doing an exact numerical integration of the functions would not be too hard (not doable in Excel) and maybe I could do your data more exactly in say MathCAD.
To do a comparison of measurement software, I think that a narrow directivity speaker should be used because those will exhibit the greatest problems. My software assumes HolmImpulse as the data source, but I could probably read in any data files if I knew the format. But since Holm is easy and free, it seems like a good choice. I prefer angles of 0,5,10,15,20 30,40,50,60,80,100,120,150 and 180 as these are optimized for data resolution in the front. Equal angles are not ideal and angles less than 5 degress in the forward direction are too coarse. Any way that you could do this? Kind of a complicated problem actually, which is why it hasn't happened to date.
Hi Earl,
I have built a jig that will allow for angle measurements. I can adjust the angles but my degree readings may be a degree or two off as it was really roughed for basic power measurement estimates using multiple reads, finding an average, and designing my crossovers based upon this. DLR did this and I opted to compare crossovers designed that way to straight on measurements. I found that the room signature really dicatates which crossover I like better.
Regarding a high directivity speaker, though, I have not built any horns or angle controlled speakers to use. I may be able to find one, though.
Thanks for the feedback on the spreadsheet. I can certainly cause any numbers lower than 4 or 5 x the radius to gray out.
I have not used MathCAD but I don't mind trying to learn it. I think that there is some free version of it available (it may work through IE). I believe that there are ways to do integration with Excel, I would have to research this further.
Again, thanks,
Jay
I have built a jig that will allow for angle measurements. I can adjust the angles but my degree readings may be a degree or two off as it was really roughed for basic power measurement estimates using multiple reads, finding an average, and designing my crossovers based upon this. DLR did this and I opted to compare crossovers designed that way to straight on measurements. I found that the room signature really dicatates which crossover I like better.
Regarding a high directivity speaker, though, I have not built any horns or angle controlled speakers to use. I may be able to find one, though.
Thanks for the feedback on the spreadsheet. I can certainly cause any numbers lower than 4 or 5 x the radius to gray out.
I have not used MathCAD but I don't mind trying to learn it. I think that there is some free version of it available (it may work through IE). I believe that there are ways to do integration with Excel, I would have to research this further.
Again, thanks,
Jay
The complex integration that has to be done (complex in the sense of real and imaginary) needs a fairly robust convergence technique like that in MathCAD or it won't be reliable.
I don't think MathCAD is free anywhere and certianly not since it was sold. Its very expensive these days (I bought version 2.0 for $49.95, upgrades weren't too bad, until it was sold and then they got ridiculous.)
Yes, graying out uncertain data is a very good idea.
Oh, and with my technique small angular erros are not critcal because the software fits in angle and it will tend to ignore small erros like those. It would certainly be a bigger issue with ARTA and the like.
I don't think MathCAD is free anywhere and certianly not since it was sold. Its very expensive these days (I bought version 2.0 for $49.95, upgrades weren't too bad, until it was sold and then they got ridiculous.)
Yes, graying out uncertain data is a very good idea.
Oh, and with my technique small angular erros are not critcal because the software fits in angle and it will tend to ignore small erros like those. It would certainly be a bigger issue with ARTA and the like.
Last edited:
MathCad was great when it used the Maple symbolic engine - for years after the switch following the sale you would still see heavyweights on the user forum asking to see worksheets posted in version 11
I don't know a good free replacement - Sage Math project has DOE/MacSyma symbolic engine and a new "sketch-pad" interface
for straight numerical work several free MatLab workalikes exist: SciLab is actively developed, Octave is reputed to be more directly MatLab compatable
NumPy/SciPy python is supposed to be good too - but you are then pretty explicitly writing programs rather than having lots already set up for you in the other packages
I don't know a good free replacement - Sage Math project has DOE/MacSyma symbolic engine and a new "sketch-pad" interface
for straight numerical work several free MatLab workalikes exist: SciLab is actively developed, Octave is reputed to be more directly MatLab compatable
NumPy/SciPy python is supposed to be good too - but you are then pretty explicitly writing programs rather than having lots already set up for you in the other packages
Actually what I now think is the best package is Mathmatica. It has a resonably priced "home" version and is really head and shoulders above anything else that I have seen. I have a copy, but alas, as with any good software the learning curve is steep. MAtLab is popular but its too numerical for me, I prefer the more symbolic approaches like Maple and Mathmatica. Maple was in MathCad up through version 14. When I showed my son it doing his complex algebra at the touch of a button he was blown away.
Actually, Excel can do complex numbers but not sure how robustly it can do integration. I have seen some evidence that it can do simple integrations but not certain how complex it can get.
Earl, you are correct, MathCad is not available in a free form. I thought that I had seen it somewhere. There are some on the net that suggest that Smath Studio (freeware) is excellent software that is like MathCad lite and can even do calculus. I have to evaluate it to see.
jcx, I will have to take a look at Sagemath and Scilab.
If you would like me to do measurement comparisons, I should be able to get to it within a week or two and I would be happy to do it.
Jay
Earl, you are correct, MathCad is not available in a free form. I thought that I had seen it somewhere. There are some on the net that suggest that Smath Studio (freeware) is excellent software that is like MathCad lite and can even do calculus. I have to evaluate it to see.
jcx, I will have to take a look at Sagemath and Scilab.
If you would like me to do measurement comparisons, I should be able to get to it within a week or two and I would be happy to do it.
Jay
Semi off topic
Mathcad, ugh.
While it might be expensive since they were bought out, a much bigger reason to avoid the new version is the company PTC that sells it now.
We had Mathcad back at intersonics starting at a very early version, later on, I bought it for home starting with version 7 and occasional upgrades throughout the years.
A couple years ago I decided to upgrade again to the latest version. To be honest, I can’t say I had used it long enough to notice where /how it was different than 2001 when my computer HD crashed.
This is where they departed from what I was used to from real software companies.
With the new version was an additional licensing / authorization stage and after re-building my computer and re-installing all the software, I came to MathCAD which would not register / authorize.
I contacted PTC and explained the problem with the computer and after a number of e-mails back and forth trying to get it re-installed, they stopped replying.
Apparently, you are allowed to UN-install it and transfer to another computer, but you are only permitted one installation / license. Bottom line, they would not allow me to re-authorize the software and I had to re-install MathCAD back to version 8 to get to to run without the CD in the drive.
In my last e-mail I explained to PTC that I thought this was really crummy product support and whenever it was convenient, I would tell others of my experience with them, hence this post, YMMV.
Best,
Tom
Mathcad, ugh.
While it might be expensive since they were bought out, a much bigger reason to avoid the new version is the company PTC that sells it now.
We had Mathcad back at intersonics starting at a very early version, later on, I bought it for home starting with version 7 and occasional upgrades throughout the years.
A couple years ago I decided to upgrade again to the latest version. To be honest, I can’t say I had used it long enough to notice where /how it was different than 2001 when my computer HD crashed.
This is where they departed from what I was used to from real software companies.
With the new version was an additional licensing / authorization stage and after re-building my computer and re-installing all the software, I came to MathCAD which would not register / authorize.
I contacted PTC and explained the problem with the computer and after a number of e-mails back and forth trying to get it re-installed, they stopped replying.
Apparently, you are allowed to UN-install it and transfer to another computer, but you are only permitted one installation / license. Bottom line, they would not allow me to re-authorize the software and I had to re-install MathCAD back to version 8 to get to to run without the CD in the drive.
In my last e-mail I explained to PTC that I thought this was really crummy product support and whenever it was convenient, I would tell others of my experience with them, hence this post, YMMV.
Best,
Tom
Just an update. I have now done measurements on several individual drivers in enclosures. I theorized that using a Bessel function to estimate the amount of effective reflection off of a baffle in combination with the trigonometric approach that Pat Brown (http://www.etcinc.us/tech/nl043_far_field_criteria.pdf) uses. My sense was that his approach overestimated the necessary farfield distance as it did not seem to take into account the loss of dispersion of sound as the frequency rises.
What I have found (and I will post my data and a new calculator that is consistent with my theory and my data) is that (this is not news) both the enclosure and driver play roles in the determination of farfield, but each contributes differently dependent upon the frequency range of interest.
One of the things that had always bothered me about Zaph's distortion measurements is that he (if I recall correctly) measured at 4 inches for tweeters and called this farfield. I do not believe this to generally be the case; certainly not from my measurements - realize, though, that there are different degrees of error that decrease as you measure further back and the definition of farfield really depends upon how much error you are willing to accept.
The calculator that I have developed (which, I am certain, also has some error associated with it as it is theoretical) using the Solver function in Excel. I am just trying to figure out how to make it so that you don't have to unlock the program, activate the solver function, and then relock the program. I think that I am pretty close in this regard.
As always, I welcome discussion and will welcome input once I can post this program.
One other comment is that my estimates do not seem to be as effective when wave guides are used. I suspect this is due to the change in beaming functions.
Jay
What I have found (and I will post my data and a new calculator that is consistent with my theory and my data) is that (this is not news) both the enclosure and driver play roles in the determination of farfield, but each contributes differently dependent upon the frequency range of interest.
One of the things that had always bothered me about Zaph's distortion measurements is that he (if I recall correctly) measured at 4 inches for tweeters and called this farfield. I do not believe this to generally be the case; certainly not from my measurements - realize, though, that there are different degrees of error that decrease as you measure further back and the definition of farfield really depends upon how much error you are willing to accept.
The calculator that I have developed (which, I am certain, also has some error associated with it as it is theoretical) using the Solver function in Excel. I am just trying to figure out how to make it so that you don't have to unlock the program, activate the solver function, and then relock the program. I think that I am pretty close in this regard.
As always, I welcome discussion and will welcome input once I can post this program.
One other comment is that my estimates do not seem to be as effective when wave guides are used. I suspect this is due to the change in beaming functions.
Jay
For a simplified yet experimentally verified version...I tested this with half a dozen drivers and various enclosures. Please feel free to comment, agree, disagree, or to grab a snack and ignore. Hope this is helpful. I appreciate all of the input along the way as my understanding is dramatically improved based upon your feedback.
Thanks,
Jay
Thanks,
Jay
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.