Ron E said:
OK, thanks. FYI, I have had BSOD in some of my programs earlier on some systems. In those cases I tracked it down to some specific motherboards. I wrote a web page about this:
http://www.tolvan.com/blue_screen_Q_A.htm
In particular there is a link to a program (not by me) that tests the computer if the moherboard is faulty.
It is possible that you do not experience problems with any other programs and I realise that this appears to make it look like my programs are at fault. Last time I investigated this it turned out they were not. I frequently use a feature that it appears as most programs don't (floating point exceptions).
I am sorry if you feel I am using you as an unpaid beta tester. I presume it is because I hinted that I might use whatever comes out of this in another simulation program that possibly not will be available for free.
I will, however *never* request any payment for The Edge, and actually I feel as if I am doing the community some service by posting this program. My the interest for previous version diffract.exe stunned me, so I thought that you all would be interested in a version that was a *bit* more user friendly. After all, the diffract.exe was only a hack to test a method that I read about. I am sorry about the bugs and quirks of The Edge, but I think it is better than the previous version, anyway.
OK, thank you everyone who has provided input regarding my baffle step simulator, "The Edge". I have now put version 1.1 on the website. It has these new features:
-In addition to circular drivers, it can now also simulate elliptical, square and rectangular membrane shapes.
-Multiple response curves can be shown for comparison. The piece of ice on the toolbar freezes the current curve and new ones can be drawn on top of it.
-Redrawing is done without recalculation if possible. This speeds things up a bit.
-Redrawing can be shut off. For systems with many sources, the automatic redrawing was quite annoying.
A few of you seems to have experiences blue screens. I have not done anything to improve on this, and I think I cannot. I am sorry about this. I have had this problem with other programs of mine, and then I tracked it down to certain combinations of hardware and OS'es.
Back then, I wrote a web page on the topic:
http://www.tolvan.com/blue_screen_Q_A.htm
Thank you again for your input.
Download The Edge
-In addition to circular drivers, it can now also simulate elliptical, square and rectangular membrane shapes.
-Multiple response curves can be shown for comparison. The piece of ice on the toolbar freezes the current curve and new ones can be drawn on top of it.
-Redrawing is done without recalculation if possible. This speeds things up a bit.
-Redrawing can be shut off. For systems with many sources, the automatic redrawing was quite annoying.
A few of you seems to have experiences blue screens. I have not done anything to improve on this, and I think I cannot. I am sorry about this. I have had this problem with other programs of mine, and then I tracked it down to certain combinations of hardware and OS'es.
Back then, I wrote a web page on the topic:
http://www.tolvan.com/blue_screen_Q_A.htm
Thank you again for your input.
Download The Edge
An externally hosted image should be here but it was not working when we last tested it.
Vikash said:
Hmm... This is a hardware combination that I have not seen cause it before.
If you have the time, could you try this program:
http://www.econos.de/fpu
either of the two versions is ok.
It is not written by me, and tests the system for a bug. It would help me a lot. If this program does *not* crash your computer, I'll have to start digging.
Thanks alot!
Svante said:
And now I have to change my mind again and confess that the Edge has a bug; the phase comes out with the wrong sign, when exporting the response. The amplitude is right, though. This will be fixed in the next version and also visible on the Egde screen. Here is a preview:
An externally hosted image should be here but it was not working when we last tested it.
Obviously the red (compensated) phase response is better than the green (not compensated). And more important, so are the amplitude responses, and *that* is what is important about baffle step compensation.
Svante, could you tell us a bit about how the program works? From what I can guess, each point around the perimeter of the baffle re-radiates and attenuates the sound, and the result is a summation of hundreds of signals with different amplitudes and phases. (If I'm completely wrong here, please tell me.)
My question is: what does each "edge point" represent? Do they represent 90 degree corners (like an infinite extrusion) or 180 degree corners (like a flat cut-out)? What gets calculated when an edge point sits on a "corner" so the radiation resistance drops by 1.75 times, or some other amount depending on the actual angle made by the creative DIYer?
Ideally, a way to model the enclusures in 3D would be good
.
CM
My question is: what does each "edge point" represent? Do they represent 90 degree corners (like an infinite extrusion) or 180 degree corners (like a flat cut-out)? What gets calculated when an edge point sits on a "corner" so the radiation resistance drops by 1.75 times, or some other amount depending on the actual angle made by the creative DIYer?
Ideally, a way to model the enclusures in 3D would be good
.CM
Look for papers by Bews and Hawksford, or Vanderkooy in the JAES.
You are essentially correct in your guesses. If no edge angle is specified, I would assume it represents a sharp 90 degree edge and that the side panel extends back infinitely, such that the reradiation from the baffle edge is all there is....
Curved edges don't actually reduce diffraction, they merely distribute it over the whole curved surface.
You are essentially correct in your guesses. If no edge angle is specified, I would assume it represents a sharp 90 degree edge and that the side panel extends back infinitely, such that the reradiation from the baffle edge is all there is....
Curved edges don't actually reduce diffraction, they merely distribute it over the whole curved surface.
CeramicMan said:Svante, could you tell us a bit about how the program works? From what I can guess, each point around the perimeter of the baffle re-radiates and attenuates the sound, and the result is a summation of hundreds of signals with different amplitudes and phases. (If I'm completely wrong here, please tell me.)
My question is: what does each "edge point" represent? Do they represent 90 degree corners (like an infinite extrusion) or 180 degree corners (like a flat cut-out)? What gets calculated when an edge point sits on a "corner" so the radiation resistance drops by 1.75 times, or some other amount depending on the actual angle made by the creative DIYer?
Ideally, a way to model the enclusures in 3D would be good
CM
It assumes a 180 degree corner, so it is really a better open baffle simulator than it is a box simulator. However, the difference from a 90 degree edge is probably small. A 90 degree edge would have a more gradual increase from half-space to full space as I understand it. However, there will be radiation in full space at sufficiently low frequencies and so the radiation resistance will drop 2 times, even if that is not how The Edge models it.
I have tried to describe what the program does on the home page, did you find that?
What interests me is a computer model of diffraction, that could also show resonances caused by changes in acoustic-radiation impedance. AFAICT the ripples in sensitivity shown by The Edge are relatively benign, because they are basically early reflections/echoes as opposed to ringing. Although it shows that diffraction effects cause peaks and troughs in the amplitude response, it doesn't take into account recursive interactions between point sources.
I wonder if a solution to this could be possible by giving the speaker point-sources some theoretical T/S parameters, so that reflections from diffraction could slightly influence each source in the same way that a real cone would be influenced? For example, each speaker point-source could generate an impulse (or whatever generic test signal) as before, but that impulse would be coloured by variations in pressure from diffraction, recursively.
Heck, even that isn't needed for a basic diffraction model with ringing. The point-sources around the perimeter can be modelled with ringing even if the speaker sources aren't involved. Take 2 parallel sides for example: reflections will travel from side to side, and each time some frequencies will be reinforced or attenuated more than others. A time window would have to be set - just like with real measurements - so the calculation doesn't take forever. What the microphone would then pick up would change with time, and could provide valuable step-response data and waterfall plots.
...And if that sounds too easy, the speaker cone could be modelled so it's not just a perfect piston. Now, where did I put that Pentium 4 cluster?
CM
I wonder if a solution to this could be possible by giving the speaker point-sources some theoretical T/S parameters, so that reflections from diffraction could slightly influence each source in the same way that a real cone would be influenced? For example, each speaker point-source could generate an impulse (or whatever generic test signal) as before, but that impulse would be coloured by variations in pressure from diffraction, recursively.
Heck, even that isn't needed for a basic diffraction model with ringing. The point-sources around the perimeter can be modelled with ringing even if the speaker sources aren't involved. Take 2 parallel sides for example: reflections will travel from side to side, and each time some frequencies will be reinforced or attenuated more than others. A time window would have to be set - just like with real measurements - so the calculation doesn't take forever. What the microphone would then pick up would change with time, and could provide valuable step-response data and waterfall plots.
...And if that sounds too easy, the speaker cone could be modelled so it's not just a perfect piston. Now, where did I put that Pentium 4 cluster?
CM
I'm speculating here, but I think that higher order reflections drop in amplitude pretty quickly. In the real measurements I've seen, there is sometimes a second order reflection, but very rarely a higher order than so.
I don't think the way to solve this is to add a transfer function to the edge sources. The only thing I think should be added is directivity, and that and the amplitude should be independent of frequency.
But as I said, I am on sort of thin ice here.
I don't think the way to solve this is to add a transfer function to the edge sources. The only thing I think should be added is directivity, and that and the amplitude should be independent of frequency.
But as I said, I am on sort of thin ice here.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.