Frequency resolution from time data of 5 ms can't be made better than 200Hz. I think that we all can finally agree on that. Sure, if the loudspeaker response is so smooth it doesn't need higher resolution than this to be fully described, i.e. it's IR decays completely within these 5 ms, there's nothing more to do - you have all the data (and resolution), can't be more happy.
Now, what if the IR doesn't decay within the refection-free time? Here it becomes tricky. Is there any really reliable indication whether this time interval is sufficient to give "good enough" data? How you exactly recognize that the response has decayed sufficiently, so you can claim that the frequency response will be identical as in anechoic chamber? I think this is now the reasonable question. Earl seeems to know something about this. Could you please elaborate?
Now, what if the IR doesn't decay within the refection-free time? Here it becomes tricky. Is there any really reliable indication whether this time interval is sufficient to give "good enough" data? How you exactly recognize that the response has decayed sufficiently, so you can claim that the frequency response will be identical as in anechoic chamber? I think this is now the reasonable question. Earl seeems to know something about this. Could you please elaborate?
Last edited:
I answered this, and made the same points that you are making, before. The impulse ringing decays exponentially - and is guaranteed to do so monotonically - so the benefits of a longer time window are logarithmic. Just look at the impulse response and you can see if it has decayed to whatever level you deem necessary. As far as I am concerned if it is not visible in the trace for say 1 ms. then that's enough. Every counter example that has been shown in this thread has an obvious tail at the window truncation - that's wrong, don't do that!!
Do people out there actually do stupid things like that - of course they do. Most of us here would, or should, know better.
Do people out there actually do stupid things like that - of course they do. Most of us here would, or should, know better.
Now if you want my opinion (which I am sure you don't), this technique is completely wrong (trivial to implement sure, but wrong.)
While most speakers are "virtually" minimum phase many of their most insidious problems are not. When the DUT has excess group delay then sliding the window will actually chop off these effects, not just smooth them, when they are real issues within the speaker itself.
I won't even use this technique in HOLM because it is so fraught with errors.
Please see http://www.diyaudio.com/forums/mult...quency-response-resolution-3.html#post3642705 and move discussions there.
But no, I would not agree.
But no, I would not agree.
Oh c'mon, don't be so high maintenance.
Frequency dependent windowing is used in Acourate to obtain a curve for room correction. In that context I believe the approach is correct. Not sure about the actual numbers though.
The discussions about measurements are definitely worthwhile, and I'm glad to see they have been given their own thread.
Back on topic, one of the things mentioned recently in a related thread was the difference between round horns and those with elliptical or rectangular mouths having shorter vertical dimension. I'd like to bring that into focus, here.
Of course, axisymmetrical horns have the same pattern in all directions. Asymmetrical horns do not. Where the dimension is smaller, the pattern control will be lost first. Some call this pattern flip, but it's really not that simple. It's directivity that changes at a different point in each dimension, and at a different rate. But in the trade-off, one gains the ability to have less vertical distance between sound sources, and this provides a taller forward lobe. It also allows for a smaller area expansion, which can make smoother response in some cases. So these are worthwhile traits, in my opinion.
Of course, vertical spacing is not the only factor that sets the distance between vertical nulls. It is also a function of crossover frequency. However, if two horns are compared, both having the same horizontal dimension and flare profile, then they will have approximately the same horizontal directivity. Therefore, if the goal is to match directivity in the horizontal plane between a direct radiating midwoofer and the tweeter horn, then crossover will be the same frequency whether the horn is round or elliptical/rectangular. The only difference is in the response along movement in the vertical dimension. The loudspeaker system with closer vertical spacing will have a taller clean forward lobe through the crossover region, with the vertical nulls and secondary lobes spaced further apart.
Back on topic, one of the things mentioned recently in a related thread was the difference between round horns and those with elliptical or rectangular mouths having shorter vertical dimension. I'd like to bring that into focus, here.
Of course, axisymmetrical horns have the same pattern in all directions. Asymmetrical horns do not. Where the dimension is smaller, the pattern control will be lost first. Some call this pattern flip, but it's really not that simple. It's directivity that changes at a different point in each dimension, and at a different rate. But in the trade-off, one gains the ability to have less vertical distance between sound sources, and this provides a taller forward lobe. It also allows for a smaller area expansion, which can make smoother response in some cases. So these are worthwhile traits, in my opinion.
Of course, vertical spacing is not the only factor that sets the distance between vertical nulls. It is also a function of crossover frequency. However, if two horns are compared, both having the same horizontal dimension and flare profile, then they will have approximately the same horizontal directivity. Therefore, if the goal is to match directivity in the horizontal plane between a direct radiating midwoofer and the tweeter horn, then crossover will be the same frequency whether the horn is round or elliptical/rectangular. The only difference is in the response along movement in the vertical dimension. The loudspeaker system with closer vertical spacing will have a taller clean forward lobe through the crossover region, with the vertical nulls and secondary lobes spaced further apart.
- Measurement of vertical nulls (video)
Since "room correction" itself is "wrong" what difference could it make? I know that a sliding window is wrong for the direct field measurement.
I strongly disagree.
Just tossing out some thoughts because the whole topic is highly complex and broad:
Our ears aren't microphones. You've said it yourself, our hearing works in bands, it's not aware of all resonances. The physical part of our hearing seems to reduce information (pressure input at the ears) whereas higher brain functions seem to be the place where the magic happens (e.g. backward and forward masking).
Stereo reproduction is listening in an interference field. Everything that reduces ambiguity has to be beneficial.
What we hear at low frequencies (I'm not just talking about <200Hz) is dominated by the room. This doesn't mean that we couldn't locate sounds at low frequencies. It's just means that the room imposes a filter which delivers ambiguous cues. Again, anything that reduces ambiguity has to be beneficial.
Last edited:
Markus, this is an old discussion.
If you mean EQ in the modal range then sure that makes sense and maybe a sliding window might be some benefit there although I would still contend that steady state is what we hear in the modal range.
But what we want above the modal range is a "flatish" direct field and a reverberant field that is not too dissimilar. If the speakers are CD and the direct field is "flatish" then there is nothing that a room EQ could do but degrade this situation.
If you don't have CD speakers and/or if they aren't flatish direct field, then maybe room EQ can help. I don't have speakers like that so its not an issue for me.
If you mean EQ in the modal range then sure that makes sense and maybe a sliding window might be some benefit there although I would still contend that steady state is what we hear in the modal range.
But what we want above the modal range is a "flatish" direct field and a reverberant field that is not too dissimilar. If the speakers are CD and the direct field is "flatish" then there is nothing that a room EQ could do but degrade this situation.
If you don't have CD speakers and/or if they aren't flatish direct field, then maybe room EQ can help. I don't have speakers like that so its not an issue for me.
Hello,
My answer willbe probably off topic but for myself in the Matlab routine I wrote and use to analyse impulse responses I am using a raised cosine (the advantage of which is to have only a real part, no complex part) the width of which is 12 wavelengths (so varying with frequency). This window is set symetrical on the point of the IR where the power is maximum.
I choose that value of 12 after many time spent to compare measurements obtained in different environments and analysis conditions. (In fact in my Matlab routine I can choose whatever number of wavelengths but since many years I did'nt find the need to use anything else than 12 wavelengths...). Only with multiways speakers having very bad alignment of the acoustics center between the several loudspeakers, the method proved to be weak.
I am using this also for wavelet reconstruction of an impulse response from a simulated frequency response (magnitude and phase). (e.g. as in the IR module Iwrote for David McBean which is included in Hornresp)
Best regards from Paris,
Jean-Michel Le Cléac'h
Last edited:
Attachments
There's a large region around the Schroeder frequency which is not fully dominated by steady state nor direct sound. It's also the region where interchannel amplitude cues are tranformed into interaural timing cues.
The exact role of direct and indirect sound is unknown. The sound field also varies greatly with room and speaker. If the indirect sound field is soft the speaker's direct response dominates. If it's the other way around, the diffuse field dominates and equalizing the steady state response becomes more appropriate again.
I believe the most important region is the frequency range where your speakers are no longer CD. The room filter creates ambiguous cues for stereo reproduction.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.