There have been many discussions for the past a few months about early diffractions and reflections, some of which were based on earlier studies. We have basically come to an understanding that
Diffractions / reflections arrived within the first 10ms are perceived as part of the original sound therefore cause damage to the original sound and are best avoided.
Diffractions / reflections arrived within the first 2ms are most harmful.
Diffractions / reflections arrived from 10ms to 50ms may be desirable as they artificially create hall effect that simulates live music experience.
But I feel like that we have not had the full picture yet. I am making an assumption that the above theory may only apply to certain frequencies, not the entire audio bandwidth. Most likely it applies to the most critical frequency range, including 1kHz - 4kHz in which our ears are most sensitive.
For example, if we consider the success of Dr. Geddes's multiple sub approach, we would know that there is no way at low frequencies it obeys the 10ms window rule, as our ears are far less sensitive at low frequencies. For 100Hz one cycle is already 10ms. From my memory Dr. Geddes uses 500ms window for low frequency measurements, and that says something.
So to complete the whole picture for our understanding we need to derive a function of the "diffraction window" vs frequencies. Of course, this is based on my proposition that there is such a relationship between the diffraction window and frequencies, and it is likely that the window expands towards lower frequencies.
Note that this is an important piece of knowledge we should acquire. It is the basis to support the pursuit of constant / controlled directivity, low frequency management, etc.
I am not equipped with the knowledge, time, funding to conduct any such scientific studies, but would love to hear the thoughts from the experts and professionals.
Regards,
Bill
Diffractions / reflections arrived within the first 10ms are perceived as part of the original sound therefore cause damage to the original sound and are best avoided.
Diffractions / reflections arrived within the first 2ms are most harmful.
Diffractions / reflections arrived from 10ms to 50ms may be desirable as they artificially create hall effect that simulates live music experience.
But I feel like that we have not had the full picture yet. I am making an assumption that the above theory may only apply to certain frequencies, not the entire audio bandwidth. Most likely it applies to the most critical frequency range, including 1kHz - 4kHz in which our ears are most sensitive.
For example, if we consider the success of Dr. Geddes's multiple sub approach, we would know that there is no way at low frequencies it obeys the 10ms window rule, as our ears are far less sensitive at low frequencies. For 100Hz one cycle is already 10ms. From my memory Dr. Geddes uses 500ms window for low frequency measurements, and that says something.
So to complete the whole picture for our understanding we need to derive a function of the "diffraction window" vs frequencies. Of course, this is based on my proposition that there is such a relationship between the diffraction window and frequencies, and it is likely that the window expands towards lower frequencies.
Note that this is an important piece of knowledge we should acquire. It is the basis to support the pursuit of constant / controlled directivity, low frequency management, etc.
I am not equipped with the knowledge, time, funding to conduct any such scientific studies, but would love to hear the thoughts from the experts and professionals.
Regards,
Bill
I raised this because I am facing a new challenge.
I bought my house in probably the last such purchase / change of house for my life, as I am heading to 50's in a few years time. I have now dedicated a room for music. Given that I will unlikely be moving again, I can design and build my (last) pair of speakers for the room.
This is in contrast to most speakers that are designed for perfect response in an anechoic chamber, which may perform badly when room acoustic is combined. For me, I would like to design the room and speakers as one system.
I have been doing some MLS measurements of the tweeters, mid-woofers, woofers and sub-woofers in the room and been amazed to find different graphs while expanding the gated window 1ms by 1ms. It was a revelation.
Based on the theory of 10ms window, for frequencies between 1kHz and 4kHz, I guess the XO design should be based on the measurement using a 6-ms window. My tweeters are crossed at 2kHz. The response should be targeted to that of the anechoic chamber, with no floor, ceiling or wall reflections included. Although the floor and ceiling bounces come within 5ms, I have sufficient acoustic damping to counter the problems.
But for the bass with the XO point at around 180Hz, I suspect that the 10ms window theory does not apply, not to say the data points are insufficient.
I also believe that the XO / EQ should include the near boundary reflections, including floor, ceiling and front wall reflections.
I am struggling on what gated measurements I should use for the design of the low XO / EQ. An answer given to the previous post would provide an answer to my questions here.
I bought my house in probably the last such purchase / change of house for my life, as I am heading to 50's in a few years time. I have now dedicated a room for music. Given that I will unlikely be moving again, I can design and build my (last) pair of speakers for the room.
This is in contrast to most speakers that are designed for perfect response in an anechoic chamber, which may perform badly when room acoustic is combined. For me, I would like to design the room and speakers as one system.
I have been doing some MLS measurements of the tweeters, mid-woofers, woofers and sub-woofers in the room and been amazed to find different graphs while expanding the gated window 1ms by 1ms. It was a revelation.
Based on the theory of 10ms window, for frequencies between 1kHz and 4kHz, I guess the XO design should be based on the measurement using a 6-ms window. My tweeters are crossed at 2kHz. The response should be targeted to that of the anechoic chamber, with no floor, ceiling or wall reflections included. Although the floor and ceiling bounces come within 5ms, I have sufficient acoustic damping to counter the problems.
But for the bass with the XO point at around 180Hz, I suspect that the 10ms window theory does not apply, not to say the data points are insufficient.
I also believe that the XO / EQ should include the near boundary reflections, including floor, ceiling and front wall reflections.
I am struggling on what gated measurements I should use for the design of the low XO / EQ. An answer given to the previous post would provide an answer to my questions here.
HiFiNutNut said:
I understood things other way. Reflections in 50ms interval for musical signal are suppressed by precedence efect and contribute to loudness and timbre rather than hall effect. Sense of space is a matter of delays in the +80ms range with adequate level - probably stronger than you can reach in domestic conditions. Beside it is lower frequency phenomenon.
Reflections under 10ms for 1KHz+ range are:
- not necessarily harmful, probably pleasant because of contributing to ASW (according Toole)
- harmful because they do contribute nothing and can smear exact localisation. (Geddes)
I agree with Bill's data for the most part.
And yes Floyd and I disagree on the first 10 ms. Floyd's data is based on subjective preference data and yes <10 ms. sound adds to spatiousness, but it also adds coloration (confirmed in discussions with Blauert). So it depends on what you are looking for. You can have spatiousness with VER, but you have to accept coloration and lower image resolution. On the other hand, my approach is to minimize the VER for good image and low coloration while maximizing the >10 ms. reflections for good spatiuosness. This, I claim, achieves the best of both worlds, but requires a more complex system (speaker and room) design to achieve.
As regards VER with frequency, it is clear that VER at VLF are benign. We basically hear in the steady state at LF. Below 500 Hz our perception of reflections falls until below 100 Hz in a small room its all steady state percpetion with no awarness of reflections at all.
VER below 10 ms and above 500 Hz are not desireble IMO, but above 20 ms are clearly positive. 10-20 ms. is unclear. < 2 ms. is certainly bad for imaging above 1 kHz. but it can also lead to nonlinear types of sound effects.
And yes Floyd and I disagree on the first 10 ms. Floyd's data is based on subjective preference data and yes <10 ms. sound adds to spatiousness, but it also adds coloration (confirmed in discussions with Blauert). So it depends on what you are looking for. You can have spatiousness with VER, but you have to accept coloration and lower image resolution. On the other hand, my approach is to minimize the VER for good image and low coloration while maximizing the >10 ms. reflections for good spatiuosness. This, I claim, achieves the best of both worlds, but requires a more complex system (speaker and room) design to achieve.
As regards VER with frequency, it is clear that VER at VLF are benign. We basically hear in the steady state at LF. Below 500 Hz our perception of reflections falls until below 100 Hz in a small room its all steady state percpetion with no awarness of reflections at all.
VER below 10 ms and above 500 Hz are not desireble IMO, but above 20 ms are clearly positive. 10-20 ms. is unclear. < 2 ms. is certainly bad for imaging above 1 kHz. but it can also lead to nonlinear types of sound effects.
Member
Joined 2003
Recommended reading before you commit to anything permanent: "Sound Reproduction - Loudspeakers and Rooms" by Floyd Toole. One personal observation though, Floyd is somewhat ambiguous on the topic of very early reflections...like the comment "not necessarily harmful". He is not ambiguous on many other subjects, like direction of the reflections, etc.
Excellent book, very easy read. You might even change your mind about stereo vs multi-channel.
Excellent book, very easy read. You might even change your mind about stereo vs multi-channel.
Floyd was ambiguous about VER in his book, but in long conversations with him I nailed him down on the subject. He does say that VER "appear" (from the limited data) to be "prefered" to not having them. He also found this surprising since it is not what he expected either. However, the tests were such that what I recommend was never tried. He looked at situations with and without VER and found a preference for them, but that preference was based on a preference between the various alternatives and not all the alternatives were represented. In Floyd's style of tests it's very hard to control all the variables. For example, his directivity tests were based on a narrow directivity Quad speaker versus other wider directivity sources. But that's as much a test between the Quad's over-all sound quality and the other speaker as it is about directional versus non-directional. It's hard on some of this stuff to draw conclusions from the tests because so many variables have been changed.
His book is excellent and the only place where Floyd and I disgree is in the first 10 ms. Other than that we are completely on the same page. And in this 10 ms region, I find his data and his conclusions to be somewhat weak, certainly weaker than all else in the book.
I also find his fixation on multi-channel disappointing because I don't think that two channel is going away any time soon.
His book is excellent and the only place where Floyd and I disgree is in the first 10 ms. Other than that we are completely on the same page. And in this 10 ms region, I find his data and his conclusions to be somewhat weak, certainly weaker than all else in the book.
I also find his fixation on multi-channel disappointing because I don't think that two channel is going away any time soon.
Member
Joined 2003
Yes, I'm doing multi-channel too, but the lack of source material is a problem. Almost nothing that I listen to is in multi-channel. And even some of that is poorly mixed. It will be a decade before a wide array of quality source is available. And if current trends to music distribution and playback continue maybe even longer than that. I see movis and live concert playback going multi fairly quickly, I mean they already have, but daily music for the mass market may always be 2 channel.
Partially agreed, and certainly at LFs, but above some frequency, I would guess about 1 kHz, it should be based on time. The ear actually responds more this way, sensing things sychronously at LF, i.e. cycles related, but more based on spectral intensity above 1 kHz, i.e. more like a time related sense.
Thanks, Dr. Geddes. I am always happy to read your text. Although this is not a function of the time window vs frequency (perhaps no one has done such tests), you have given the right answer to my questions clearing out half of the clouds in my mind.
Now based on that, if you were to design the bass XO (say, at or below 200Hz) and EQ for loudspeakers dedicated to a room, would you include all reflections, including floor, ceiling, side and front wall bounces, into your measurements then design your XO and EQ based on those measurements? And if so, what time window would you use?
Regards,
Bill
HiFiNutNut said:
Lets start at the extremes. At LFs, < transition frequency, to do a crossover independent of the room is pretty much pointless. It won't work like that once in the room. So I do the "setup" of everything below 200 Hz based on steady state measurements which have been sptially averaged. Try and get the smoothest spatial and spectral response form all sources.
Now at the other end it's easy to say that the crossover needs to be done based on the total polar response for a time window of not less than 10 ms, but should contain no room reflections. If the crossover is not below about 1 kHz then this is not a problem. Between 200 Hz and 1 kHz, its all up in the air, but thank goodness this is where our hearing is the least accute and seems to be pretty tolerant of errors and such. So I don't know the right answer for this region, but I do know that it doesn't matter as much either.
HiFiNutNut said:
Hi Bill,
Interesting approach I just started my MLS measurements but I am not familiar with the background off the measurement technique.
So I am not fully understanding what I am doing.
So my question I do not have a picture by the example of a 10msec (FFT)window or 500msec. Could you explain this more detailed for me so I can understand the discussion.
Regards, Helmuth
When MLS or any technique is used to find the impulse response, it can be "windowed". This means that the impulse response data is use only within a defined space of time usually starting just before the impulse and then lasting for some time, but usually before the first reflections (whcih can sometimes be hard to identify). Outside of this "window" the data is "assumed" to be zero as it would be if there were no reflections, i.e. "anechoic". This then yields the response of the source independent of the room. But this technique is only valid above some LF which depends on the length of the window. Shorter windows higher LF point of validity. No measure at VLF can be independent of the room.
gedlee said:
Thank you Dr geddes.
With in Hobbybox V6 I can choose the measuring points 256 and up
(256 X2 and so on).
With this low sample rate I can't measure low frequencies due the few samples 20Hz-20000Hz.
So I shout maybe limit the frequency range?
My question:To measure 10msec or 500msec long within a FFT-window, what shout I do I still not understand??
I did a water fall with 2048 samples that's the maximum that HBX supports for waterfall measurement.
I can than also see the result of the impulse.
This pulse.
I asked a very intelligent colleague how to understand this pulse.
He thought that the MLS uses a square wave to measure because all frequencies are in it.
And the impulse shout be a the result as this graphic seen from the side.
But that wasn't good he thought so he did not know. So I thought it is may be the average of the hole measurement?
So my question where is the impulse the result from.
I know that the shorter the time range of the pulses better the response.
I can't read German, so I can't be of much help. Perhaps that software can't do windowing, I don't know. Sounds to me like you need some fundamental background in signals and systems, like "What is an Impulse Response?" That discussion would go too far a stream of this topic I am affraid. Try downloading my book chapters and see if this helps.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.