I have dealt with it.
Recording/Playback is information storage and retrieval system that is primarily a linear time invariant system.
Direct sound of speaker is primary focus. This includes primary effects of room where room may be considered part of speaker.
If listener wants system that is corrected across entire room, then a sufficient collection of measurements may be made for room related transfer function (RRTF), much as with HRTF, and used with head tracking system. As listener moves about in room, positional information is used for morphing between associated RRTFs.
Recording/Playback is information storage and retrieval system that is primarily a linear time invariant system.
Direct sound of speaker is primary focus. This includes primary effects of room where room may be considered part of speaker.
If listener wants system that is corrected across entire room, then a sufficient collection of measurements may be made for room related transfer function (RRTF), much as with HRTF, and used with head tracking system. As listener moves about in room, positional information is used for morphing between associated RRTFs.
Yes, may. We need more than a "may".
Nice idea but it doesn't exit. So we have to deal with what we have. This is the response of one speaker at my listening position within a 20x20cm cube:
A perfectly valid description from a "information transmission system" point of view. Does it show what we hear? Not so much...
Attachments
Last edited:
Why not apply some smoothing to those measures to better represent what we hear: "The ear tends to combine the sound within critical bandwidths, which are about 1/6 octave wide (historically thought to be 1/3 octave)." Music and the Human Ear
What does the impulse response look like?
Here are 6 measures of my right (digitally corrected) speaker around a 6' x 2' area at the listening position centered 10' away. Attached is an overlay with 1/6 octave smoothing applied. Also attached is an overlay of the 6 corresponding impulse responses.
The gold curve is the furthest measure from the speaker. Average response is about +- 5dB across the listening area.
To assist in the understanding of my results: Linear filter - Wikipedia, the free encyclopedia A few salient points:
"A linear time-invariant (LTI) filter can be uniquely specified by its impulse response h, and the output of any filter is mathematically expressed as the convolution of the input with that impulse response. The frequency response, given by the filter's transfer function
, is an alternative characterization of the filter."
With respect to IIR versus FIR filters: "Another advantage of FIR filters is that their impulse response can be made symmetric, which implies a response in the frequency domain that has zero phase at all frequencies (not considering a finite delay), which is absolutely impossible with any IIR filter."
What does the impulse response look like?
Here are 6 measures of my right (digitally corrected) speaker around a 6' x 2' area at the listening position centered 10' away. Attached is an overlay with 1/6 octave smoothing applied. Also attached is an overlay of the 6 corresponding impulse responses.
The gold curve is the furthest measure from the speaker. Average response is about +- 5dB across the listening area.
To assist in the understanding of my results: Linear filter - Wikipedia, the free encyclopedia A few salient points:
"A linear time-invariant (LTI) filter can be uniquely specified by its impulse response h, and the output of any filter is mathematically expressed as the convolution of the input with that impulse response. The frequency response, given by the filter's transfer function
With respect to IIR versus FIR filters: "Another advantage of FIR filters is that their impulse response can be made symmetric, which implies a response in the frequency domain that has zero phase at all frequencies (not considering a finite delay), which is absolutely impossible with any IIR filter."
Attachments
Yes, one can read that a lot but what does it even mean for two speakers in a room playing a stereo recording? And what about the time domain?
I am sure you can do your own research ;-) Already linked to James D. (jj) Johnston Home Page research on this thread, who arguably knows more about how our hearing, perceptual coding, and room correction works than most.
Specific to Acourate re: stereo and time domain, try searching IACC on the Acourate forum or ask Uli directly.
Specific to Acourate – solved 3 years ago: http://files.computeraudiophile.com/2013/1202/AcouratePRCen.pdf (PDF). Don't you have Acourate?
The acoustic measures I posted have no preringing. If anything, there is too much preringing compensation, and excess phase correction. Time to adjust and remeasure.
Are you going to post any impulse responses from the measurements you made?
mitchba, it should be clear by now that I know all that literature. Re-linking those papers don't answer what I've asked. If the answers would be there I wouldn't ask. It's not as simple as correcting a 1/6 octave smoothed response and expect to have removed all room/speaker distortion. For example, our ability to determine pitch is extraordinarily accurate despite the fact that our cochlea seems to represent just a coarse filter bank.
You've quoted an excerpt that states that symmetrical IRs are an advantage of FIR over IIR. That's what I was referring to. It's NOT an advantage. It's a disadvantage and can be detrimental. That's why Acourate has the ability to reduce pre-ringing.
Sure, what do you want to see?
Last edited:
Yes, may. We need more than a "may".
Nice idea but it doesn't exit. So we have to deal with what we have. This is the response of one speaker at my listening position within a 20x20cm cube:
A perfectly valid description from a "information transmission system" point of view. Does it show what we hear? Not so much...
I see a room in those measurements riddled with bad reflections and room modes. Typical situation for 99%. If you cannot separate the speaker from the room and address the room itself then the point of this mental exercise is what exactly? Does not matter if we read all the papers if we do not address the real issues. No amount of nit picking will solve the major issues when we are only talking about the finest details. Forest from the trees
^
Of course we have to separate the room from the source (speaker). That was the whole point of my graph. The problem is NOT just as one dimensional as fixing a "information transmission system" problem.
But even if we have separated room and source, what do we perceive? How do distortions of speaker and room influence our perception of sound (loudness, spectrum) and space (locatedness, AWS, LEV). This is at the core of the topic "room correction". If that sounds like nit picking to you then you're in the wrong thread because without answering those questions were just playing hit or miss.
Of course we have to separate the room from the source (speaker). That was the whole point of my graph. The problem is NOT just as one dimensional as fixing a "information transmission system" problem.
But even if we have separated room and source, what do we perceive? How do distortions of speaker and room influence our perception of sound (loudness, spectrum) and space (locatedness, AWS, LEV). This is at the core of the topic "room correction". If that sounds like nit picking to you then you're in the wrong thread because without answering those questions were just playing hit or miss.
Last edited:
The point is to address the room, once done go to finer granularity, not the other way around. Can't see how you will ever be happy without doing the research for yourself.
I have noticed many things that haven't been mentioned but few bits of literature scattered about here and there. Have noted these things and will research them for myself once a little further along in my project. These "things" were interesting acoustic effects and oddities observed over 20 years ago so yes there is plenty of work yet to be done. This thread sadly does not address any of them, so nothing to be learned here except how to not go about it. Why? You're making the assumption that the room would dominate everything and only changes in the room itself would improve things. You're limiting yourself by these assumptions.
You are assuming I am making uneducated assumptions and thus imposing self limitations unknowingly. Why do you repeatedly JACK people around with such conjecture?
You are assuming I am making uneducated assumptions and thus imposing self limitations unknowingly. Why do you repeatedly subject others to such conjecture at every turn?
You sir are trying to fenesse out the finest of needles in a hay stack of crud by ignoring the big picture while nit picking about the subtle nuances. Your use of low quality measurement equipment not designed for the task is appalling. Do you know anything other than sensitivity and response of your non traceable uncertified microphones? If you repeat with they are calibrated, I will laugh, as they would be considered inadmissable in court for even SPL measurements. I at least understand the limitations of the UMIK-1's I own and what calibration means having run a NIST traceable cal lab on hundreds of equipment of all shape and form.
Last edited:
Yes, may. We need more than a "may".
Nice idea but it doesn't exit. So we have to deal with what we have. This is the response of one speaker at my listening position within a 20x20cm cube:
A perfectly valid description from a "information transmission system" point of view. Does it show what we hear? Not so much...
I'll clarify: substitute "is to" for "may". Just as speaker's have baffle step of decreasing frequency with increasing size, speaker is to be considered baffled be the room in which it is set up. Speaker, intended listening area, and effect of room on sound received at intended listening area.
Of course we have to separate to room from the speaker, which is identical to separating the speaker from the room.
Your measurement set for 20cm x 20cm x 20cm cube looks to have about 12 measurements. Windowing appears to be at least 16k samples for 44.1kHz or 48kHz sample rate, perhaps much higher. This is not what is heard, hearing is perception of continuously fluctuating air pressure across the eardrum, to which very good approximation is fluctuation of air pressure at opening of ear canal.
Mitchba proposes smoothing. This is broad in terms of how it is carried out. On surface this would be applying smoothing algorithm to frequency response data for a given window. Better to start with window/gating that shows direct response of speaker. This is readily compared to windowing/gating out direct response, leaving only that of room. Total response equals superposition of direct response and reflected response for system of speaker, receiver. This holds because system is dominantly linear and time invariant.
You already know primary effects of room on perceived sound: increased spaciousness, increased loudness, and changes in timbrel characteristics do to modal properties of room at receiver for given source location.
Windowing/gating for direct response that includes earliest reflections captures very much what the waveguide characteristics of human ear/head/torso system. Human hearing is most sensitive off of forward facing axis in direction close to human visual blind spot with eyes fixed forward. (Coincidence? I think not.) High frequency sensitivity for free field source is highest along this vector as pinna focuses sound into ear canal by reflection.
It is easy to have reflected sounds of higher intensity than the direct sound. This works both for reflections from pinna, and from room for room low frequency reflections have highly uniform behavior, as seen in your data set. At higher frequencies, room reflections from different points arrive at receiver at differing delay times, and superposition results in frequency dependent constructive and destructive interference, viewable in frequency domain as combing of response, and in time domain as delayed impulse responses of direct sound.
Your blanket statements about FIR filter ringing is ill conceived. Transfer function of transmission system is all pass filter applied by convolution to continuous signal. For FIR system pass band for reflection free system is symmetrical. For FIR based crossover, all pass sum is Dirac function, exactly the same as first order IIR system.
I would suggest to those who can't treat the room but have two free corners to use a corner loaded horn. With the right horn the vertical dispersion is limited to 30 degrees and the horizontal to 45 degrees. And the response is very uniform down to schroeder. There's some beaming in the very highs obviously, but that's not much of importance.
Can be built as a two-way or a three-way. Far better then a speaker that has a poor off-axis response where one is trying to correct it with room correction. Again, EQ will not work very well as there is no 1:1 correlation between phase and amplitude.
Can be built as a two-way or a three-way. Far better then a speaker that has a poor off-axis response where one is trying to correct it with room correction. Again, EQ will not work very well as there is no 1:1 correlation between phase and amplitude.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.