keith_correa
Well-Known Member
Those of a certain "vintage" may know who Ken is. For those who don't, Google is your friend!
This is a link to an interview given by him:
www.stereophile.com
And the following is what this post is about:
"OK, I was asked what I thought the major remaining problems in loudspeaker design were. Here is my answer.
-----
To borrow a concept from Artificial Intelligence, I divide the loudspeaker question into two domains: the "hard" problems and the "soft" problems. While I am personally most interested in the hard problems, let's start with the soft problems. These include things that, while they may be very difficult to accomplish, are easy to conceptualize; things like getting a proper frequency response on axis and in a room, controlling radiation pattern, impulse response, distortion, power handling, etc. It's important to note that my short list addresses what is left to be done, not what is most important in the big picture. For example, on-axis frequency response is very important to the sound, but it isn't a very interesting problem anymore. With good drivers and EQ/DSP, it's pretty easy to get flat axial response these days. So, here is my arbitary and non-comprehensive list of the top soft problems in loudspeaker design:
1- Independent control of axial and power response.
2- Radiation pattern control and the elimination of non-coincident sources.
3- Achieving a proper power response in a room without creating a cluttered impulse response within the fusion time.
4- Dealing with the fact that bass response is very dependent on the modulation envelope of the signal. (Bass response for signals with long-time-constant modulation envelopes is quite different from the bass response of short-time-constant modulation envelopes.) I believe this is a widely overlooked phenomenon.
5- Dealing with the variable transmission characteristics of room boundaries at low frequencies, and the attendant issue that deep bass in-room response typically takes between 10 and 30 seconds to stabilize.
Now, on to the "hard" problems, which concern the basic role of the loudspeaker in the record-reproduction chain. These are the ones that interest me, because these are the ones that will allow our hifi systems to truely approach virtual reality, instead of merely hinting at it.
1- Getting it through our collective heads that loudspeakers are not amplifiers and that transfer function metrics like frequency response and distortion are misapplied and a conceptual dead end. They do not appropriately characterize the system. Improving frequency response and distortion past a certain point does not yield further meaningful improvement in realism. (I know this is anathema to most audio people, but I stand by it 100%.)
2- A loudspeaker's sole job is to undo what a microphone does. (Or reproduce what an engineer/musician hears in the studio.) It is folly to suggest there is an ideal loudspeaker when every recording is made in a different way. The plethora of available "state-of-the-art" loudspeakers is a testament to this basic problem. Speaker design cannot progress without standardization of recording techniques. Good luck with that...
3- Signal transforms that reduce dimensionality are inherently irreversible. In other words, there are an infinite number of ways a speaker can reproduce a given microphone signal which are all mathematically correct. Information is lost in the recording process, and that information is gone forever. We are lucky that speakers sound as good as they do!
4- Reproducing a sound field in a listening room is a spatial sampling problem. Unless you have twice the number of loudspeakers as the shortest wavelength you want to reproduce requires, spatial aliasing is >inevitable<. I am pretty certain this is why it is generally easy to tell a live sound from a recorded one under anything but highly controlled listening situations and particular locations.
5- Making speakers than can work in many listening rooms and for multiple listening postions. This is a far more difficult problem than people who make "room correction" systems want to admit. (I know: I worked on AR's pioneering Adaptive Digital Signal Processor room correction system back in the late 70's.)
6- Better perceptually-relevant measurement techniques."
This is a link to an interview given by him:
Ken Kantor: Now Hear This
Stereophile Editor John Atkinson walked into our office brandishing a CD (footnote 1). "Guess what Ken Kantor did? He took a year off from running NHT (footnote 2) to make this disc." "That's really cool," I responded, drifting into a fantasy about what I'd do with a year off. "I'd like to meet...
And the following is what this post is about:
"OK, I was asked what I thought the major remaining problems in loudspeaker design were. Here is my answer.
-----
To borrow a concept from Artificial Intelligence, I divide the loudspeaker question into two domains: the "hard" problems and the "soft" problems. While I am personally most interested in the hard problems, let's start with the soft problems. These include things that, while they may be very difficult to accomplish, are easy to conceptualize; things like getting a proper frequency response on axis and in a room, controlling radiation pattern, impulse response, distortion, power handling, etc. It's important to note that my short list addresses what is left to be done, not what is most important in the big picture. For example, on-axis frequency response is very important to the sound, but it isn't a very interesting problem anymore. With good drivers and EQ/DSP, it's pretty easy to get flat axial response these days. So, here is my arbitary and non-comprehensive list of the top soft problems in loudspeaker design:
1- Independent control of axial and power response.
2- Radiation pattern control and the elimination of non-coincident sources.
3- Achieving a proper power response in a room without creating a cluttered impulse response within the fusion time.
4- Dealing with the fact that bass response is very dependent on the modulation envelope of the signal. (Bass response for signals with long-time-constant modulation envelopes is quite different from the bass response of short-time-constant modulation envelopes.) I believe this is a widely overlooked phenomenon.
5- Dealing with the variable transmission characteristics of room boundaries at low frequencies, and the attendant issue that deep bass in-room response typically takes between 10 and 30 seconds to stabilize.
Now, on to the "hard" problems, which concern the basic role of the loudspeaker in the record-reproduction chain. These are the ones that interest me, because these are the ones that will allow our hifi systems to truely approach virtual reality, instead of merely hinting at it.
1- Getting it through our collective heads that loudspeakers are not amplifiers and that transfer function metrics like frequency response and distortion are misapplied and a conceptual dead end. They do not appropriately characterize the system. Improving frequency response and distortion past a certain point does not yield further meaningful improvement in realism. (I know this is anathema to most audio people, but I stand by it 100%.)
2- A loudspeaker's sole job is to undo what a microphone does. (Or reproduce what an engineer/musician hears in the studio.) It is folly to suggest there is an ideal loudspeaker when every recording is made in a different way. The plethora of available "state-of-the-art" loudspeakers is a testament to this basic problem. Speaker design cannot progress without standardization of recording techniques. Good luck with that...
3- Signal transforms that reduce dimensionality are inherently irreversible. In other words, there are an infinite number of ways a speaker can reproduce a given microphone signal which are all mathematically correct. Information is lost in the recording process, and that information is gone forever. We are lucky that speakers sound as good as they do!
4- Reproducing a sound field in a listening room is a spatial sampling problem. Unless you have twice the number of loudspeakers as the shortest wavelength you want to reproduce requires, spatial aliasing is >inevitable<. I am pretty certain this is why it is generally easy to tell a live sound from a recorded one under anything but highly controlled listening situations and particular locations.
5- Making speakers than can work in many listening rooms and for multiple listening postions. This is a far more difficult problem than people who make "room correction" systems want to admit. (I know: I worked on AR's pioneering Adaptive Digital Signal Processor room correction system back in the late 70's.)
6- Better perceptually-relevant measurement techniques."
Last edited:
