Hi Guys
I was thinking about what properties an ideal reproducing speaker would have or how two speakers with a similar response can sound totally different..
For the ideal:
One would want the speaker to produce “all the notes” as loud as needed with equal amplitude, where a response curve comes in.
One would want the speaker to produce as little as possible of anything not in the input signal, this extra sound can be considered “noise”. This would include a number of things including harmonic distortion weighted by audibility.
As music is a time variant signal, in theory, one would want “all the notes” to be reproduced at the same time also, this is where group delay or acoustic phase comes in.
Assuming your not suspended in an anechoic chamber, all of these factors need to be considered at the listening point, not at 1 meter.
The room makes it complicated, none of them are the same but in general the further you are from the speaker, the greater the reverberant field is compared to the direct sound.
An example from another area makes the problems in a living room more understandable.
Pretend you are in a large empty gymnasium, in front of you is a stool with a tiny but super accurate speaker, reproducing a person talking.
From 2 feet, the voice is 100% understandable, you are totally in the near field and as you back away, the sound level falls off –6 dB each time you double the distance.
After some paces away, the sound level falls off more slowly AND the voice is getting harder to understand. Far enough away the sound level pretty much stops falling altogether, you are totally in the reverberant field, you can’t understand a word but it sounds like voices.
If one wanted to enlarge the near field distance in the gymnasium, to increase how far the “information” can be reliably transmitted, then one has to have less reverberant field. The catch 22 here is the reverberant field level is proportional the total radiated acoustic power and room absorption VS frequency. This is why intelligibility goes down with increasing level in problem rooms.
In large scale applications like what I design for at work, the solution is using a speaker that has directivity, the sound (ideally) is confined to an angle occupied by the listening area.
Because it is desirable to have the reverberant field have a spectrum similar to the direct field spectrum, a “Constant Directivity” horn or other system is used to proving a fixed coverage angle VS frequency.
My choice for the commercial systems at work has been Constant directivity horns as well.
Measuring a speakers radiation is a pain in the rear, here are two companies I have used to take the spherical measurements needed for the commercial area.
http://www.nwaalabs.com/
http://www.etcinc.us/Data.htm
It is this area where hifi speakers (it seems to me) are particularly poorly documented or understood. For example, when one has two sources of sound, such as an upper and lower driver at crossover, unless they are about ¼ wl or less apart, they produce a new radiation pattern.
The further they are apart acoustically, the more complicated the radiation pattern becomes, the more lobes it has.
These lobes are pointed in other directions than the listener and directly excite the reverberant field with narrow frequency ranges.
While the recorded information is carried by direct sound, the unavoidable reverberant filed effects its “taste”.
Anyway, some rambles on a Sunday night.
Happy New Year
Tom Danley
I was thinking about what properties an ideal reproducing speaker would have or how two speakers with a similar response can sound totally different..
For the ideal:
One would want the speaker to produce “all the notes” as loud as needed with equal amplitude, where a response curve comes in.
One would want the speaker to produce as little as possible of anything not in the input signal, this extra sound can be considered “noise”. This would include a number of things including harmonic distortion weighted by audibility.
As music is a time variant signal, in theory, one would want “all the notes” to be reproduced at the same time also, this is where group delay or acoustic phase comes in.
Assuming your not suspended in an anechoic chamber, all of these factors need to be considered at the listening point, not at 1 meter.
The room makes it complicated, none of them are the same but in general the further you are from the speaker, the greater the reverberant field is compared to the direct sound.
An example from another area makes the problems in a living room more understandable.
Pretend you are in a large empty gymnasium, in front of you is a stool with a tiny but super accurate speaker, reproducing a person talking.
From 2 feet, the voice is 100% understandable, you are totally in the near field and as you back away, the sound level falls off –6 dB each time you double the distance.
After some paces away, the sound level falls off more slowly AND the voice is getting harder to understand. Far enough away the sound level pretty much stops falling altogether, you are totally in the reverberant field, you can’t understand a word but it sounds like voices.
If one wanted to enlarge the near field distance in the gymnasium, to increase how far the “information” can be reliably transmitted, then one has to have less reverberant field. The catch 22 here is the reverberant field level is proportional the total radiated acoustic power and room absorption VS frequency. This is why intelligibility goes down with increasing level in problem rooms.
In large scale applications like what I design for at work, the solution is using a speaker that has directivity, the sound (ideally) is confined to an angle occupied by the listening area.
Because it is desirable to have the reverberant field have a spectrum similar to the direct field spectrum, a “Constant Directivity” horn or other system is used to proving a fixed coverage angle VS frequency.
My choice for the commercial systems at work has been Constant directivity horns as well.
Measuring a speakers radiation is a pain in the rear, here are two companies I have used to take the spherical measurements needed for the commercial area.
http://www.nwaalabs.com/
http://www.etcinc.us/Data.htm
It is this area where hifi speakers (it seems to me) are particularly poorly documented or understood. For example, when one has two sources of sound, such as an upper and lower driver at crossover, unless they are about ¼ wl or less apart, they produce a new radiation pattern.
The further they are apart acoustically, the more complicated the radiation pattern becomes, the more lobes it has.
These lobes are pointed in other directions than the listener and directly excite the reverberant field with narrow frequency ranges.
While the recorded information is carried by direct sound, the unavoidable reverberant filed effects its “taste”.
Anyway, some rambles on a Sunday night.
Happy New Year
Tom Danley
454Casull said:
They should have an edge if they existed.... but again it is only 1 tiny part of the equation.
dave
all this talk of measurements, we are making the big mistake that some of the hifi mags make, in that assuming one is best, and all else is grot.
There are myriad of loudspeakers, some excellent, some good, some less good, all viable, just to write one off on frequency response alone isn't adequate at all.
What matters most is, do you like the music they make? That's all there is to it.
AS I said b4, trying to hear the sound of a loudspeaker via a freq. response is like trying to hear a violin from its freq. response.
There are myriad of loudspeakers, some excellent, some good, some less good, all viable, just to write one off on frequency response alone isn't adequate at all.
What matters most is, do you like the music they make? That's all there is to it.
AS I said b4, trying to hear the sound of a loudspeaker via a freq. response is like trying to hear a violin from its freq. response.
I agree that a single axis freq measurement will not tell you a lot. But a lot of measurements may shed some light on the matter (freq on multiple axis, impulse, decay, (T)HD, IMD). If done well, they will correlate nicely with what you hear. Ultimately the ear is the final judge, but I still rely heavily on my mic during the design process.lt cdr data said:
A couple of random thoughts.
Why is FR important?
In order to reproduce a transient correctly, the FR has to be flat.
I cannot emphasize this enough. Flat FR+minimimum phase behavior=correct transient response. Having either significant FR variability or non minimum phase behavior will distort transients.
So flat FR is theoretically desireable. Conversely, a speaker with marked FR abnormalities will have poor transient behavior in that FR range. PERIOD. No way around this.
Now listening distance is important, as it determines the amount of reverberant sound that is added to the direct sound, as Tom mentioned above. In fact, many of you listen beyond the critical distance and the reverberant sound is more important, so to speak.
So, a smooth power response is just as important, perhaps, as a smooth on axis FR.
However, it's somewhat of a stretch to think that a unit with poor FR variability on axis somehow becomes smooth off axis, making the power response smooth.
So again, this leads to a requirement of FR flatness. Yes, it may be silly to argue about small differences in FR linearity, but the rule is sound. The FR needs to be relatively flat.
I would not condem a car that went 0-60 in 20sec, but to say that "speed" is subjective or in the eye of the beholder is only partially true.
Why is FR important?
In order to reproduce a transient correctly, the FR has to be flat.
I cannot emphasize this enough. Flat FR+minimimum phase behavior=correct transient response. Having either significant FR variability or non minimum phase behavior will distort transients.
So flat FR is theoretically desireable. Conversely, a speaker with marked FR abnormalities will have poor transient behavior in that FR range. PERIOD. No way around this.
Now listening distance is important, as it determines the amount of reverberant sound that is added to the direct sound, as Tom mentioned above. In fact, many of you listen beyond the critical distance and the reverberant sound is more important, so to speak.
So, a smooth power response is just as important, perhaps, as a smooth on axis FR.
However, it's somewhat of a stretch to think that a unit with poor FR variability on axis somehow becomes smooth off axis, making the power response smooth.
So again, this leads to a requirement of FR flatness. Yes, it may be silly to argue about small differences in FR linearity, but the rule is sound. The FR needs to be relatively flat.
I would not condem a car that went 0-60 in 20sec, but to say that "speed" is subjective or in the eye of the beholder is only partially true.
ucla88 said:
minimum phase behaviour is just important to your equation. And that eliminates many, many candidates. Getting flat FR most often ends up screwing up the phase response seriously. One of the reasons a lot of people put up with full-ranges. They tend to be much better behaved phase wise.
attached it a 30 degree off axis of the same speaker in posted on-axis for before.
dave
Attachments
Well, "flattening the FR" is not what causes non minimum phase behavior. Basically, all individual drivers are min phase, except for some very infrequent oddities. And even this is debatable.
The nonminimum phase comes from the xover, not from trying to get the system flat.
Now single drivers are theoretically appealing, but they cannot do a good job with either nonlinear distortion or smooth power response-in general. (I realized there are exception).
As far as your off axis plots, well you're right in your example. This is likely just diffraction effects. Whch do ge better off axis. Which is why diffractio may be overrated as a problem
The nonminimum phase comes from the xover, not from trying to get the system flat.
Now single drivers are theoretically appealing, but they cannot do a good job with either nonlinear distortion or smooth power response-in general. (I realized there are exception).
As far as your off axis plots, well you're right in your example. This is likely just diffraction effects. Whch do ge better off axis. Which is why diffractio may be overrated as a problem
planet10 said:
I bet people like to listen that speaker slightly off axis 🙂
I am not much experienced with full-ranges but the reason why they have a fan base may be more related to them being closest thing to point sources and the resultant nice imaging. I don't see the point in sacrificing the amplitude response flattness for the sake of some phase linearity. How much of a phase distortion is audible as far as know is still being debated, but I don't think many will debate that a 1db peak on upper midrange will change the sound of a speaker profoundly. You could have a nice phase response at the expense of having a nasty FR amplitude response. What will this gain? The linear distortion is caused by both phase and amplitude nonlinearities. If the argument is "I have the phase right, so what amplitude is screwed" is assumed correct, then what is wrong with this argument "I got the amplitude flat, so what the phase is screwed"? As a mathematical model, both divergence from flat FR amp and linear phase results in transient distortion. Add two sine waves and record it. Then change the amplitude of one sine wave without touching its phase, the sum will be distorted version of the previous sum of two sine waves. In addition to this, our hearing is more sensitive to FR amp irregularities than non-linear phase issues.
I can't recall of the top of my head which exact studies they were, but I remember reading a few studies that resulted in people preferring speakers with flat anechoic FR to others. I think this was related to how humans process sound, the early sound and then the reflections are processed differently. In this sense, the humar hearing is not different then the measuring software that gates out the reflections to yield pseude-anechoic measurements. Even though when put in a real room the speaker's FR will change, its anechoic FR signature is still heard.
Something like these anyways 🙂
Feyz said:
Point source & single driver are both phase things. They are preferred by some because of smooth phase response with no discontinuities. The brain picks out those discontinuities like a bull sees a red cape.
dave
I must both agree and disagree. Whereas the point source will have superior phase characteristics, the audibility of this is still debatable. What isn't debatable is the superior power response of the point source. That perhaps combined with the phase characteristics, is what might make some like single drivers.
However, distortion bass and breakup treble make them non-ideal IMHO. A 10 inch cone tweeter is a poor reproducer of cymbals. A 4 inch subwoofer can't recreate a kick drum. The hopeless dilema of a "fullrange" single driver.
That I agree with totally. A high performance coaxial driver like the TAD of KEF reference will be similar, but vastly superior to a fullrange.
Absolutely. The NRC research showed this. Something like this:
http://www.harman.com/wp/pdf/Loudspeakers&RoomsPt2.pdf
which sort of brings us back to the essence of the thread.
No, Wilson has no clue how to design an accurate loudspeaker, or perhaps they do but choose not to. They absolutely do know what they are doing when it comes to parting (hi-end) fools and their money. So essentialy, yes, they do know what they are doing😉 .
Cheers and Happy New Year(s)
AJ
Every great amplifier should have a high-quality speaker system
otherwise youd miss out on amp precision
Regards
otherwise youd miss out on amp precision
Regards
454Casull said:
Not very flat above 2k, so no it isn't close (althou in its passband, under the conditions it was measured it is pretty flat). You'd have to add a tweeter at the minimum which, in most cases, means a discontinuity.
dave
Over 15 years ago, I would say the Wilson Audio System I Listened to were very good. Now, having done much more listening and testing, I don't know if I would have the same opinion.
Better Phase characteristics produce a better sound stage. Frequency response is also very important due to coloration of sound. Since lots of people don't have the time to listen to live performances, even less have the opportunity to listen to performances without the amplification, too many people just try to imagine or just tend to listen to sound balance, or what ever that sounds soothing. However, the better you can get the phase, the more you can feel how the instrument is positioned in the performance hall and how it relates with the room timber.
Not all full range drivers are the same, but they are easier to work with. In the process of phase and frequency compensation, there still is a trade-off. Generally if you can keep the phase response within 30 degrees at least above 1K, you are pretty good, and can focus on the response as long as the phase is in this range.
Better Phase characteristics produce a better sound stage. Frequency response is also very important due to coloration of sound. Since lots of people don't have the time to listen to live performances, even less have the opportunity to listen to performances without the amplification, too many people just try to imagine or just tend to listen to sound balance, or what ever that sounds soothing. However, the better you can get the phase, the more you can feel how the instrument is positioned in the performance hall and how it relates with the room timber.
Not all full range drivers are the same, but they are easier to work with. In the process of phase and frequency compensation, there still is a trade-off. Generally if you can keep the phase response within 30 degrees at least above 1K, you are pretty good, and can focus on the response as long as the phase is in this range.
AJinFLA said:
One thing to remember is that some Wilsons (Grand SLAMM for sure, perhaps others) have an auxiliary tweeter on the back panel, crossed over at 6 dB/octave at about 10 or 12 kHz, which helps to flatten power response in the room since tweeters start getting directional above 5 kHz. The second driver, even though it doesn't face the listener, provides a sense of air.
Wilsons also have dead cabinets, which removes a source of stored energy re-radiating music.
I'm still not enchanted with their choice of drivers, though.
Francois.
planet10 said:
Point source vs multi driver is not only about phase. A multi driver can be made linear phase using certain xovers, and a single driver can be non-minimum phase, the ones that have a small cone in the middle (I forgot the term for it) for high frequency extension are likely to be non-minimum phase. A multi driver which is linear phase will not sound like a single driver which is also linear phase, because multi driver's off axis performance will have lobes.
As for phase response discontinuity, what exactly are you refering to? Phase wrappings at +180/-180 are not actually discontinuties, they look like that because of the plotting. If the phase is not wrapped, they wouldn't look like discontinuities. For instance if the phase goes to from 180 to 190, it will show like a discontinuity with phase wrapping (-180/+180), because it will look like it goes from 180 to -170 (190 - 360). But if the phase is unwrapped it won't show as discontinuity, it would just go from 180 to 190. Even if phase is not unwrapped but plotted from 0 to 360, instead of -180 to +180, that will not show as a discontinuity.
Actually you should be looking at the group delay, not just the phase itself. And how much of a group delay under what signal or music is audible is to my knowledge is still under study and not very well defined. By the way if phase had discontinuties, the group delay being its negative derivate, would have infinite values at the discontinuties. Neither of the commonly used Linkwitz or Butterworth xovers have such things in their sum's group delays.
Feyz said:
At every XO point (including where a whizzer cone takes over from a main cone in a FR)
dave
Feyz said:
That would be great if the drivers were perfect. Bu they are the most variable part in the system.
- Status
- Not open for further replies.