Ardacer
Active Member
- Joined
- Jun 13, 2018
- Messages
- 106
- Likes
- 134
I'll update this post as I have time, with images and text. Whoever wishes to contribute in any way, please do.
Most generally, there are 2 main ways one can go about this to achieve similar results - that is, good sound.
1) make everything from few components of really high quality
2) med-high quality, multiple components
3) a cake made of shit is still shit
Fist way is usually a 2-way or 3-way, big bass reflex, with quality drivers and cabinets or huge horns and compression drivers. Usually employs analog crossovers of low order, with very high quality amplifiers.
Multi-amping fest is the second way. It has a drawback of added complexity, but gives many benefits in turn.
When a single driver (or amp) is forced to play a large chunk of the audible range, or the whole of it, intermodular distortion (IMD) will increase the noise floor very substantially while playing complex waveforms, masking the finer detail in sound. There are ways around this but it requires very high quality components. What you can do instead, which is arguably even better - is divide the spectrum into 3 or more drivers, and have an amp for each driver. This also alleviates this problem nicely. Amps don't have to be top notch, they only drive inductive loads and a single driver, and drivers themselves only have to be good at the smaller chunk of the spectrum that they are set for. This also requires an active crossover, either digital or analog (electronic crossover) - it is also a bit easier to mix and match drivers that might not work as nicely together otherwise due to impedance, sensitivity, etc.
In general when selecting a driver for a project, it would be good to see an IB loaded FR plot (1m distance or near field, on axis) with distortion plots at something like 75-80db Z weighted (normal listening levels) and 95-100db, for transients. Manufacturers rarely do this, they instead advertise an "useable range", in which the thd+n is under 1-2% or so under normal conditions. There are some websites like Zaph audio, that host driver measurements, which is very helpful here.
THD under 1% (or 40db under fundamental) throughout the entire range it will perform, is what you aim for. If you plan to use lower order crossovers, it needs to behave well a lot further into the range of the other driver. With very steep, 48db/oct digital filters for example, you pretty much have nothing to worry about with regards to driver behaviour over/under the crossover frequency (8order LR needs to behave well just a half of octave above/below, 4th order about an octave, 2nd 2 octaves, 1st order 4 octaves).
That's what I know of problems with IMD and THD+noise.
Now...
What do you want from your speakers? That will decide the optimal form factor.
Will you be be using them in nearfield? Do you want to experience the singers and instruments as if they were in your room, there with you? Do you want to feel like you are in a concert hall, or out in the open?
Does the WAF factor apply to you?
Nearfield
If your goal is to have a simple, small speakers that you'd use in nearfield, you can get away with a lot of things. They will sit really close to you, so the main incoming soundwave will inherently be a lot stronger than any reflected sound. This means you can focus mainly on distortion and linearity of the on axis response. It is always a good idea to have the drivers positioned on the main baffle as assymetrical as possible, and as close to each other as possible. Properly done (which is HARD) coaxials are probably the best choice here (example - KEF, genelec). Sealed box is always better than ported for the bass control and quality - it's just the nature of things. Proper ported designs will sound good, however. Optimal choice in my opinion is a small, closed box with a good coaxial driver, and a dedicated sub (or 2) for frequencies <100-120 Hz. Ported for quantity, sealed for quality.
If you want both, then a sealed box with high excursion quality drivers mounted in opposing directions (I'll add image), so that, when one pushes, other pulls, and this way you get 6 more db of output, and total vibration cancelling - the box is still. Keep in mind that doubling the distance means -6db of perceived sound pressure level. Also, every big reflective surface that is very close to the driver (not so close to literally not allow it to breathe) will also add +6db to the output - meaning wall1, wall2, ceiling, floor, etc.
2 subs is usually better than one, but for nearfield configuration it might not matter.
Baffle step starts at 1/8 * (13560/smallest dimension in inches), stops at 2 * (13560/smallest dim.). You will have to correct the transition somehow, from 4pi to 2pi radiation, to keep the main sound front linear (positioning actually matters here too, with regards to reflective surfaces).
Room integration - far fields
This part is rather speculative on my part, I admit, but it is grounded in works of other people smarter than me (Linkwitz, gainphile, John K), and my background as a neuroscientist (neuroophthalmologist with passion for hearing, and soon a PhD in neuroradiology).
Consider how our hearing apparatus is built.
There is, first, your earlobe. It's a glorified reverse horn. Then, there are bones and membranes that serve to bridge the acoustic impedance of air and in-ear liquid. Once the sound information is inside this liquid, it goes to the nautilus shaped inner ear. It is basically a continuous spiral chamber (image will be added).
Think of it as a continuous spectrum analyzer basically. It forms standing waves, for each frequency at a dedicated location, with nerve fibers collecting this data towards the higher structures in the brain.
These higher structures integrate this information over time - this is quite crucial. In doing so, together with other senses, we gain perception of reality around us.
Sounds in our environment naturally reflect from objects in our room. If they didn't, you'd get something like a headphone soundstage all the time (considering no head movement).
These reflections, together with the main sound front, help us create the imgage of space in our brain, sort of how like 2 eyes, or one eye changing location helps create a sense of depth.
So eliminating the reflections means eliminating the sense of something happening inside the room. You might want this, you might not. In any case, reflections are usually spectraly identical to the sounds that cause them in nature. That's why it's /very important/ to keep the spectral energy within the room contained, if you care about such stuff.
That's why it's the biggest problem when a speaker has no control, or it was not on the mind of the designer, of the sound directivity/dispersion. The worst offender is a poor crossover - usually a beaming woofer crossed to steep to an omnidirectional tweeter (at that specific frequiency) (image inserted here).
That causes a sharp change in the spectrum of reflections and the main wave, causing the perceived "box" coloration. Also, if there is a box, it must be properly braced and configured to minimize stored energy which can also raise the noise floor and mask tiny details in the sound (stuff Purrin calls plankton) I guess.
For the best "experience the singers and instruments as if they were in your room" type of speaker, it would surely be:
A full range omni speaker. That's just impossible or, at least, stupidly impractical, for now. The other thing one can do is an infinite baffle speaker configured for total 2pi radiation. It sort of behaves like an omni, because the backwave.. there is literally a wall there - there is no back wave. To do this, you'd need at least a 4-way or even a 5-way speaker with drivers quite close to each other.
Each speaker driver should have a total 2pi radiation (except for the very high end, over 15khz or so) in the part of the spectrum it plays.
If you position 2 drivers one over another, there will be a length between their acoustic centers. If this length is smaller than that of the lambda (wavelength) of the crossover frequency, there will never be a full cancelation in any direction. If it is rather small compared to the wavelength, the crossover frequency radiation to the extreme vertical points will be attenuated, slightly. At the distance = 1/4 lambda, that attenuation is around 4.5db. At 1/8, it's 2db, and at 1/2, it's 12 db. Keeping crossovers very steep, like 8th order, helps keep the attenuated reflected window very tiny part of the spectrum, keeping the reflections basically the same spectrally as the front wave, everywhere.
This also requires dealing with the walls - depending on the situation, waf is either huge, or 0.
Other posibilities are some interesting smart omni designs, but as far as I know, they are almost never really true omni. It might be good enough, of course, ymmv. Properly controlled directivity speakers will have a gradual transition from an omni to a directed sound (image). It's not ideal in the sense that the spectrum of reflections is not the same as that of the front wave, but at least there are no huge leaps, no big breaks, and it's tolerated better.
In order to truly kill the room as much as possible, to
"feel like you are in a concert hall, or out in the open"
you would surely need..
the frond radiation of a dipole, with the back bubble missing, and have no reflective surface behind you.
This, too, is impossible or at least stupidly impractical.
Full range dipoles (image), when configured properly, close to a reflective surface behind them, or a diffusive surface behind them, and a diffusive surface behind you, is the best bet.
Without any care for the surfaces, they still deaden the room a LOT. You are left with the original soundstage of the stereo recording as much as possible. Similar to them are the full range (or partial range) cardioids, or cardioid/dipole hybrids - but they actually radiate more to the surfaces left and right. Nothing beats a proper dipole. They are quite hard to model (examples, nao note, lx521, orions, gainphile). They will absoultely need a complex crossover and dsp, and a lot of very poweful, very well built (well vented spyders, high excursions, etc) woofers to work properly.
This is what I have for you for now.
If you made it this far, hats off, and thank you.
Most generally, there are 2 main ways one can go about this to achieve similar results - that is, good sound.
1) make everything from few components of really high quality
2) med-high quality, multiple components
3) a cake made of shit is still shit
Fist way is usually a 2-way or 3-way, big bass reflex, with quality drivers and cabinets or huge horns and compression drivers. Usually employs analog crossovers of low order, with very high quality amplifiers.
Multi-amping fest is the second way. It has a drawback of added complexity, but gives many benefits in turn.
When a single driver (or amp) is forced to play a large chunk of the audible range, or the whole of it, intermodular distortion (IMD) will increase the noise floor very substantially while playing complex waveforms, masking the finer detail in sound. There are ways around this but it requires very high quality components. What you can do instead, which is arguably even better - is divide the spectrum into 3 or more drivers, and have an amp for each driver. This also alleviates this problem nicely. Amps don't have to be top notch, they only drive inductive loads and a single driver, and drivers themselves only have to be good at the smaller chunk of the spectrum that they are set for. This also requires an active crossover, either digital or analog (electronic crossover) - it is also a bit easier to mix and match drivers that might not work as nicely together otherwise due to impedance, sensitivity, etc.
In general when selecting a driver for a project, it would be good to see an IB loaded FR plot (1m distance or near field, on axis) with distortion plots at something like 75-80db Z weighted (normal listening levels) and 95-100db, for transients. Manufacturers rarely do this, they instead advertise an "useable range", in which the thd+n is under 1-2% or so under normal conditions. There are some websites like Zaph audio, that host driver measurements, which is very helpful here.
THD under 1% (or 40db under fundamental) throughout the entire range it will perform, is what you aim for. If you plan to use lower order crossovers, it needs to behave well a lot further into the range of the other driver. With very steep, 48db/oct digital filters for example, you pretty much have nothing to worry about with regards to driver behaviour over/under the crossover frequency (8order LR needs to behave well just a half of octave above/below, 4th order about an octave, 2nd 2 octaves, 1st order 4 octaves).
That's what I know of problems with IMD and THD+noise.
Now...
What do you want from your speakers? That will decide the optimal form factor.
Will you be be using them in nearfield? Do you want to experience the singers and instruments as if they were in your room, there with you? Do you want to feel like you are in a concert hall, or out in the open?
Does the WAF factor apply to you?
Nearfield
If your goal is to have a simple, small speakers that you'd use in nearfield, you can get away with a lot of things. They will sit really close to you, so the main incoming soundwave will inherently be a lot stronger than any reflected sound. This means you can focus mainly on distortion and linearity of the on axis response. It is always a good idea to have the drivers positioned on the main baffle as assymetrical as possible, and as close to each other as possible. Properly done (which is HARD) coaxials are probably the best choice here (example - KEF, genelec). Sealed box is always better than ported for the bass control and quality - it's just the nature of things. Proper ported designs will sound good, however. Optimal choice in my opinion is a small, closed box with a good coaxial driver, and a dedicated sub (or 2) for frequencies <100-120 Hz. Ported for quantity, sealed for quality.
If you want both, then a sealed box with high excursion quality drivers mounted in opposing directions (I'll add image), so that, when one pushes, other pulls, and this way you get 6 more db of output, and total vibration cancelling - the box is still. Keep in mind that doubling the distance means -6db of perceived sound pressure level. Also, every big reflective surface that is very close to the driver (not so close to literally not allow it to breathe) will also add +6db to the output - meaning wall1, wall2, ceiling, floor, etc.
2 subs is usually better than one, but for nearfield configuration it might not matter.
Baffle step starts at 1/8 * (13560/smallest dimension in inches), stops at 2 * (13560/smallest dim.). You will have to correct the transition somehow, from 4pi to 2pi radiation, to keep the main sound front linear (positioning actually matters here too, with regards to reflective surfaces).
Room integration - far fields
This part is rather speculative on my part, I admit, but it is grounded in works of other people smarter than me (Linkwitz, gainphile, John K), and my background as a neuroscientist (neuroophthalmologist with passion for hearing, and soon a PhD in neuroradiology).
Consider how our hearing apparatus is built.
There is, first, your earlobe. It's a glorified reverse horn. Then, there are bones and membranes that serve to bridge the acoustic impedance of air and in-ear liquid. Once the sound information is inside this liquid, it goes to the nautilus shaped inner ear. It is basically a continuous spiral chamber (image will be added).
Think of it as a continuous spectrum analyzer basically. It forms standing waves, for each frequency at a dedicated location, with nerve fibers collecting this data towards the higher structures in the brain.
These higher structures integrate this information over time - this is quite crucial. In doing so, together with other senses, we gain perception of reality around us.
Sounds in our environment naturally reflect from objects in our room. If they didn't, you'd get something like a headphone soundstage all the time (considering no head movement).
These reflections, together with the main sound front, help us create the imgage of space in our brain, sort of how like 2 eyes, or one eye changing location helps create a sense of depth.
So eliminating the reflections means eliminating the sense of something happening inside the room. You might want this, you might not. In any case, reflections are usually spectraly identical to the sounds that cause them in nature. That's why it's /very important/ to keep the spectral energy within the room contained, if you care about such stuff.
That's why it's the biggest problem when a speaker has no control, or it was not on the mind of the designer, of the sound directivity/dispersion. The worst offender is a poor crossover - usually a beaming woofer crossed to steep to an omnidirectional tweeter (at that specific frequiency) (image inserted here).
That causes a sharp change in the spectrum of reflections and the main wave, causing the perceived "box" coloration. Also, if there is a box, it must be properly braced and configured to minimize stored energy which can also raise the noise floor and mask tiny details in the sound (stuff Purrin calls plankton) I guess.
For the best "experience the singers and instruments as if they were in your room" type of speaker, it would surely be:
A full range omni speaker. That's just impossible or, at least, stupidly impractical, for now. The other thing one can do is an infinite baffle speaker configured for total 2pi radiation. It sort of behaves like an omni, because the backwave.. there is literally a wall there - there is no back wave. To do this, you'd need at least a 4-way or even a 5-way speaker with drivers quite close to each other.
Each speaker driver should have a total 2pi radiation (except for the very high end, over 15khz or so) in the part of the spectrum it plays.
If you position 2 drivers one over another, there will be a length between their acoustic centers. If this length is smaller than that of the lambda (wavelength) of the crossover frequency, there will never be a full cancelation in any direction. If it is rather small compared to the wavelength, the crossover frequency radiation to the extreme vertical points will be attenuated, slightly. At the distance = 1/4 lambda, that attenuation is around 4.5db. At 1/8, it's 2db, and at 1/2, it's 12 db. Keeping crossovers very steep, like 8th order, helps keep the attenuated reflected window very tiny part of the spectrum, keeping the reflections basically the same spectrally as the front wave, everywhere.
This also requires dealing with the walls - depending on the situation, waf is either huge, or 0.
Other posibilities are some interesting smart omni designs, but as far as I know, they are almost never really true omni. It might be good enough, of course, ymmv. Properly controlled directivity speakers will have a gradual transition from an omni to a directed sound (image). It's not ideal in the sense that the spectrum of reflections is not the same as that of the front wave, but at least there are no huge leaps, no big breaks, and it's tolerated better.
In order to truly kill the room as much as possible, to
"feel like you are in a concert hall, or out in the open"
you would surely need..
the frond radiation of a dipole, with the back bubble missing, and have no reflective surface behind you.
This, too, is impossible or at least stupidly impractical.
Full range dipoles (image), when configured properly, close to a reflective surface behind them, or a diffusive surface behind them, and a diffusive surface behind you, is the best bet.
Without any care for the surfaces, they still deaden the room a LOT. You are left with the original soundstage of the stereo recording as much as possible. Similar to them are the full range (or partial range) cardioids, or cardioid/dipole hybrids - but they actually radiate more to the surfaces left and right. Nothing beats a proper dipole. They are quite hard to model (examples, nao note, lx521, orions, gainphile). They will absoultely need a complex crossover and dsp, and a lot of very poweful, very well built (well vented spyders, high excursions, etc) woofers to work properly.
This is what I have for you for now.
If you made it this far, hats off, and thank you.