When we talk about time domain response in speakers, I think of issues related to sealed verses ported enclosure designs. Porting is often used to extend low frequency response by exploiting resonance effects. While extending bass frequency response, ports tend to also make bass flabby and poorly defined. They make it hard to accurately reproduce essential differences between a 40Hz bass guitar and a 40Hz kick drum.
Unless I am missing something, those aspects of speaker performance can be reasonably measured with waterfall plots and impulse response measurements. Are you saying one can accurately derive waterfall and impulse response graphs from modeled frequency response and phase graphs?
Last edited:
I believe it is ARTA that can switch between the two at the click of a mouse from an MLS measurement.
Jan
Here: ARTA Software it says Arta can measure both. I don't see that it has any ability to model frequency response then deduce from that a transient model.
Again, the point I have been asking about relates to modeling and prediction tools, not measurements after a system has already been built.
I don't see this giving the right answer except in a strictly minimum phase system.
Jan, Bill is right, you're mostly wrong. Again the only common place where the resonate frequency matters (which is the bump in impedance) is if it's in a spot where it's preventing the desired slope you want, or it causes a big bump. For woofers it's hard to cause a big bump because it's almost exclusively in a range that is already down from the general sensitivity of the driver, meaning any increase is beneficial. But it mostly means the FS is below the appreciable range (well below the -3db).
So because the amp doesn't care, and your crossover point is say at 2khz, and we're talking about and FS of say 43hz, and your woofer's -3db is at 55hz... It's a moot point. Care not.
If it's a midrange you'll probably have the FS below the crossover point as well. Let's say FS is 158hz, and your cross at 250hz. In this case it's exactly like a tweeter, the only reason you reduce the impedance is to knock down the hump that otherwise is interfering with your desired crossover slope.
So therefor the impedance of the driver at FS has little to do with the crossover design besides when it interferes with slopes. Now if you want to talk about the relatively benign impedance from the inductance of driver, that's a little different. First, it's basically compensated for by the amplifier, but some designers prefer it to be rather low. When it's low a speaker may sound better depending on which range the low inductance driver plays. SOO for a subwoofer it's moot, but for a Fullrange driver it may bring a bit of clarity. And you probably know there are trade offs.
But really when making a crossover accounting for inductance is no big deal. When you look at a frequency response plot of the driver you'll notice pretty much nothing of interest at some higher points inductance. That may be partly because as frequency climbs so does directionality of the produced sound waves. Generally a speaker will lose ability to reproduce the frequencies before the inductance starts to do much.
The point is that you're barking up the wrong tree, as Bill is saying.
Mark,
Port noise can be improved by regulating the velocity of the port. I've heard many speakers that produce bass that is extremely articulate, while ported. It's possible to go either way. There's a number of factors at play.
Why would you need to model anything beyond FR and phase data? As Jan says, that is time domain in a way that's useful for designing a speaker. Aside from that there's no reason to model them really, you just measure drivers and pick the ones you like. The crossover and box won't change from the basic T/S parameters, FR, and phase. It's the subjective sound where you'll note the differences in drivers transient response.
For tweeters it could be useful to know the polar plot won't be as pretty as the given, if the waterfall is poor for a tweeter. But it doesn't help you much unless you're designing a horn, wave guide, phase plug, or something. A simulator for that would be pretty neat... and complicated.
So because the amp doesn't care, and your crossover point is say at 2khz, and we're talking about and FS of say 43hz, and your woofer's -3db is at 55hz... It's a moot point. Care not.
If it's a midrange you'll probably have the FS below the crossover point as well. Let's say FS is 158hz, and your cross at 250hz. In this case it's exactly like a tweeter, the only reason you reduce the impedance is to knock down the hump that otherwise is interfering with your desired crossover slope.
So therefor the impedance of the driver at FS has little to do with the crossover design besides when it interferes with slopes. Now if you want to talk about the relatively benign impedance from the inductance of driver, that's a little different. First, it's basically compensated for by the amplifier, but some designers prefer it to be rather low. When it's low a speaker may sound better depending on which range the low inductance driver plays. SOO for a subwoofer it's moot, but for a Fullrange driver it may bring a bit of clarity. And you probably know there are trade offs.
But really when making a crossover accounting for inductance is no big deal. When you look at a frequency response plot of the driver you'll notice pretty much nothing of interest at some higher points inductance. That may be partly because as frequency climbs so does directionality of the produced sound waves. Generally a speaker will lose ability to reproduce the frequencies before the inductance starts to do much.
The point is that you're barking up the wrong tree, as Bill is saying.
Mark,
Port noise can be improved by regulating the velocity of the port. I've heard many speakers that produce bass that is extremely articulate, while ported. It's possible to go either way. There's a number of factors at play.
Why would you need to model anything beyond FR and phase data? As Jan says, that is time domain in a way that's useful for designing a speaker. Aside from that there's no reason to model them really, you just measure drivers and pick the ones you like. The crossover and box won't change from the basic T/S parameters, FR, and phase. It's the subjective sound where you'll note the differences in drivers transient response.
For tweeters it could be useful to know the polar plot won't be as pretty as the given, if the waterfall is poor for a tweeter. But it doesn't help you much unless you're designing a horn, wave guide, phase plug, or something. A simulator for that would be pretty neat... and complicated.
Why? I would say take a look at all the graphs for the speakers in this document: http://dt7v1i9vyp3mf.cloudfront.net/assetlibrary/n/ns10m.pdf?jQWj8tYIeZeymRCNXitG9Qfwq9mLf1t0
What I notice is that speakers with the best frequency responses don't have the the best transient response, and vice versa. That being the case, if all one has is a modeling program that directly let's one see performance in terms of frequency response, I don't think it would be clear to anyone by looking at data such a program produces, what the transient response is likely going to turn out to be when eventually measured.
So what is likely to happen? One proceeds to design a flat response speaker with poor transient response and then after the speaker is built, measure it and find out about the transient response performance. If they don't like what they see at that point, and they go back to their speaker design software program, and what should they do then? What can they do? Only design another speaker with flat frequency response.
Well I am just trying to disprove someone else's ideas. His point is that at a speaker resonance point, not only is the magnitude quite high like 20 or 30 ohms, but also the phase goes through a rapid change. This is shown in the measurement by Stereophile some pages back.
I agree with you that 'the amp doesn't care' but my opponent claims it is harder for the amp to cope with the phase shift between output voltage and current. So he says he wants to use an impedance correction network // to the speaker as a whole to make it look resistive and thus constant, like 8 ohms, to the amp.
His 2nd point is that if you do the impedance corrections to each individual driver you make it easier to design the xover for that driver as the xover design process assumes resistive load.
I'm just asking what you guys think about this. Please do read what I write.
Jan
I agree with you that 'the amp doesn't care' but my opponent claims it is harder for the amp to cope with the phase shift between output voltage and current. So he says he wants to use an impedance correction network // to the speaker as a whole to make it look resistive and thus constant, like 8 ohms, to the amp.
His 2nd point is that if you do the impedance corrections to each individual driver you make it easier to design the xover for that driver as the xover design process assumes resistive load.
I'm just asking what you guys think about this. Please do read what I write.
Jan
That I don't know. Not sure how it works in detail but the functionality is there. Have been using it extensively with Vacuphile last week measuring speakers.
Jan
Well I'd be interested to hear a purely resistive network. The challenge to get the FR somewhat flat would likely mean dropping the sensitivity a lot to shrink the changes that all the impedance compensations make. Which is sorta counter intuitive because the amplifiers that benefit the most are low power.
Mark, the problem with speaker drivers is the ideal one has a linear increase in volume with frequency. Basically you need a horn to compensate for it, in direct inverse. Point being everything else is a massive compromise. When it comes to understanding transient response you're onto something, basically it takes some experience from what I've been told by a guy who designs speakers and drivers. But I'd say stuffing and box shape are going to play big factors, as where the crossover is debatable. There are some considerations to the complex-impedance one may want to think on. But in general I can only imagine you're referring to using just a couple of capacitors? Otherwise I'd assume Q values, cone mass, BL, and such dominate over the crossover. Maybe I just need an example?
Mark, the problem with speaker drivers is the ideal one has a linear increase in volume with frequency. Basically you need a horn to compensate for it, in direct inverse. Point being everything else is a massive compromise. When it comes to understanding transient response you're onto something, basically it takes some experience from what I've been told by a guy who designs speakers and drivers. But I'd say stuffing and box shape are going to play big factors, as where the crossover is debatable. There are some considerations to the complex-impedance one may want to think on. But in general I can only imagine you're referring to using just a couple of capacitors? Otherwise I'd assume Q values, cone mass, BL, and such dominate over the crossover. Maybe I just need an example?
Jan
In the past I did test control of speaker impedance at resonance by acoustic means.
Notice the beneficial change in reactive power.
Look at the diagrams here
http://www.diyaudio.com/forums/loun...ch-preamplifier-part-ii-7910.html#post4646076
http://www.diyaudio.com/forums/loun...ch-preamplifier-part-ii-7920.html#post4648832
George
Suppose I want to design a speaker that is down maybe only 3dB at 40Hz, but that also decays down 40dB at 40Hz within 30ms after the end of a tone burst? In other words, I want to specify waterfall performance, in addition to frequency response performance. And I don't care too much about sensitivity, which I am willing to sacrifice to get the other specs right. Can I simulate the required design parameters in boxsim or any other program people use? And if I can't find a solution for exactly what I want, can I play around with design trade offs for the specified parameters in my software program, whatever program is best for that?
The point I have been trying to make is I think there is a problem in that the design process is currently focused on frequency response to the exclusion of other things, primarily because that's what the design tools are good at.
Last edited:
Hi Jan.
Re: crossovers. Not sure if this adds anything to the discussion but I seem to recall that about 20-30years ago(?) KEF made a big deal about this and marketed very complex networks that were advertised as "conjugate load matching networks" (See ad' for their R103/3) E.g. the R104/2 at that time had a 30 element circuit! They must have felt it was important. The "selling point" in their advertising brochures being that it made it an easier load for the amp' to drive.
(If you Google "KEF conjugate matching crossovers" the 4th site has excerpts from John Borthwick's book on Loudspeakers and Headphones with a schematic and impedance curve for the R104.)
Cheers Jonathan
Re: crossovers. Not sure if this adds anything to the discussion but I seem to recall that about 20-30years ago(?) KEF made a big deal about this and marketed very complex networks that were advertised as "conjugate load matching networks" (See ad' for their R103/3) E.g. the R104/2 at that time had a 30 element circuit! They must have felt it was important. The "selling point" in their advertising brochures being that it made it an easier load for the amp' to drive.
(If you Google "KEF conjugate matching crossovers" the 4th site has excerpts from John Borthwick's book on Loudspeakers and Headphones with a schematic and impedance curve for the R104.)
Cheers Jonathan
Last edited:
Why didn't you say
. This was the bit that isn't 'mostly' and the real biatch with passive systems and why an awful lot of 2-ways are so inefficient. Of course if you biamp you can just bang some series resistance in to tip towards current drive, but TBH not sure whether there is much you can do without ending up with something with the efficiency of an LS3/5a.A linkwitz transform before the power amp will do what you want, but of course you are cutting into power handling as soon as you start boosting bass.
90% of the problem is overdamped alignments of porting. ATC use underdamped ports to maximise power handling and no one I am aware of has accused any of their studio mains as having flabbly bass. I've only heard the 100As and they certainly impressed the hell out of me in all bar one parameter.
As long as you can normalise. I've not had a chance to print out the link you published to compare, but as the NS-10 are 10dB down from anything approaching 'fidelity' there is some normalisation to work what the waterfall plots actually mean.
Jan-
1) in a sense there is an issue because you need the necessary voltage capacity to have adequate drive at resonance and you also need sufficient current capacity for the lowest impedance with the worst phase angle. Similar to building utility power distribution- you need to provide for the worst case VA but only get paid for the Watts. However adding a parallel impedance across the driver to reduce its Z doesn't help with the voltage requirement and makes the power requirements at that peak worse (more devices and heat sink). That is a situation where as an amp designer I don't really want the "help".
2) For a crossover- designing a passive crossover (which seems really retro today) you can start with the conventional filter solutions but they are not going to get you far. If you correct the Z of the driver to be flat it makes the arithmetic work but is far from an optimum solution. If you can make a decent model of the drivers impedance and throw the lot into a spice simulator you can start getting to a solution but you still need to focus the actual acoustics of the driver since that's the output that matters. Unfortunately the acoustics are 3D and spice is 1D so the ideal on axis response may be really bad 20 degrees off axis. Correcting for a lot of small resonances with a passive crossover is not likely to be successful so get better drivers if you have problems. Waterfall plots are great for identifying problems but painful if you can't fix them, like cone resonances. Low frequency waterfalls are hard to get. You need a large space. I use ground plane in a warehouse where all the reflections are 50' away with a windowed measurement. Still the ambient noise is a limitation at low frequencies for waterfalls. Even for headphones.
Mark-
Virtually all bass reproduction solutions use resonance to some degree. Some solutions like the Bose tube effectively cascade the resonances so they have a high Q and loose coupling and just ring on. Sealed boxes are the other extreme with pretty low Q but less efficiency at low frequencies. If you have enough power and a closed loop servo (speaking from experience here) you can make resonance free extended bass as low as you want. However many won't hear the ringing and peaking in the mid-bass that is interpreted as bass.
The other great insult to good bass is summing the bass. All out of phase content is effectively removed. Microphones are not all in the exact same place at low frequencies so they will have phase differences at low frequencies. I always advised customers who could not afford two subwoofers to connect one sub to only one channel.
Bill, I am not saying it's impossible to design some kind of ported speaker with a good resulting waterfall plot.
I'm saying that available design tools that people actually use focus on designing for, and displaying predicted performance graphs in terms of, frequency response.
Let's suppose for a moment a speaker design is strictly minimum phase, so time domain performance can be inferred from frequency and phase. The problem would still be that time domain performance is not displayed in a format that makes it easy to visualize. I want to know how many dB of decay at LF as a function of time. Phase vs frequency plots do not directly make that easy to see, or to design to a given specification.
And if the system is not minimum phase, then there would not be enough information in frequency and phase plots to infer time domain performance.
In any case, it is part of human nature that human brains tend to think in terms of what is readily visible. If only frequency and phase are visible, many people will not invest much effort into design for time domain performance, or even realize that it may be very important for some speaker applications.
To put it another way, what we have now for speaker design appears to be a lot like SPICE was for circuit analysis back in the old days before transient analysis was added to it. And many people here in fact use SPICE, with transient analysis probably being the most commonly used feature.
But for speaker design, we are still stuck in the old days of frequency mode analysis. Too bad. Makes it harder, more time consuming and more costly, although not impossible, to design to time domain specifications.
I'm saying that available design tools that people actually use focus on designing for, and displaying predicted performance graphs in terms of, frequency response.
Let's suppose for a moment a speaker design is strictly minimum phase, so time domain performance can be inferred from frequency and phase. The problem would still be that time domain performance is not displayed in a format that makes it easy to visualize. I want to know how many dB of decay at LF as a function of time. Phase vs frequency plots do not directly make that easy to see, or to design to a given specification.
And if the system is not minimum phase, then there would not be enough information in frequency and phase plots to infer time domain performance.
In any case, it is part of human nature that human brains tend to think in terms of what is readily visible. If only frequency and phase are visible, many people will not invest much effort into design for time domain performance, or even realize that it may be very important for some speaker applications.
To put it another way, what we have now for speaker design appears to be a lot like SPICE was for circuit analysis back in the old days before transient analysis was added to it. And many people here in fact use SPICE, with transient analysis probably being the most commonly used feature.
But for speaker design, we are still stuck in the old days of frequency mode analysis. Too bad. Makes it harder, more time consuming and more costly, although not impossible, to design to time domain specifications.
Good point. And according to Bob Katz there is some bass localization as well, which mitigates towards using stereo subwoofers, if possible.
The Gedlee multiple sub approach has a lot going for it, not least that the measurements show it works. I've got one of those servo feedback sealed box subs, and one day I'll get it working...
I should point out that I got into audio when flat earthism was in full swing and my first high end experience was celestion SL-600s so I have a natural anti-port mentality. Small English living rooms also make high bass output unnecessary. And sadly a huge number of commercial designs use underdamped designs with a big peak at 60-80Hz then falling off rapidly. I'll take a 100Hz F3 and second order roll off thank you. My opinions may change if I ever get said sub actually working.
But I have realised ports are not evil if done right. Almost no one does them right so I run screaming back to sealed boxes.
I should point out that I got into audio when flat earthism was in full swing and my first high end experience was celestion SL-600s so I have a natural anti-port mentality. Small English living rooms also make high bass output unnecessary. And sadly a huge number of commercial designs use underdamped designs with a big peak at 60-80Hz then falling off rapidly. I'll take a 100Hz F3 and second order roll off thank you. My opinions may change if I ever get said sub actually working.
But I have realised ports are not evil if done right. Almost no one does them right so I run screaming back to sealed boxes.
- Status
- Not open for further replies.