Markbakk,
As a matter of fact the frequency response -for an experienced eye that is- already tells us most of that story. Impulse, frequency/phase and CSD are all mathematically related . If you have one, you have all the others.
But hey, I am preaching to the choir here, innit?
As a matter of fact the frequency response -for an experienced eye that is- already tells us most of that story. Impulse, frequency/phase and CSD are all mathematically related . If you have one, you have all the others.
But hey, I am preaching to the choir here, innit?
They are indeed mathematically related. But isn't that only valid in a perfectly (un-)linear world. One could think of heat etc. for instance making some very strange things happen.
If one sweep steady state sinus slowly to get the FR and then calculate the impulse response and compare it to a measured actual impulse I would suspect that the impact of a pulse will trigger other non-linearites than the sweep and thus - not be the mathematically correspondence to the FR?
What do you say?
//
If one sweep steady state sinus slowly to get the FR and then calculate the impulse response and compare it to a measured actual impulse I would suspect that the impact of a pulse will trigger other non-linearites than the sweep and thus - not be the mathematically correspondence to the FR?
What do you say?
//
I think first we need to get past the idea that transient response simply means it can't do its job fast enough, and instead identify those things (or whatever is happening) in terms that relate to the things Boden and others have described.
I'm afraid not. This is just something I have been pondering for a while. My math is not stellar in any way but I acknowledge the relations so I don't in any way question Forier or Laplace or any other guru. Just that I have a hunch that the jerk in an impulse just exposes non-lineareties that simply don't show up at sinus and therefore, the relation might not hold in reality. This is just speculation from my side so far. I would like to do some test and measurements some day but I also respect the challenge to make them in a correct way which I don't think is easy.
For this reason, I would like to try to playback a real impulse and record it and from this, create the needed EQ to make the impulse look good. And then play music. I wonder how that sounds? And would a sweep be flat now?
//
For this reason, I would like to try to playback a real impulse and record it and from this, create the needed EQ to make the impulse look good. And then play music. I wonder how that sounds? And would a sweep be flat now?
//
Last edited:
I don't think so - I cant on top of my head remember any software that does that as easy as e.g. REW does it in the other/normal direction (sweep -> FR -> impulse). But I would like to be told differently. I am a low-power user of rePhase 🙂
//
//
Last edited:
Maybe the good ol square wave (band limited) is the way to go after all 🙂 if one can do as good as ESL-63 we are fine I suppose!?
Attached a square wave limited to 20-20k with 2nd order filters.
//
Attached a square wave limited to 20-20k with 2nd order filters.
//
Attachments
Last edited:
Yes. It was my assumption the previous poster was intimating that conventional drivers were inertia limited.
My assumption is that air is lighter so inertia is less.
I multiquited the wrong post,
But Charles has a point I did not consider, thermal inertia.
My question would then be, what can produce sound with less inertia than dynamic drivers, ESL, plasma drivers?
Do field coil drivers have an advantage here?
I do not pretend to be a master of the mathematics involved... but I do remember reading that it is very difficult to produce and measure an actual impulse from a speaker. In order for the S/N ratio to be low enough to be useful, the pulse must have a very high SPL, above the linear range of most loudspeakers. I am paraphrasing this, and probably filtering it through my somewhat limited Fourier understanding. The upshot is that a 95 dB swept sine signal can duplicate a 95 dB impulse, but if you measured a 95 dB impulse, there may not be enough data there to convert it into a useful frequency response.
thoughts?
j.
Attachments
Yes, the crest factor of a pulse is not very advantageous for being used as measurement stimulus. That's why other types of stimuli are used and the impulse response derived from the responses to these.
If you want to achieve really high cutoff frequencies you could use any of these principles but you would need a specialised supertweeter which will most probably be much smaller that your average 1" dome. Just don't expect a driver to be capable of reproducing from 2 to 100 kHz equally well. But then you will again need a crossover that is capable of integrating that tweetrer/supertweeter arrangement such that the impulse is perfect. And it will only be perfect on one axis.
Regards
Charles
If you want to achieve really high cutoff frequencies you could use any of these principles but you would need a specialised supertweeter which will most probably be much smaller that your average 1" dome. Just don't expect a driver to be capable of reproducing from 2 to 100 kHz equally well. But then you will again need a crossover that is capable of integrating that tweetrer/supertweeter arrangement such that the impulse is perfect. And it will only be perfect on one axis.
Regards
Charles
Really, check out Fincham. They did that, they've been there at KEF from 1975 onwards. Even to the part where short-period overload came into play. Jim and Charles: the pulse level indeed caused trouble getting enough results IIRC.
To reproducing square waves: the interesting thing is that the phase relationship is quite important from an electronic POV. If our hearing system is able to discern that phase relation however is questionable to the least.
To reproducing square waves: the interesting thing is that the phase relationship is quite important from an electronic POV. If our hearing system is able to discern that phase relation however is questionable to the least.
If cone breakup depends on force exerted by coil (that seems legit), and force is I*B*L, then there should be a force threshold below which it doesn't breakup, and above which it does.
Then spreading the energy of the pulse in a sweep will not measure this.
I'm not going to measure this, so I'll offer some handwaving: personally I would:
Play Sheffield Drum & Track pretty loud.
Record voltage on woofer terminals with the soundcard, calibrated in volts with a scope.
Extract the highest sharpest most dynamic peak from it ; this has to be done after crossover filtering so you know what signal the woofer actually gets, it's no use looking at slew rate of high frequency if the woofer never sees that signal. An alternative would be to apply a digital filter that corresponds to the crossover.
Take note of peak values for scale
Construct a pulse with the same slew rate as the drum hit, for example a raised cosine, or a wavelet.
Playback and record the acoustic output, either outdoors or in an anechoic room.
Average the hell out of it to get rid of the noise.
Now that wouldn't give much useful information, so another signal has to be added, for example a very low amplitude high frequency sine wave.
Now plot the amplitude of the recorded HF sine wave according to the time position in the simulated drum kick pulse.
Without the pulse you should get a constant amplitude. When the coil force creates nonlinear effects there should be a change in amplitude of the HF tone.
Cone breakup is basically a linear effect.
I feel as though phase_accurate implied this in post #79, whereas markbakk implied a non-linearity in post #78, though I may be mistaken?
I feel as though phase_accurate implied this in post #79, whereas markbakk implied a non-linearity in post #78, though I may be mistaken?
Cone breakup behavior can obviously be approached with linear functions. Be it rather complex ones, you’d have to solve the wave equation on a point in space from an out of shape oscillating conical surface. Bell modes would be doable 😉 but I think most engineers would grab a BEM tool for this.
The breakup itself can be considered as elastic. It’s obviously not only acceleration and cone mass but also speed and acoustic impedance working on any point of the cone. Combine that with the elastic properties of every point on the cone and presto! Seems also linear, but not really first order math either.
WRT sweep, impulse or level required, I think the small differences in outcome vanish in the noise of the measurement or that of our ears/listening environment.
The breakup itself can be considered as elastic. It’s obviously not only acceleration and cone mass but also speed and acoustic impedance working on any point of the cone. Combine that with the elastic properties of every point on the cone and presto! Seems also linear, but not really first order math either.
WRT sweep, impulse or level required, I think the small differences in outcome vanish in the noise of the measurement or that of our ears/listening environment.
I've had a massive headache for the last couple days (and still have it), so my brain's not processing things the best right noe. But I read the review. It's basically measurements of the driver's parameters up to its limits, quantifying non-linearities within them. Unfortunately things aren't separated into individual components. Overall the method seems great for system design purposes, just not identifying the exact underlying causes contributing to them. This is like having Qts - it's enough to design with, but it doesn't tell the whole story.
I hate how things disappear online. It always seems to be the most useful things too... I'll check out out remaining docs soon, when I've recovered
What do you mean by radiation impedance? Resistance of air to diaphragm movement? Higher with a horn?
Yes, but we're talking about one of the things which cause a driver to respond non-linearly. Attenuating the signal.
The causes for attenuated transients are many, but one of them is the motor not being able to do its job fast enough (slippage, post 58). Most often this accounts for only a small amount. Quanti/qualifying it is my aim
Last edited:
Impulse and sine sweep will be the same. Any non-linearities will show up equally in both.
The thing i believe is overlooked when thinking about capturing either frequency response, or impulse response, is both measurements need as wide a bandwidth as possible to allow for good FFT math to gives valid measurements.
When bandwidth is limited, phase is difficult to pin down.
[(by frequency response i mean magnitude (what most folks simply call freq response) and phase.]
So for freq response, we want 20 to 20kHz sine sweeps or noise.
And for impulse response, we want a pulse that includes spectral energy from every freq, 20-20k.
A very fast pulse...traditionally a gun shot (blanks), a balloon pop, etc.
I've used a Dirac pulse digitally....a single sample containing the entire spectrum many folks are familiar with.
If you use any of those pulses for just a subwoofer, at the same drive level you would use the pulse for a full-range speaker, all you will get a pretty quiet 'thump'.
For the tweeter, you'll get a nice loud 'crack'.
Mid, in between of course 🙂
Point here is, if you want to hear "speed", it's a function of freq response....
Perfect freq response = perfect transient (impulse) response = equals perfect "speed response" to coin a BS term Lol
Very true.
Although a woofer's square wave will only look good when it's response extends high enough to provide the needed harmonics.
To add a little credence to that and also to my prior post....here's a set of square waves i took from a DIY MTM, designed to run from 100Hz up.
