Jim,
Did you look at Dan Wiggins test results? He added mass and the driver still responded to impulse with the same speed. Of course it changes the frequency response, but not necessarily it's dynamic ability.
Qts is the measure of a driver's propensity to resonate at its resonant frequency, not a measure of its dynamic ability. Properly matching a driver to an enclosure is of far greater importance in determining the transient response. A high Qts driver can be quite dynamic and have good transient response, however, a high Le driver cannot.
Did you look at Dan Wiggins test results? He added mass and the driver still responded to impulse with the same speed. Of course it changes the frequency response, but not necessarily it's dynamic ability.
Qts is the measure of a driver's propensity to resonate at its resonant frequency, not a measure of its dynamic ability. Properly matching a driver to an enclosure is of far greater importance in determining the transient response. A high Qts driver can be quite dynamic and have good transient response, however, a high Le driver cannot.
xplod1236 said:
I'm not sure what you mean by "too tight" and it would really depend on what you are asking your sub to do. If it's operating as a true subwoofer with a very low upper frequency cutoff, then a higher Le of 3mH or even higher can work well in a proper enclosure because in those very low frequencies you are not asking it to be very dynamic, just to move a lot of air and provide SPL. If you are talking about a sub that you want to really act like a woofer (not a sub woofer) and give you quality bass performance from say 50 and up then look for Le's of around 1mH, like pro sound woofers have, but you are going to sacrifice that very low end unless you use a big bass horn.
If you get a dual coil sub, then be sure that you can wire it in parallel, because the Le is 4 times as high if you wire the coils in series.
High frequency response and good transient response are exactly the same thing. If this is not true someone needs to re-educate me on linear continuous time system (Fourier) analysis.
Deja Vu, dammit!
Who knows what the "transient" response of a woofer really means? By definition a frequency response graph can be generated by a transient, specifically: a unit impulse or a DC step. It allows us to see how good the frequency response is. All sounds can be said to be transients: now you hear them, now you don't.
I find adireaudio website link nearly every other week now, and it's already been criticised and somewhat discredited in another thread (by those who probably happened to take physics as a subject some time in their lives).
If anyone wants fast transient response, guess what: they've invented this new type of woofer! It's called a tweeter, and it has really tight bass, super fast transient response, low mass, low inductance and more! 🙄 😱
Like I've said before, the 2 nearest things to a graph that shows a woofer's transient response are:
Waterfall plots, AKA: cumulative spectral decay plots.
cumulative = of an accumulation; accumulating; summing
spectral = of a spectrum; of a range of frequencies
decay = to fade away; diminish; get quieter after a period of time
plot = story-line; Hollywood excuse for a story-line 😀 😉
Waterfall plots are used mostly for mids and tweeters to show where their break-up modes are, and are generally made from step-response measurements. Group-delay plots are usually theoretical, and can be calculated from the phase response. Subwoofer Simulator can show group delays, and I thought that winISD could too but maybe I'm wrong.
CM
Who knows what the "transient" response of a woofer really means? By definition a frequency response graph can be generated by a transient, specifically: a unit impulse or a DC step. It allows us to see how good the frequency response is. All sounds can be said to be transients: now you hear them, now you don't.
I find adireaudio website link nearly every other week now, and it's already been criticised and somewhat discredited in another thread (by those who probably happened to take physics as a subject some time in their lives).
If anyone wants fast transient response, guess what: they've invented this new type of woofer! It's called a tweeter, and it has really tight bass, super fast transient response, low mass, low inductance and more! 🙄 😱
Like I've said before, the 2 nearest things to a graph that shows a woofer's transient response are:
Waterfall plots, AKA: cumulative spectral decay plots.
cumulative = of an accumulation; accumulating; summing
spectral = of a spectrum; of a range of frequencies
decay = to fade away; diminish; get quieter after a period of time
plot = story-line; Hollywood excuse for a story-line 😀 😉
Waterfall plots are used mostly for mids and tweeters to show where their break-up modes are, and are generally made from step-response measurements. Group-delay plots are usually theoretical, and can be calculated from the phase response. Subwoofer Simulator can show group delays, and I thought that winISD could too but maybe I'm wrong.
CM
I withdraw any reference I've made using the words "transient response" as it relates to this woofer discussion and replace them with "detailed, articulate, accurate bass frequency notes". I definitely don't want to get into a "fast bass" debate and will just agree to disagree whether or not a woofer is part of "transient response".
Sorry, I was probably a bit too sarcastic. I just wanted to emphasize that people shouldn't look to various meaningless descriptions like "transient", but try figure out what really makes some woofer designs better than others. If things like cone acceleration had anything to do with it, then eliminating the practice of low-pass filtering of woofers would be an obvious first step. Should we pick out a woofer with an extra-low equivalent inductance, and then put a large inductance in series as part of a passive crossover network?
I think it's not so much acceleration, but overshoot and ringing at certain low frequencies that can be problematic. Also, I find it interesting that when a woofer is placed in an oversized box, it's said to be "overdamped". I'd call that "underdamped", and I think it's time someone experimented a bit and tested my claims about group-delay on winISD or something, because you might be very surprised.
CM
I think it's not so much acceleration, but overshoot and ringing at certain low frequencies that can be problematic. Also, I find it interesting that when a woofer is placed in an oversized box, it's said to be "overdamped". I'd call that "underdamped", and I think it's time someone experimented a bit and tested my claims about group-delay on winISD or something, because you might be very surprised.
CM
CM,
I agree with you and the coils in xovers has been nagging at me since I first started believing the Le factor. Connecting VC's in parallel vs series seems to make a big sonic difference, but plopping down a big coil in front of a driver in series doesn't seem to have the same effect. What is the difference?
Ringing, I always address by damping the driver basket.
Overshoot is definitely a problem and I would think that this should show up in a driver's BL, or am I not understanding BL correctly? I thought as long as the motor's strength was sufficient to control the moving mass, then overshoot would be caused by excessive Le, since electrically the driver was fighting the changes in direction.
Jim,
I own a Maelstrom and have it in the big EBS enclosure. The low end performance is great. A friend of mine used to play bass guitar in a band and he said he had never heard anything play so low. I wouldn't, however, try to make it play like a woofer though despite Adire's claim that it will play up to 1Khz. As a sub, it's worth every bit of the $300 .
I agree with you and the coils in xovers has been nagging at me since I first started believing the Le factor. Connecting VC's in parallel vs series seems to make a big sonic difference, but plopping down a big coil in front of a driver in series doesn't seem to have the same effect. What is the difference?
Ringing, I always address by damping the driver basket.
Overshoot is definitely a problem and I would think that this should show up in a driver's BL, or am I not understanding BL correctly? I thought as long as the motor's strength was sufficient to control the moving mass, then overshoot would be caused by excessive Le, since electrically the driver was fighting the changes in direction.
Jim,
I own a Maelstrom and have it in the big EBS enclosure. The low end performance is great. A friend of mine used to play bass guitar in a band and he said he had never heard anything play so low. I wouldn't, however, try to make it play like a woofer though despite Adire's claim that it will play up to 1Khz. As a sub, it's worth every bit of the $300 .
Do I understand this correctly?
I've picked up a lot of this audio theory stuff from guys like you over the years and then I try to apply what I know from enginnering school. Sometimes there is a disconnect so help me fill in the gaps if you would. Here is some of what I think I know. To accurately reproduce any audible waveshape all audible frequencies must be reproduced equally, i.e. flat frequency response from 20 to 20kHz. Can a square wave be used to judge transient response? A square wave is just a sum of sinusoidal waves starting with the fundamental frequency plus odd harmonics. It's the higher order harmonics which give a square wave its sharp edges. If you can't reproduce higher harmonics then the perfect square wave gets distorted with rounded edges and slanted sides. When a square wave is fed into an amplifier and out from the speakers comes a square wave with slanted sides and rounded edges is that not poor transient response? If the sides of the square wave are really really sloped and the edges are really really rounded is that not really poor transient response? In my mind poor high frequency response results in rounded edges therefore poor transient response is directly proportional to poor high frequency response. Do I understand this correctly?
Obviously the goal is to equally reproduce all frequencies from 20Hz to 20kHz. If one uses a low Q, low Le woofer with low Fs then one can efficiently reproduce low and high frequency. Correct? Isn't the response curve of a low Q woofer broad and flat? If the low Q, low Le woofer can efficiently reproduce frequencie up to the kilohertz range then one can choose a horn midrange and still have lots of overlap at the crossover frequency. This will result in a flat response from about 30Hz up to several kilohertz. Add a high quality tweeter and now the system is able to faithfully reproduce any waveshape, including an inpulse with wide spectral content. Is that not what good transient response is all about?
I've picked up a lot of this audio theory stuff from guys like you over the years and then I try to apply what I know from enginnering school. Sometimes there is a disconnect so help me fill in the gaps if you would. Here is some of what I think I know. To accurately reproduce any audible waveshape all audible frequencies must be reproduced equally, i.e. flat frequency response from 20 to 20kHz. Can a square wave be used to judge transient response? A square wave is just a sum of sinusoidal waves starting with the fundamental frequency plus odd harmonics. It's the higher order harmonics which give a square wave its sharp edges. If you can't reproduce higher harmonics then the perfect square wave gets distorted with rounded edges and slanted sides. When a square wave is fed into an amplifier and out from the speakers comes a square wave with slanted sides and rounded edges is that not poor transient response? If the sides of the square wave are really really sloped and the edges are really really rounded is that not really poor transient response? In my mind poor high frequency response results in rounded edges therefore poor transient response is directly proportional to poor high frequency response. Do I understand this correctly?
Obviously the goal is to equally reproduce all frequencies from 20Hz to 20kHz. If one uses a low Q, low Le woofer with low Fs then one can efficiently reproduce low and high frequency. Correct? Isn't the response curve of a low Q woofer broad and flat? If the low Q, low Le woofer can efficiently reproduce frequencie up to the kilohertz range then one can choose a horn midrange and still have lots of overlap at the crossover frequency. This will result in a flat response from about 30Hz up to several kilohertz. Add a high quality tweeter and now the system is able to faithfully reproduce any waveshape, including an inpulse with wide spectral content. Is that not what good transient response is all about?
A square wave is all very well, but the shape of it after passing through a loudspeaker is far mode affected by the phase/frequency characteristics of the XO and drive units than by their amplitude response.
Ouroboros
Ouroboros,
Can you please elaborate about phase problems? From what I understand, phase shift starts to occur at resonance in bass-reflex enclosures. I also understand certain types of active and passive crossovers cause phase shift at the crossover frequency. Can someone give a good explanation of crossover types and phase shift or point to a good source for this info? By the way, I recently lucked out and had a used Tektronix 100MHz oscilloscope given to me! I'm really jazzed! I look forward to doing more in-depth testing of phase response and transient response, etc. with my new O'scope. Now, if I could only convert my garage to an anechoic room...
Jim
P.S. I hope you all noticed from my previous post I can't spell 'enginnering'. What a moron I am!
Ouroboros,
Can you please elaborate about phase problems? From what I understand, phase shift starts to occur at resonance in bass-reflex enclosures. I also understand certain types of active and passive crossovers cause phase shift at the crossover frequency. Can someone give a good explanation of crossover types and phase shift or point to a good source for this info? By the way, I recently lucked out and had a used Tektronix 100MHz oscilloscope given to me! I'm really jazzed! I look forward to doing more in-depth testing of phase response and transient response, etc. with my new O'scope. Now, if I could only convert my garage to an anechoic room...
Jim
P.S. I hope you all noticed from my previous post I can't spell 'enginnering'. What a moron I am!
I'm not sure what the best resource is, but fundamentally all crossover networks induce phase shifts (except for digital finite impulse response designs) in the output. That applies to first order through n-order, passive and active. Using your square wave as an example, the woofer in a three way such as you described might reproduce the fundamental and perhaps the first odd harmonic, the midrange the next several, and the tweeter from there up. If the relative phases of woofer, midrange, and tweeter have been altered by the crossover network used, then the square wave harmonic components will not be aligned at output and you won't see anything resembling a square wave.
First order networks sum acoustically at a single point to a flat phase response, so in practice the benefit is dubious. Linkwitz-Riley crossover are next best, with even cycle phase shifting throughout the frequency range. Other crossover types must be adjusted so that phases are aligned at the crossover frequency, but they will become increasingly misaligned away from that point.
First order networks sum acoustically at a single point to a flat phase response, so in practice the benefit is dubious. Linkwitz-Riley crossover are next best, with even cycle phase shifting throughout the frequency range. Other crossover types must be adjusted so that phases are aligned at the crossover frequency, but they will become increasingly misaligned away from that point.
RHosch said:
"...fundamentally all crossover networks induce phase shifts... Using your square wave as an example, the woofer in a three way such as you described might reproduce the fundamental and perhaps the first odd harmonic, the midrange the next several... If the relative phases of woofer, midrange, and tweeter have been altered by the crossover network used, then the square wave harmonic components will not be aligned at output and you won't see anything resembling a square wave..."
Having said that doesn't it make for better transient response (of stuff like drums) to use a woofer with extended high frequency response. Then the woofer can reproduce the higher harmonic components without 'handing off' to the midrange driver?
"...fundamentally all crossover networks induce phase shifts... Using your square wave as an example, the woofer in a three way such as you described might reproduce the fundamental and perhaps the first odd harmonic, the midrange the next several... If the relative phases of woofer, midrange, and tweeter have been altered by the crossover network used, then the square wave harmonic components will not be aligned at output and you won't see anything resembling a square wave..."
Having said that doesn't it make for better transient response (of stuff like drums) to use a woofer with extended high frequency response. Then the woofer can reproduce the higher harmonic components without 'handing off' to the midrange driver?
RHosch,
You've got me interested in digital finite impulse design crossovers. I am an EE but DSP is not my forte. How does one go about implementing DSP active crossovers? My guess is someone has already done this and the code exists somewhere.
Jim
You've got me interested in digital finite impulse design crossovers. I am an EE but DSP is not my forte. How does one go about implementing DSP active crossovers? My guess is someone has already done this and the code exists somewhere.
Jim
jimseko said:
Funny you should mention it. The thread "DIY DSP Engine" is about finding an appropriate DSP dev board for that very purpose. FIRs themselves are dead easy - they're just a dot product. Finding the coefficients is where the Deep Magic lies. I use ScopeFIR from Iowegian. It's a nice graphical package for $100.
Francois.
Some people obviously think so, but speaker design isn't that simplistic. What you are describing is essentially a single driver fullrange system. While there are some obvious benefits, there are also several disadvantages... higher distortion products of all types, limited SPL, restricted frequency response at the highest and lowest ends of the spectrum, power compression, etc. Some of these can be designed around... some really cannot. Of course, you might simply be suggesting to use a high bandwidth woofer in a more traditional multi-way system. It's a bit sensless to use a high bandwidth driver and then limit it's operating bandwidth with a crossover, so unless you are going to have unusually high XO points for a woofer, or a gradual (6db) slope, you won't reap the "benefits".jimseko said:
Speaker design is always a compromise. What is more audible to you... phase distortion, or harmonic distortion? Limited frequency response, or irregular polar pattern? Pick your poison, and then do your best to optimize a design within that framework.
There is plenty of computer code out there that can transform your PC into a multi-way FIR digital crossover. Be warned however, that old "always a compromise" kicker rears its head. Controlled behavior in the frequency domain (sharp cutoff, no or linear phase shift) can result in bizarre problems in the time domain... ringing, pre-echo. Nothing is ever really a free lunch.jimseko said:
454Casull said:
Subtractive FIRs (highpass = delay-lowpass, or vice versa) are the way to go. I use them in my day job, and they work just fine. You do have to make sure the passband is *very* flat, as any passband ripple will be different from 1.0, and show itself as imperfect stopband for the subtracted filter.
From a practical point of view with speakers you want to use a standard FIR for the highpass and subtract from it to generate the lowpass. Another point - the filter is subtracted not from the input signal, but from the input with delay equal to (FIR length - 1)/2.
The off-axis behaviour is surprisingly good; the models turned out much better than I'd expected, with maximum 1 dB of ripple over 1 octave at 30 degrees off axis for a mid-tweeter combo separated by 1 wavelength at the crossover frequency (15 cm @ 2200 Hz). The response irregularity was largest right at the crossover frequency, of course, and diminished rapidly away from it. This was, I might add, with a not terribly aggressive filter.
Francois.
- Status
- Not open for further replies.