What I did was to put a microphone half way between the centers of a woofer and tweeter pair put a sine wave of a particular frequency into each alternately withought moving the microphone, and measure the total time of flight difference between the two.
As I stated in this particular instance the differences were 14 and 22 mm. at 1.5k and 3kHz.
It is certainly true that theoretically you can determine non minimum phase behavior from the impulse response, but a wonder if noise and various variables of microphones and digitation etc. are degrading the signal to the extent where the information about shift in acoustic center is being lost or being made ambiguous.
rcw
As I stated in this particular instance the differences were 14 and 22 mm. at 1.5k and 3kHz.
It is certainly true that theoretically you can determine non minimum phase behavior from the impulse response, but a wonder if noise and various variables of microphones and digitation etc. are degrading the signal to the extent where the information about shift in acoustic center is being lost or being made ambiguous.
rcw
So the difference between the ACs acually shifted by 8mm right ?
This is IMO negligible since this is only 23 us and this is nothing compared to the duration of one period at 1.5 kHz (666us) or 3 kHz ( 333 us).
Using the right crossover topology wirh these drivers you could still end up with a system that is worlds apart in terms of time coherency to a system using one of the standard crossover types.
Regards
Charles
This is IMO negligible since this is only 23 us and this is nothing compared to the duration of one period at 1.5 kHz (666us) or 3 kHz ( 333 us).
Using the right crossover topology wirh these drivers you could still end up with a system that is worlds apart in terms of time coherency to a system using one of the standard crossover types.
Regards
Charles
rcw said:
This sounds like the differences in excess-delay between the drivers due to the widely different driver lowpass Fc and the likely varying lowpass slopes between the two on top of the AC offset. It reduces to the minimum-phase differences between the two coupled with the actual AC offsets. The excess-delay will be different between two drivers at the same frequency and altering the frequency moves to points where the rotation caused by the excess-delay is different again. The tweeter rotation probably changes little between these two frequencies whereas the midrange or certainly a woofer is in the range of larger changes in delay. The larger driver is not near its point of nearly constant delay, the tweeter probably is.
There may a minor issue with mics, but I doubt that it's significant with a properly calibrated mic. Noise becomes an issue primarily at the low end of measurements due to the relatively good S/N ratio attainable by an MLS system, especially if one averages multiple impulse measurements. I always use an 8-MLS series measurement that will, if I recall, provided an extra 3db of noise reduction for every doubling of the measurements, so I should be getting an additional 9db improvement. Generally even single measurement methods are very good in a reasonably quiet environment. One could always measure with MLS in an anechoic chamber to make the point moot. I'd bet big manufacturers do just that. The low end issue has to do with window width/type for resolution and precision. But since the offset is related to the lowpass slope delay, the low end has very little influence in most AC offset measurement. It's possible to get a good model of the low end highpass anyway, using T/S measurements.
The other benefit of MLS systems is when using one in two-channel mode, the cross-correlation of the MLS used to extract the impulse response effectively eliminates all upstream influence from the feedback sampling point. The influences of any preamp, amp, interconnect and cabling are all effectively eliminated.
The remaining issues, such as calibrating for the feedback cables and cards used are minimal to almost insignificant, but it's always possible to calibrate this to ensure that it's insignificant. That's only a concern when working towards absolute AC, anyway. For relative AC offset, since the same system is used to measure both drivers, the same low-level errors would be in both, so it would effectively "cancel out".
Dave
The amount of shift in the acoustic center I described accounts for an excess of about 18 degrees more than you would expect from a 100Hz. -5kHz. Butterworth bandpass filter that the driver approximates.
This is admittedly not a huge amount, but it might be getting lost somewhere in the measurement tolerances, and may lead to errors in modelling.
rcw.
This is admittedly not a huge amount, but it might be getting lost somewhere in the measurement tolerances, and may lead to errors in modelling.
rcw.
rcw said:
The AC is not shifting for the AC point considered to be the point of minimum-phase in the driver. Modeled summed response shows exceedingly good correlation with subsequent measurements throughout the XO area whether using measured data directly or using modeled SPL with HBT-generated phase when done properly easily up to 20KHz. All CAD software I've ever used assumes a fixed AC and minimum-phase raw driver measurements. All have exhibited excellence correlation. This could not occur were the AC not fixed.
Dave
Interesting conversation here, but I'd like to back up and ask the original poster what exactly he is looking for? You use the phrase "this time coherency thing" and by this do you mean time alignment, which is usually used in the context of trying to have a phase-linear design, but not necessarily. In fact, your second order "in phase" electrical filters are not expected to be linear phase, so it somewhat contradicts your first question.
Or are you simply asking how to "time align" the drivers so that they follow your design intent?
Pete B.
Or are you simply asking how to "time align" the drivers so that they follow your design intent?
Pete B.
Hi rcw,
Likely your sine based microphone measurements were a steady resultant, being the summation in time, of entire cone aperture and baffle edge related components; this in a manner which would change with both microphone angle away from driver axis and with baffle shape/size, and thus be something which would not automatically be revealed by short period impulse analysis.
Your measurements are however - factual - and must especially be related to audible observations by those who use extended range mid-bass and broad mid-band drivers, where electrical waveform compensation, from t=0 (BSC or EQ) can counter *steady sine* measured effects which do not arise until t=x.uS, after surface waves reach a baffle edge or radiated waves interact within a cabinet or room, whereby there appears to be a listening position satisfaction of the linear phase requirement -
"Linear phase reduces to a pure time delay which introduces no transient distortion." at the beginning of John K's link above.
WITHOUT there any longer being a satisfactory impulse (dynamic/ transient) response from t=0 !
From Twisted85's original photographs does the tweeter dome not need to be better aligned with the mid-bass domes, such that the waveguide idea is sensible for preserving dynamic coherence, due to crossover impulse output still starting from t=0 ?
Cheers ........ Graham.
Likely your sine based microphone measurements were a steady resultant, being the summation in time, of entire cone aperture and baffle edge related components; this in a manner which would change with both microphone angle away from driver axis and with baffle shape/size, and thus be something which would not automatically be revealed by short period impulse analysis.
Your measurements are however - factual - and must especially be related to audible observations by those who use extended range mid-bass and broad mid-band drivers, where electrical waveform compensation, from t=0 (BSC or EQ) can counter *steady sine* measured effects which do not arise until t=x.uS, after surface waves reach a baffle edge or radiated waves interact within a cabinet or room, whereby there appears to be a listening position satisfaction of the linear phase requirement -
"Linear phase reduces to a pure time delay which introduces no transient distortion." at the beginning of John K's link above.
WITHOUT there any longer being a satisfactory impulse (dynamic/ transient) response from t=0 !
From Twisted85's original photographs does the tweeter dome not need to be better aligned with the mid-bass domes, such that the waveguide idea is sensible for preserving dynamic coherence, due to crossover impulse output still starting from t=0 ?
Cheers ........ Graham.
Guys, I would be interested in an answer to whether a filter can introduce a real time delay?
As I said before, I didn't think it could, and that it is just a phase shift or slope. But people do keep talking about delay from filters, and the acoustic source does appear to change with group delay.
Can someone help me better understand the difference between a phase shift like that of a filter, and a delay?
I understand the delay part, and that it will produce a phase shift. But what is the exact behavior of a phase shift or slope when not caused by a true delay?
As I said before, I didn't think it could, and that it is just a phase shift or slope. But people do keep talking about delay from filters, and the acoustic source does appear to change with group delay.
Can someone help me better understand the difference between a phase shift like that of a filter, and a delay?
I understand the delay part, and that it will produce a phase shift. But what is the exact behavior of a phase shift or slope when not caused by a true delay?
Excuse me if this appears to simplistic.Tenson said:
An analog filter definitely does introduce a real time delay. That time delay is not constant at all frequencies, that may the reason for some of the confusion. Phase shift at a frequency and delay are the same thing viewed from different perspectives.
An analog electrical filter and an acoustic "filter" definitely introduce time delays. Drivers are essentially inherently filtering devices as well, they are all a bandpass device. The time delay for both can be represented as a curve that is commonly called the group delay. It is simply a graph of the delay at each frequency in the graph. The delays are real.
For example, a lowpass (electrical or acoustic) has no delay at DC (0 Hz) the point referred to as "constant" delay, since at that point the slope of the curve is essentially flat. As frequency increases there is a point where the graph starts to curve down. This is due to the increasing amount of delay at each frequency. A highpass is the inverse. At infinity, the delay is 0. As frequency decreases, the delay increases.
For driver/XO combination, the only thing that really matters is the acoustic response of the combination, that is, the driver acoustic output when connected to a crossover.
Maybe the easier way to think about it is to consider how a driver and a filter operate. A driver only outputs an acoustic signal when a current if flowing through the coil. In order for that to occur, a voltage is applied to the coil. For a direct connection, the filter effect of the motor and the mechanic parts control the output.
If the voltage applied to the coil is delayed at some frequency, then the response of the driver is delayed equivalently. In order to control the driver output, filters are used that alter the voltage frequency "profile" applied to the driver. This has both voltage amplitude and phase (delays) such that when applied to the driver inputs, the resultant output is the target response.
Dave
rcw said:
I know I'm a little late to reply . Been on vacation. But I guess you failed to read one of my earlyer posts, "The all pass response includes all aspects of the crossover required to match the acoustic targetes for the crossover." , or missunderstood. By acoustic targets it is meand to shape the drivers response to match the desired target, on axis.
I would also point out that the tweeter my measurements were made relative to a tweeter to can be modeled as a high pass filter at 1.2kHz. This means that an additional 20degrees of phase shift must be added to the aforementioned 18 degrees, i.e. I now have 38degrees of phase shift that I cannot account for in any other way than to suppose that it is a result of the acoustic center moving due to cone decoupling.
rcw.
rcw.
Hi rcw,
This is the only photo I have of a B200 which was modified in February to counter the very effect you report. (Click to enlarge, then click on that too.)
The driver is here mounted on an experimental baffle (to check several OB and widerange reproduction aspects though about to be scrapped/rebuilt) and running as a directly connected wide-range, except for simple series C bass cut, with an additional 10 ohm damping resistor across the driver terminals to limit mechanical resonance due to C impedance introducing series tuning and increasing Qes at Fs. (The amplifier is low ohm capable and all drivers are separately wired back to its output terminals.)
The centre dustcap was replaced by a range of simple paper cone sizes from approx 3/4" to 4" in length attached directly to the voice coil former. Reproduction was judged by ear (not equipment) and repeated checks for the best sounding cone dimension eventually led to one of length of 65mm (#2.6") being made and used.
My judgement of B200 reproduction had been that the transduced phase shift (voice coil impedance/cone shape and mechanics) had effectively shifted the acoustic centre so far back at high frequencies where the underhung voice coil still transduces efficiently, that there had been a within band phase inversion which sounded as dissonant as if an out of phase tweeter were permanently connected.
Using another tweeter could not fully correct this without filtering the B200, which then affected wanted reproduction.
This new centre cone goes towards compensating for the AC shift at hf you mention, and reproduction sounds so much more time coherent and pleasant to listen to as a result.
The foam fingers disperse some hf, dissipate some otherwise symmetrical air-side wave motion, and improve the horizontal radiation pattern.
Cheers ......... Graham.
This is the only photo I have of a B200 which was modified in February to counter the very effect you report. (Click to enlarge, then click on that too.)
The driver is here mounted on an experimental baffle (to check several OB and widerange reproduction aspects though about to be scrapped/rebuilt) and running as a directly connected wide-range, except for simple series C bass cut, with an additional 10 ohm damping resistor across the driver terminals to limit mechanical resonance due to C impedance introducing series tuning and increasing Qes at Fs. (The amplifier is low ohm capable and all drivers are separately wired back to its output terminals.)
The centre dustcap was replaced by a range of simple paper cone sizes from approx 3/4" to 4" in length attached directly to the voice coil former. Reproduction was judged by ear (not equipment) and repeated checks for the best sounding cone dimension eventually led to one of length of 65mm (#2.6") being made and used.
My judgement of B200 reproduction had been that the transduced phase shift (voice coil impedance/cone shape and mechanics) had effectively shifted the acoustic centre so far back at high frequencies where the underhung voice coil still transduces efficiently, that there had been a within band phase inversion which sounded as dissonant as if an out of phase tweeter were permanently connected.
Using another tweeter could not fully correct this without filtering the B200, which then affected wanted reproduction.
This new centre cone goes towards compensating for the AC shift at hf you mention, and reproduction sounds so much more time coherent and pleasant to listen to as a result.
The foam fingers disperse some hf, dissipate some otherwise symmetrical air-side wave motion, and improve the horizontal radiation pattern.
Cheers ......... Graham.
PB2 said:
Sorry, my questions are probably a bit vague as I'm a little confused about these things and different terms.
The thing I'm after, is placing the drivers so that the signal from each driver would reach the measurement/listening spot at exact same moment (or as close as possible in this case). And this means Time Aligned?
Just by looking at the speaker at the moment, I'm guessing that the tweeter AC would be closer to the measurement point than the woofers. And this should be seen in the step response, right?
So I'm thinking of moving the tweeter further back with a small DIY waveguide and at the same time maybe achieve a better directivity characteristics.
I tried to generate the Step response with SoundEasy yesterday, but probably didn't do it right.
The only thing the manual refers is "Step response is calculated from Hilbert-Bode transfrom".
I made MLS far field measurement from 1meter, near-field measument of the woofer and port, and merged all these together as I've always done. Then I tried to use the HB transform, but couldn't get the phase to match at all above the near-/farfield merge point. (even with different slopes/delay settings)
What did I do wrong?
Twisted85 said:
Sorry, my questions are probably a bit vague as I'm a little confused about these things and different terms.
The thing I'm after, is placing the drivers so that the signal from each driver would reach the measurement/listening spot at exact same moment (or as close as possible in this case). And this means Time Aligned?
Just by looking at the speaker at the moment, I'm guessing that the tweeter AC would be closer to the measurement point than the woofers. And this should be seen in the step response, right?
So I'm thinking of moving the tweeter further back with a small DIY waveguide and at the same time maybe achieve a better directivity characteristics.
I tried to generate the Step response with SoundEasy yesterday, but probably didn't do it right.
The only thing the manual refers is "Step response is calculated from Hilbert-Bode transfrom".
I made MLS far field measurement from 1meter, near-field measument of the woofer and port, and merged all these together as I've always done. Then I tried to use the HB transform, but couldn't get the phase to match at all above the near-/farfield merge point. (even with different slopes/delay settings)
What did I do wrong?
A couple of things. First, If you want the signal from the different drivers to arrive at the same time you need to align the acoustic centers. Call that anything you like. Second, If you want an accurate step response from the system then you must design it with a crossover that introduces no phase distortion. I refer to these as transient perfect crossovers. Third, the other option to crossover design is the Spica approach I discussed at the
botton of my web page. I have rewritten this a little because my recollection of the Spica crossover was in error. Thanks to a form reader for pointing this out to me. The corrections don't change the outcome, however. This type of crossover isn't transient perfect but is much closer than standard crossovers. Fourth, you can not use SoundEasy to compute the step response of a system unless A) you measure the system response and B) the system is minimum phase (that is, uses a correctly implimented transient perfect crossover.) The reason the system must be minimum phase is beacuse in SE the step response in the measurement screen is computed by first perfroming a Hilbert-Bode transformation which generates the minimum phase from the amplitude response. Thus if the measured system response is not minimum phase (which will be the case for all systems that do not use a correctly implimented transient perfect crossover) the step response computed in SE will not be representative of the step response of the system.
This has nothing to do with the MTM I wanted to discuss earlier, but also has to do with phase matching. I wanted to play with the Peerless 830970 2" widerange driver as it looked promising and was very cheap, so I bought a pair for a little testing.
I had also a pair of Seas P21REX 8" woofers lying around, so I thought I would try to match those to the Peerless driver.
This is just an experimental project about phase matching and how the widerange drivers sound at their best (I'm thinking of using them in a car)
So I made separate enclosures for each that can be put one on the other and also allows moving the tweeter section back and forth relatively to the woofer. I made also very large roundings around the Peerless enclosure to avoid harmful baffle diffraction effects.
I set the enclosures for the measurements so, that the enclosure front surfaces were on the same depth. The peerless driver was flush mounted to the surface, the seas woofer wasn't. I lifted the enclosure high enough to reach 5,7ms time window, and placed the mic on the tweeter axis.
After all the measurements were done, I began with the XO design, and here how it looks now:
Frequency Responses
The target slopes are 18db/oct. The peerless driver has quite bad bump at 2Khz which should be notched down. I thought 400Hz could be a good choice for xo point? This wouldn't stress the widerange too much, and the 8" woofer directivity shouldn't be a problem at 400Hz.
(After taking the screenshots, I raised the woofer lowpass from 300Hz to 340Hz for better total frequency response)
Is that 20Khz peak a problem? Or is it so high frequency that it doesn't matter?
When I made the hilbert-bode transforms, I had to use 3,7cm delay for the woofer, and 0,9cm for the tweeter to get the phases match. So this would mean, that the difference in acoustic centers are around 2,8cm? And by pushing the tweeter enclosure back, they would be aligned at the measuring point.
Here is how the phase responses looks, without any change to the tweeter location, and with the Xo described earlier:
Phases
They seem to align quite well, right?
Now here comes the thing I don't understand. When I try to move the tweeter position in SoundEasy to something let's say Z +2,5cm (more distance to the mic), I would have thought that the phase match would get better, but it doesn't, it goes worse?
I would also like to know the difference in the MLS measurement screen, between the two window types, rectangular and blackman-harris? In every SE guide, the blackman-harris is suggested. But when I make my measurements (typically with 4-5,8ms window), I get much better result with rectangular window near the lowest window frequency. With blackman-harris, the response starts to slope down before the window edge. I'm a bit confused with this.
I had also a pair of Seas P21REX 8" woofers lying around, so I thought I would try to match those to the Peerless driver.
This is just an experimental project about phase matching and how the widerange drivers sound at their best (I'm thinking of using them in a car)
So I made separate enclosures for each that can be put one on the other and also allows moving the tweeter section back and forth relatively to the woofer. I made also very large roundings around the Peerless enclosure to avoid harmful baffle diffraction effects.
I set the enclosures for the measurements so, that the enclosure front surfaces were on the same depth. The peerless driver was flush mounted to the surface, the seas woofer wasn't. I lifted the enclosure high enough to reach 5,7ms time window, and placed the mic on the tweeter axis.
After all the measurements were done, I began with the XO design, and here how it looks now:
Frequency Responses
The target slopes are 18db/oct. The peerless driver has quite bad bump at 2Khz which should be notched down. I thought 400Hz could be a good choice for xo point? This wouldn't stress the widerange too much, and the 8" woofer directivity shouldn't be a problem at 400Hz.
(After taking the screenshots, I raised the woofer lowpass from 300Hz to 340Hz for better total frequency response)
Is that 20Khz peak a problem? Or is it so high frequency that it doesn't matter?
When I made the hilbert-bode transforms, I had to use 3,7cm delay for the woofer, and 0,9cm for the tweeter to get the phases match. So this would mean, that the difference in acoustic centers are around 2,8cm? And by pushing the tweeter enclosure back, they would be aligned at the measuring point.
Here is how the phase responses looks, without any change to the tweeter location, and with the Xo described earlier:
Phases
They seem to align quite well, right?
Now here comes the thing I don't understand. When I try to move the tweeter position in SoundEasy to something let's say Z +2,5cm (more distance to the mic), I would have thought that the phase match would get better, but it doesn't, it goes worse?
I would also like to know the difference in the MLS measurement screen, between the two window types, rectangular and blackman-harris? In every SE guide, the blackman-harris is suggested. But when I make my measurements (typically with 4-5,8ms window), I get much better result with rectangular window near the lowest window frequency. With blackman-harris, the response starts to slope down before the window edge. I'm a bit confused with this.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.