Yes Just like West Virginia. It's all relative.
Was simply curious about what would happen if the peak is locked to time zero. Since the peak of the impulse is mostly the highest frequencies of the measurement, what does that do to the relative phase of the drivers. It's relative, but relative to what? Seems it doesn't matter. It works.
This afternoon I set up measurement and locked the time to the soundcard loop-back. That took the latency out of the measurement. Then ran a sweep. The impulse offset was 7.5 meters. Guess how far the mic was from the speaker.
Some interesting reading for a Sunday afternoon:
https://www.google.com/url?sa=t&rct...KDR0SQK8xibdzI0fg&sig2=WoXjh1LzXindlaN6A2-LGA
https://www.google.com/url?sa=t&rct...KDR0SQK8xibdzI0fg&sig2=WoXjh1LzXindlaN6A2-LGA
An excellent review. The kind of background that everyone discussing the issues here should have. But it doesn't really address the main issues being discussed here lately.
gedlee- As I understand, you make box loudspeakers and are some sort of professor at a university with much acoustics equipment. How about you take one specific speaker and microphone and do a plot using HOLM and then do a two tone series of plots without moving the speaker or the mic? Post the results of these two phase plots here. No one would believe me if I did this but you have GURU status and access to the means. Would add you start the two tone series one octave above the resonance frequency of the driver with that likely being within the useful bandwidth of that normal driver. Say your woofer starting at 100Hz you could plot points every 50Hz up to at least 1kHz and show us all how well HOLM and this plot agree, not.
I recall at an ALMA meeting when all the test gear manufactures lined up with their new phase plot feature and how great that was. Vance Dickeson said I want to test your systems on stage using the B&K mic and this specific speaker. They all agreed to their chagrin. Vance and the manufactures all ran this series of test and then the data was presented. You know what the main question from Vance was? Why is it every one of these test procedures results do not even closely resemble each other? The answer? Each manufacturer claimed all the others were wrong. Likely all were wrong. In this case that would include HOLM.
I recall at an ALMA meeting when all the test gear manufactures lined up with their new phase plot feature and how great that was. Vance Dickeson said I want to test your systems on stage using the B&K mic and this specific speaker. They all agreed to their chagrin. Vance and the manufactures all ran this series of test and then the data was presented. You know what the main question from Vance was? Why is it every one of these test procedures results do not even closely resemble each other? The answer? Each manufacturer claimed all the others were wrong. Likely all were wrong. In this case that would include HOLM.
SumAudioGuy
I am surprised that you responded and given your last rude response to me I am surprised that I am responding to you.
First, I was never a Prof at any univ., but I have been an adjunct at a few. I am also retired so I neither have a large array of equipment nor do I do anything that I don't want to do, and proving your point (whatever it is) falls in the category of "not interested".
I am sure that others, like Pano, (who I almost always disagree with, but would certainly trust his ability to do this test,) would be interested. But here is what I would expect to happen.
First, your test implies some sort of absolute phase that exists without a reference to something else. I have already stated that no such thing exists - phase is always relative to something and it must be clear what that something is and that it is the same something in both tests. Second, if the two tests differ, this in no way would indicate that yours was right. This will put someone (?) in the position of determining what is right and what is wrong (no thanks, that doesn't interest me either since I already know the answer.)
I already described a test at AES that concluded the exact opposite of what you are claiming at ALMA. But, at any rate, such a test of "phase" would have to have been a long time ago and as such things would well have changed in that time.
I am surprised that you responded and given your last rude response to me I am surprised that I am responding to you.
First, I was never a Prof at any univ., but I have been an adjunct at a few. I am also retired so I neither have a large array of equipment nor do I do anything that I don't want to do, and proving your point (whatever it is) falls in the category of "not interested".
I am sure that others, like Pano, (who I almost always disagree with, but would certainly trust his ability to do this test,) would be interested. But here is what I would expect to happen.
First, your test implies some sort of absolute phase that exists without a reference to something else. I have already stated that no such thing exists - phase is always relative to something and it must be clear what that something is and that it is the same something in both tests. Second, if the two tests differ, this in no way would indicate that yours was right. This will put someone (?) in the position of determining what is right and what is wrong (no thanks, that doesn't interest me either since I already know the answer.)
I already described a test at AES that concluded the exact opposite of what you are claiming at ALMA. But, at any rate, such a test of "phase" would have to have been a long time ago and as such things would well have changed in that time.
Two tone test. Let us start with a 100Hz and 200Hz two tone where the 100Hz signal is locked in phase to the 200Hz signal. This means for every zero crossing of the 100Hz signal there will be two zero crossing of the 200Hz signal with half of the zero crossing of the 200Hz signal being exactly at the same instant the 100Hz signal crosses zero. Now apply this to the "woofer" with a resonance of less than 50Hz in the box and measure the output from that woofer noting the time difference between the output zero crossing of the 100Hz signal and the 200Hz signal . This will tell us the time difference between these two signals. Repeat using 100Hz and 300Hz and so on until say 100Hz and 2000Hz.
This series of test will generate a array of time differences at 100Hz intervals. Given the known frequencies and the reference of 100Hz one can then calculate and plot either the time difference of the higher frequency compared to the lower frequency. Because phase is a measurement in degrees and measuring degrees is not straight forward, the two tone test series can provide an (assumed) reference time of zero for the 100Hz signal and the difference in time from zero with a phase locked higher frequency signal, a relative phase plot of the driver can be easily calculated and plotted. I still do it this way with several modifications but, lets stick to something which is not refutable.
The assumption here is the woofer at 100Hz in in the usable band and not out of band. Probably true for a 12" woofer in a sealed box. This means whatever is measured will show time variations at the test frequencies which can be, via math, turned into phase response. It us quite common for drivers to, in a very short bandwidth, reverse the phase of their output signal. For a 12" woofer this will pretty much always happen by 1200Hz. Minor variations in phase are very common. ±20° I would consider to be minimal and a very good driver if it was within this range over a decade of frequency or more. The very best driver I have ever seen for audio was able to stay within that range from 100Hz to 6500Hz which is amazing. Most drivers will not make a decade with many not even making two octaves.
gedlee- Is it not universally considered when a acoustic signal from a transducer increase the air pressure in front of that transducer to be of positive phase? Woofer goes out means positive phase. Phase therefore is not relative as this is universally accepted. In this discussing possibly we are calling change in phase across the working band the phase shift. This means all excess time is removed from the system based on measurement I have explained several time with zero phase assigned near the band center of the transducer. This would then allow anyone to know the zero phase frequency and the phase response relative to that frequency. Is that the relative you are avoiding or claiming does not exist? Sounds like the book "The Wood Effect" to me.
gedlee- "I already described a test at AES..." where can I read this? Date, issue number, etc. Journal of the AES is it in there or a separate paper. I want to see this a lot! Please list.
This series of test will generate a array of time differences at 100Hz intervals. Given the known frequencies and the reference of 100Hz one can then calculate and plot either the time difference of the higher frequency compared to the lower frequency. Because phase is a measurement in degrees and measuring degrees is not straight forward, the two tone test series can provide an (assumed) reference time of zero for the 100Hz signal and the difference in time from zero with a phase locked higher frequency signal, a relative phase plot of the driver can be easily calculated and plotted. I still do it this way with several modifications but, lets stick to something which is not refutable.
The assumption here is the woofer at 100Hz in in the usable band and not out of band. Probably true for a 12" woofer in a sealed box. This means whatever is measured will show time variations at the test frequencies which can be, via math, turned into phase response. It us quite common for drivers to, in a very short bandwidth, reverse the phase of their output signal. For a 12" woofer this will pretty much always happen by 1200Hz. Minor variations in phase are very common. ±20° I would consider to be minimal and a very good driver if it was within this range over a decade of frequency or more. The very best driver I have ever seen for audio was able to stay within that range from 100Hz to 6500Hz which is amazing. Most drivers will not make a decade with many not even making two octaves.
gedlee- Is it not universally considered when a acoustic signal from a transducer increase the air pressure in front of that transducer to be of positive phase? Woofer goes out means positive phase. Phase therefore is not relative as this is universally accepted. In this discussing possibly we are calling change in phase across the working band the phase shift. This means all excess time is removed from the system based on measurement I have explained several time with zero phase assigned near the band center of the transducer. This would then allow anyone to know the zero phase frequency and the phase response relative to that frequency. Is that the relative you are avoiding or claiming does not exist? Sounds like the book "The Wood Effect" to me.
gedlee- "I already described a test at AES..." where can I read this? Date, issue number, etc. Journal of the AES is it in there or a separate paper. I want to see this a lot! Please list.
Assigning zero phase near to band center of driver is an assumption.
Phase of flight time corrected acoustic response relative to driver response at terminals is primary reference.
With typical dynamic driver acoustic phase equals terminal phase at impedance peak, and at impedance minimum. Phase also matches at DC, obviously not a useful point of reference in typical applications.
The stepped tone method presented here is no different than using swept sine, only that it is indeed quite tedious to do manually. Simple proof is that swept sine test signal returns single sample pulse at t=0. Convolution of this pulse with any sine wave at any given phase returns the identical sine wave with same phase.
Phase of flight time corrected acoustic response relative to driver response at terminals is primary reference.
With typical dynamic driver acoustic phase equals terminal phase at impedance peak, and at impedance minimum. Phase also matches at DC, obviously not a useful point of reference in typical applications.
The stepped tone method presented here is no different than using swept sine, only that it is indeed quite tedious to do manually. Simple proof is that swept sine test signal returns single sample pulse at t=0. Convolution of this pulse with any sine wave at any given phase returns the identical sine wave with same phase.
I am not sure that this test means what it is claimed to mean. The zero crossings of the higher tone are obvious only when they coincide with those of the base tone, i.e. relative phase between them= 0, 2pi etc. When there is a relative phase shift between the tones the zero crossings cannot be said to be uniquely those of the higher tone because of superposition. It is not a clean test. (Oh and if there is any harmonic distortion in the signals then ALL bets are off.)
Sumaudioguy - phase is more than just positive pressure, it is a continuum of shift that is positive some of the time. This does not nail down a t=0 since it is ambiguous.
The AES was a workshop many years ago, maybe even 20 years. I have no details on it now. The paper that I references here earlier, and was posted, is a recent work that does not describe any issue like you are raising.
Sumaudioguy - phase is more than just positive pressure, it is a continuum of shift that is positive some of the time. This does not nail down a t=0 since it is ambiguous.
The AES was a workshop many years ago, maybe even 20 years. I have no details on it now. The paper that I references here earlier, and was posted, is a recent work that does not describe any issue like you are raising.
Sumaudioguy also says that HOLM does not detect polarity.
This is quite untrue. I use HOLM all the time to save me from wiring mistakes. All you have to do is look at the impulse or the phase plot. You'll see a polarity mistake right away. Just the other day I found a polarity flip in the right channel that was due to bad wiring of a 1/4" to XLR cable.sumaudioguy said:
I've checked my playback-recording chain several times. It's easy to tell if the system is inverting or not. HOLM certainly does tell you the polarity. Or Absolute Phase, as some people call it.
Barleywater
"Assigning zero phase near to band center of driver is an assumption." I already said that. What is your point?
Flight time is an air travel term. Do you mean excess time which is the time difference between signal applied and, as example, the time the signal is detected and processed in the measurement system?
"With typical dynamic driver acoustic phase equals terminal phase..." You are really confused here. Terminal phase is lost in some simulation model and not useful. Please do not confuse simulations and math models for real world events. A common error. Convolution cannot be completed without knowledge of bandwidth of the system under test, again, see Heiser and the AES paper, "Linear Distortion."
"The stepped tone method presented here..." Wrong, without the bandwidth knowledge a swept sine wave does not equal an impulse. That is such an incorrect understanding the only response is to go study some Heiser please. This assumption works with summing nodes in linear system like an opamp input but, not in unknown system like a transducer. Another common error.
Pano- yah, getting valid results is often a tedious task. Short cuts often mislead to erroneous errors. Hope the testing goes well and good luck. If you have the right equipment this can go really fast. As a note, once a higher frequency time difference is established, that higher frequency can be used for the new reference frequency (noting that additional if any time difference) possibly making higher frequency measurements easier and/or more accurate.
The two tone test does not provide a value for excess time which I have stated repeatedly. It does provide time differences caused by the transducer system relative to the reference frequency and this discussion in particular is about the phase response (time delay variation within the transducer) of the driver and not about excess time (delay) of all the parts in between. HOLM provides a phase graph which I say is useless. This test can be used to compare real results with the HOLM output to show HOLM does not work correctly.
I forget when the ALMA meeting was, maybe about 2000. Ask Vance Dickason.
The key point of this exercise is to create a phase plot for a real driver using test methods without assumptions. The only assumption here is with a 12" driver with a sealed box resonance below 50Hz will be operating in its pass band at 100Hz. That is not an extreme wild guess or horrid assumption. Beyond that it is all measurement.
gedlee- I bring this up because this is an essential part to speaker design. All the speaker designers and builders I regularly communicate with discuss this subject. Not having a decent model of the loudspeaker phase response means that cannot be added to the simulation of the crossover. That means the simulation cannot take into consideration the actual driver response. Unthinkable around here. Additionally- the testing done here establishes an acoustic center for the driver position from which all phase measurements are made. Your statement "... it is a continuum of shift..." is dead wrong. Phase of the driver can invert within 1Hz bandwidth and is therefore not continuous. This is caused by change in mode of the cone which is also well established. Any even bigger error of assumption is the phase of the output at a given frequency and radiation direction is smooth and changes in a continuous fashion. Super wrong. The main lobe of a single driver is in one phase and the adjacent lobe is typically 180° opposite. The intersection of the two lobes do not show any smooth change. Between the lobes there is a null and on either side of the null are two opposite phase signals. Like pregnancy, it is one or the other and not almost pregnant, not continuous. Linkwitz writes of this at length.
PANO- Just saw your post. I said HOLM does not work correctly providing erroneous data. I did not say it is inconsistent in its measurements. If the error made is consistent and the speakers are of the same design am certain what you say could be true. That does not mean the actual data is correct. It means HOLM can tell when a driver is mis-wired when comparing two pretty much otherwise identical speakers. Definitely handy!
"Assigning zero phase near to band center of driver is an assumption." I already said that. What is your point?
Flight time is an air travel term. Do you mean excess time which is the time difference between signal applied and, as example, the time the signal is detected and processed in the measurement system?
"With typical dynamic driver acoustic phase equals terminal phase..." You are really confused here. Terminal phase is lost in some simulation model and not useful. Please do not confuse simulations and math models for real world events. A common error. Convolution cannot be completed without knowledge of bandwidth of the system under test, again, see Heiser and the AES paper, "Linear Distortion."
"The stepped tone method presented here..." Wrong, without the bandwidth knowledge a swept sine wave does not equal an impulse. That is such an incorrect understanding the only response is to go study some Heiser please. This assumption works with summing nodes in linear system like an opamp input but, not in unknown system like a transducer. Another common error.
Pano- yah, getting valid results is often a tedious task. Short cuts often mislead to erroneous errors. Hope the testing goes well and good luck. If you have the right equipment this can go really fast. As a note, once a higher frequency time difference is established, that higher frequency can be used for the new reference frequency (noting that additional if any time difference) possibly making higher frequency measurements easier and/or more accurate.
The two tone test does not provide a value for excess time which I have stated repeatedly. It does provide time differences caused by the transducer system relative to the reference frequency and this discussion in particular is about the phase response (time delay variation within the transducer) of the driver and not about excess time (delay) of all the parts in between. HOLM provides a phase graph which I say is useless. This test can be used to compare real results with the HOLM output to show HOLM does not work correctly.
I forget when the ALMA meeting was, maybe about 2000. Ask Vance Dickason.
The key point of this exercise is to create a phase plot for a real driver using test methods without assumptions. The only assumption here is with a 12" driver with a sealed box resonance below 50Hz will be operating in its pass band at 100Hz. That is not an extreme wild guess or horrid assumption. Beyond that it is all measurement.
gedlee- I bring this up because this is an essential part to speaker design. All the speaker designers and builders I regularly communicate with discuss this subject. Not having a decent model of the loudspeaker phase response means that cannot be added to the simulation of the crossover. That means the simulation cannot take into consideration the actual driver response. Unthinkable around here. Additionally- the testing done here establishes an acoustic center for the driver position from which all phase measurements are made. Your statement "... it is a continuum of shift..." is dead wrong. Phase of the driver can invert within 1Hz bandwidth and is therefore not continuous. This is caused by change in mode of the cone which is also well established. Any even bigger error of assumption is the phase of the output at a given frequency and radiation direction is smooth and changes in a continuous fashion. Super wrong. The main lobe of a single driver is in one phase and the adjacent lobe is typically 180° opposite. The intersection of the two lobes do not show any smooth change. Between the lobes there is a null and on either side of the null are two opposite phase signals. Like pregnancy, it is one or the other and not almost pregnant, not continuous. Linkwitz writes of this at length.
PANO- Just saw your post. I said HOLM does not work correctly providing erroneous data. I did not say it is inconsistent in its measurements. If the error made is consistent and the speakers are of the same design am certain what you say could be true. That does not mean the actual data is correct. It means HOLM can tell when a driver is mis-wired when comparing two pretty much otherwise identical speakers. Definitely handy!
Last edited:
OK, I am getting tired of this tedious discussion.
First it is spelled Heyser, and second, I hope that you are aware of the fact that the AES would not publish Heyser's work after he died because they found it to be erroneous. His papers were investigated by a couple of Purdue Math Profs and found not to be correct.
I knew Dick Heyser and met him several times. He was a nice enough guy, but his ideas of measurements were wrong. Sorry to burst your bubble!
I have been modeling loudspeakers and crossovers probably since before you were born (judging from your naïve views) and phase has always been a key consideration and guess what!? My results always work out just fine. Your two-tone method is so full of holes that to use it as an absolute reference as to what is right is absurd. What you are saying is that your clumsy method and opinions are right and everyone else is wrong. Now that is a position that you will have to back up with more than just what you have said thus far.
First it is spelled Heyser, and second, I hope that you are aware of the fact that the AES would not publish Heyser's work after he died because they found it to be erroneous. His papers were investigated by a couple of Purdue Math Profs and found not to be correct.
I knew Dick Heyser and met him several times. He was a nice enough guy, but his ideas of measurements were wrong. Sorry to burst your bubble!
I have been modeling loudspeakers and crossovers probably since before you were born (judging from your naïve views) and phase has always been a key consideration and guess what!? My results always work out just fine. Your two-tone method is so full of holes that to use it as an absolute reference as to what is right is absurd. What you are saying is that your clumsy method and opinions are right and everyone else is wrong. Now that is a position that you will have to back up with more than just what you have said thus far.
Any cites on that? I've read a lot of his stuff and found it to be a mix of nonsense and very profound- the task was always to separate the two. Always interesting, though. He reminded me a lot of Tommy Gold.
FWIW, I've been a strong advocate of using Heyser impedance plots, which I find useful for spotting resonances and determining amplifier drive requirements. His well-known speculation on Thom theory was a perfect example of the nonsense/profound mix.
Hi Sy
True, not everything was wrong, and some of it was insightful, but too much of it was wrong for the AES to publish it.
When Heyser died there was a push to publish all his works as a tribute. I was on the board at the time. The outside people were hired for their opinion and it was as I said. No one wanted to make a big deal of this - obviously - but the fact is that nothing from Heyser was ever published posthumously.
True, not everything was wrong, and some of it was insightful, but too much of it was wrong for the AES to publish it.
When Heyser died there was a push to publish all his works as a tribute. I was on the board at the time. The outside people were hired for their opinion and it was as I said. No one wanted to make a big deal of this - obviously - but the fact is that nothing from Heyser was ever published posthumously.
I might be not understanding things properly here, but it seems to me that what this two tone test is showing is phase shift with frequency for a particular driver? ie when that driver is generating two tones far apart (but one a multiple of the other) in frequency that the higher frequency may be shifted in phase somewhat so is no longer completely in phase with the lower frequency?
As a test by itself as to how coherent (phase wise) the various frequencies a driver emits are, I can see it being a good test to separate the good drivers from poor.
Am I right that an exceptional driver would show an almost flat line (20 deg change from start to finish) for phase from left to right with increasing frequency? Is not a typical down slope of phase from a measurement such as holm produces, basically showing the same phenomenon?
Is this another of those holy grail type things, linear phase perhaps?
I've probably completely missed the point
Tony.
As a test by itself as to how coherent (phase wise) the various frequencies a driver emits are, I can see it being a good test to separate the good drivers from poor.
Am I right that an exceptional driver would show an almost flat line (20 deg change from start to finish) for phase from left to right with increasing frequency? Is not a typical down slope of phase from a measurement such as holm produces, basically showing the same phenomenon?
Is this another of those holy grail type things, linear phase perhaps?
I've probably completely missed the point
Tony.
OK, but if that is the case, then why does HOLM always show me the correct polarity of a speaker? I've never had it fail, no matter the driver. Yes, I've verified the whole system, I use sawtooth waves for the electronics and a battery for the driver. HOLM always shows the polarity it should.
Probably. Or maybe just close up. I have a hypo-echoic chamber down to ~20Hz. Should be OK. Was too busy blowing up tweeters tonight to measure phase.