Tony
In HOLM we do not have access to the impulse response of the input wave. One would have to post process the data to get this information - doable, if it actually added anything to the analysis. With two tests at different DUT to mic distances, one could calculate the sound card latency if this were of interest.
PS. of course the latency could also be easily done other simpler ways!!
In HOLM we do not have access to the impulse response of the input wave. One would have to post process the data to get this information - doable, if it actually added anything to the analysis. With two tests at different DUT to mic distances, one could calculate the sound card latency if this were of interest.
PS. of course the latency could also be easily done other simpler ways!!
Doesn't HOLMImpulse work under Windows 10 (32-bit)?
Just installed it with no apparent problems but when I try to open it up, it closes down after displaying the splash screen for a couple of seconds, so never goes to the main window. Tried it with compatibility settings for Win 7 and 8 too but to no avail.
Am I missing some system requirement, e.g. Java, .NET framework, Flash etc?
Just installed it with no apparent problems but when I try to open it up, it closes down after displaying the splash screen for a couple of seconds, so never goes to the main window. Tried it with compatibility settings for Win 7 and 8 too but to no avail.
Am I missing some system requirement, e.g. Java, .NET framework, Flash etc?
Last edited:
First this quote.- If the speaker is minimum phase over a bandwidth why does HOLM not show this? I see no minimum phase graphs produced by HOLM for any drivers. So what is wrong here? HOLM does not do this measurement correctly, that is what is wrong. Note this post by me agreeing with a band of minimum phase- http://www.diyaudio.com/forums/soft...requency-impulse-response-86.html#post3935714
JohnPM- the underlying assumption was in my suggested test was to first determine the band center and then use test signals near that band center to determine the distance. Both of these I have explained in detail in earllier post. The answer from the HOLM guy was almost an exact quote. This is wrong and everyone should realize it is wrong as I do. Further, if system delay is not constant then no measurements of phase can be made unless that too is included in the calculations. System delay is not important for modeling a loudspeaker, driver phase response is and can be very accurately measured using two tone test. It seems you are trying to use the argument for delay as validation for phase. What is that? Validation by diversion? Are you instead referring to delay differences between different drivers in the same system? Well for that two tone will not provide the answer and will not provide total delay. Everyone knows that and I made no such claim for two tone to provide delay, only phase. I am discussing the primary error of phase and not multiple driver delay which HOLM also does not measure correctly. Once the band center is known, the assumption is signals near the band center will exhibit minimum phase behavior so distance can be measured in any of several valid ways, none used by HOLM. That is the assumption with everything else in my discription being then measured by a valid process.
I made the very clear explanation in previous post so would suggest anyone interested should bother to read those to remark intelligently. arnyk did not do this clearly making a point already presented. Really, I read this whole long thread prior to contacting the HOLM supplier with these very pointed questions. I was hoping for something useful but found GIGO instead. How about trying that and focus on what I have written as a correct method instead of making the assumption HOLM is an infallible authority and anything which disagrees is wrong. Or is this tread a political process as so many things masquerading as science today and on DIYaudio are?
Hi sumaudioguy, I think the problem is that you state that the program is useless, which many have found to be false. Maybe for your particular use scenario it is, but for crossover design as we are doing it, it has proved a very useful tool.
If someone comes and tells you that what you are doing won't work, and all you need to do is do some tests to prove it. But you have already done tests that prove it DOES work, do you think that you are going to say, "Oh that must have been a fluke that my stuff worked I better run these tests to verify that it shouldn't have"?
Tony.
If someone comes and tells you that what you are doing won't work, and all you need to do is do some tests to prove it. But you have already done tests that prove it DOES work, do you think that you are going to say, "Oh that must have been a fluke that my stuff worked I better run these tests to verify that it shouldn't have"?
Tony.
In post #842 you said (with my added emphasis):
How you made the leap from the time between impulse response peaks corresponding to distance to "loudness = distance" is anyone's guess, but it doesn't seem connected in any way to logic or common sense.
It is in any event irrelevant to what you say you want, which is a valid phase response. If I understand your comments on drive unit band centre you are making the assumption that phase at or near the band centre can be treated as being more or less zero. Holmimpulse lets you easily achieve this. Simply use the control provided for adjusting the impulse response time zero reference until the phase at your band centre frequency is zero. Holmimpulse can get you close by automatically picking the impulse response peak as an initial setting for the zero time (just use the Auto Detect button with the option set to Largest Peak), so you shouldn't have to move the reference much (if at all).
How you made the leap from the time between impulse response peaks corresponding to distance to "loudness = distance" is anyone's guess, but it doesn't seem connected in any way to logic or common sense.
It is in any event irrelevant to what you say you want, which is a valid phase response. If I understand your comments on drive unit band centre you are making the assumption that phase at or near the band centre can be treated as being more or less zero. Holmimpulse lets you easily achieve this. Simply use the control provided for adjusting the impulse response time zero reference until the phase at your band centre frequency is zero. Holmimpulse can get you close by automatically picking the impulse response peak as an initial setting for the zero time (just use the Auto Detect button with the option set to Largest Peak), so you shouldn't have to move the reference much (if at all).
Tony - I have to take issue with you comment. If a certain program produces erroneous results then it is "useless" - to all of us, not just SumAudioguy. He is saying that the program is more than just useless (it could be perfectly correct and still be useless to him,) he is saying that Holm is wrong and that's a larger claim. He has yet to prove his contention however. (And I am not holding my breath that he ever will since he does not seem to hear anything else that anyone says.)
We all keep saying over and over that relative phase in Holm is perfectly correct. Then what about "absolute" phase? There is no such thing! Phase is always relative to some point in time arbitrarily called t=0 and this is always relative (unless we go back to the Big Bang.) So there is only relative phase and that is calculated quite correctly in Holm.
We all keep saying over and over that relative phase in Holm is perfectly correct. Then what about "absolute" phase? There is no such thing! Phase is always relative to some point in time arbitrarily called t=0 and this is always relative (unless we go back to the Big Bang.) So there is only relative phase and that is calculated quite correctly in Holm.
Hi Earl, I agree, but I think we (us and sumaudioguy) are talking apples and oranges... I may be being extreme in my example, but if holm were truly broken and useless then I am at a loss as to why it works for me 
Unless the thing that Sumaudioguy says is broken/useless is irrelevant to how I use it (and that was basically my point).
Tony.
Unless the thing that Sumaudioguy says is broken/useless is irrelevant to how I use it (and that was basically my point). Tony.
I tested some months ago, with early versions of Win10, no problem on my computer(s), just install and run!
Sumaudioguy,
"The highest amplitude of the test signal is time aligned with the highest amplitude of the response signal with that difference in time being the distance to the transducer."
This above calculus is done only for achieving the distance between mic and speaker which is necessary for correct phase calculus later on.
Simplified, Holm presumably works like this:
Holm does a frequency sweep, and records both amplitude and phase,
using "When-Signal-Was-Sent-From-Holm" (WSWSFH) as time zero.
Holm then takes those three variables and calculates the impulse response.
In unlocked mode, Holm then looks at this impulse response, and detects a time delay compared to the WSWSFH time point.
Holm moves those two time points to "0" at the impulse time scale.
This new "0" time point reference is then used for calculating the actual phase response over the speaker frequency band.
Holm then presents impulse, frequency and phase curves.
The moved WSWSFH time reference point can be locked, enabling the DIY:er to see the phase differences when speaker is moved or modified.
Hope this helps....
"The highest amplitude of the test signal is time aligned with the highest amplitude of the response signal with that difference in time being the distance to the transducer."
This above calculus is done only for achieving the distance between mic and speaker which is necessary for correct phase calculus later on.
Simplified, Holm presumably works like this:
Holm does a frequency sweep, and records both amplitude and phase,
using "When-Signal-Was-Sent-From-Holm" (WSWSFH) as time zero.
Holm then takes those three variables and calculates the impulse response.
In unlocked mode, Holm then looks at this impulse response, and detects a time delay compared to the WSWSFH time point.
Holm moves those two time points to "0" at the impulse time scale.
This new "0" time point reference is then used for calculating the actual phase response over the speaker frequency band.
Holm then presents impulse, frequency and phase curves.
The moved WSWSFH time reference point can be locked, enabling the DIY:er to see the phase differences when speaker is moved or modified.
Hope this helps....
Last edited:
I am interested in the phase plot, so I did some experimenting. I made a band passed signal and looked at it in several software packages to see how they displayed phase of the same signal
The signal was bandpassed at 60Hz, 2nd order Butterworth and 10K, 1st order. The different software was:
Phase at 20Hz = +150 deg. Phase at 20KHz = -60 deg, Phase at 1Khz = 0
But with one caveat.
If HOLM or REW put the impulse peak at time 0, then the phase at the top of the frequency range tends to flatten. That can be shifted in HOLM to agree with other software.
This weekend I will check with driver measurements, movements and filtering.
The signal was bandpassed at 60Hz, 2nd order Butterworth and 10K, 1st order. The different software was:
- Passive Crossover Designer
- XSim
- REW
- ARTA
- rePhase
- HOLMImpulse
Phase at 20Hz = +150 deg. Phase at 20KHz = -60 deg, Phase at 1Khz = 0
But with one caveat.
If HOLM or REW put the impulse peak at time 0, then the phase at the top of the frequency range tends to flatten. That can be shifted in HOLM to agree with other software.
This weekend I will check with driver measurements, movements and filtering.
Pano - I would have expected nothing any different. The claims made are simply wrong. All decent software is going to give the same answer.
This question was raised at the AES decades ago and it was found (in numerous workshops) that - when done right - all approaches and software agreed. And if they hadn't? Then one was wrong and it would have been corrected. How could it be any other way? Taking acoustic measurements is not a new subject. There is nothing that is going to slip through the cracks and go undetected.
There was an excellent paper on all this more than ten years ago comparing all of the different techniques. It was concluded in that paper that Log-Sweep was the most robust in terms of noise and nonlinearities - the two sources of most significant errors. This is precisely what Holm uses and precisely why I started using it. I have found it to be very immune to nonlinearities, but not so much to noise, especially impulsive noise. If someone slams a door in my house while a sweep is going on it will be seen in the result as a sharp glitch. I wish Holm still did averaging (it used to,) but I think that this may have had some problems.
This question was raised at the AES decades ago and it was found (in numerous workshops) that - when done right - all approaches and software agreed. And if they hadn't? Then one was wrong and it would have been corrected. How could it be any other way? Taking acoustic measurements is not a new subject. There is nothing that is going to slip through the cracks and go undetected.
There was an excellent paper on all this more than ten years ago comparing all of the different techniques. It was concluded in that paper that Log-Sweep was the most robust in terms of noise and nonlinearities - the two sources of most significant errors. This is precisely what Holm uses and precisely why I started using it. I have found it to be very immune to nonlinearities, but not so much to noise, especially impulsive noise. If someone slams a door in my house while a sweep is going on it will be seen in the result as a sharp glitch. I wish Holm still did averaging (it used to,) but I think that this may have had some problems.
I had hoped that sumaudioguy would reply to my question here:
but, since he has not been forthcoming, I'm going to share what I had planned to next, anyway. These are some articles and links to articles discussing phase measurement. Or rather not: most describe ways to _generate_ the minimum phase response post-measurement, and how to do so reliably.
Driver Model Accuracy and its Impact on Phase by David L. Ralph
Finding Relative Acoustic Offsets Empirically by David L. Ralph
Approximate determination of minimum-phase response of measured loudspeaker by Bohdan Raszinsky
Acoustic Centre Evaluation by Bohdan Raszinsky
It's Just A Phase I Am Going Through by John Kreskovsky
but, since he has not been forthcoming, I'm going to share what I had planned to next, anyway. These are some articles and links to articles discussing phase measurement. Or rather not: most describe ways to _generate_ the minimum phase response post-measurement, and how to do so reliably.
Driver Model Accuracy and its Impact on Phase by David L. Ralph
Finding Relative Acoustic Offsets Empirically by David L. Ralph
Approximate determination of minimum-phase response of measured loudspeaker by Bohdan Raszinsky
Acoustic Centre Evaluation by Bohdan Raszinsky
It's Just A Phase I Am Going Through by John Kreskovsky
http://www.bg.ic.ac.uk/research/g.stan/JAES_Online_version.pdf ?
The log-sweep method seems to have been popularised for audio measurements by Angelo Farina (AES paper), although the technique was previously known and used in other fields.
Yes indeed. But sometimes I just want to check for myself.
Also was interested in the time lock. What do you lock to? When I was doing 3-way crossovers, I normally locked to the midrange driver. But is that correct? Since an impulse peak is mostly composed of the higher frequencies, what effect does locking to the top end of the midrange have?
I looked at this in HOLM. If the midrange driver impulse peak is put at time zero, and then the following driver measurements are locked to that offset, what happens?
Answer: Nothing much. They all line up as they should. The driver phases and amplitudes sum as they should. I still don't see the problem. Anywhere else to look?
With direct radiating drivers I use the midrange driver as a time reference, note the physical offsets of the impulse response peaks for the other drivers and then center all the impulse response peaks on time zero to produce the phase and amplitude response files. I use the offsets as the acoustic center offsets in the modeling software. The on-axis results are as expected and the modeled off-axis response matches measurements fairly well. I think this is more realistic than assuming that the driver acoustic centers are all in the same plane.
I'm no expert, but I think horns are different. The Z-axis offset of the apparent focal point of the horn is not necessarily consistent with the time delay. I have dealt with a couple of horns by guessing where the focal point is and using a combination of Z-axis offset and time delay in modeling. Seems to work OK.
I'm no expert, but I think horns are different. The Z-axis offset of the apparent focal point of the horn is not necessarily consistent with the time delay. I have dealt with a couple of horns by guessing where the focal point is and using a combination of Z-axis offset and time delay in modeling. Seems to work OK.