This is a really long thread and have wondered about this for a long time. Sorry have not read this very long thread for a while but, would really like this clarified.
It is my understanding the stimulating impulse (test signal) has a maximum amplitude at some point along the impulse curve, the peak amplitude at time say t-zero for convenience. The response of the speaker under test also produces a maximum amplitude and then is received at the microphone returning to the analyzer at time t-zero+nTime. It is the time from the driving impulse maximum amplitude to the speaker under test maximum amplitude which is used to determine the "excess time" or delay from the speaker under test to the microphone. This is used to determine the distance from microphone to the speaker is what I understand. Small times like delay through amplifiers and so on also add to this time but are small compared the the delay from speaker to microphone.
Is this correct? If wrong, how is this time determined to allow calculation of driver phase response and so on?
It is my understanding the stimulating impulse (test signal) has a maximum amplitude at some point along the impulse curve, the peak amplitude at time say t-zero for convenience. The response of the speaker under test also produces a maximum amplitude and then is received at the microphone returning to the analyzer at time t-zero+nTime. It is the time from the driving impulse maximum amplitude to the speaker under test maximum amplitude which is used to determine the "excess time" or delay from the speaker under test to the microphone. This is used to determine the distance from microphone to the speaker is what I understand. Small times like delay through amplifiers and so on also add to this time but are small compared the the delay from speaker to microphone.
Is this correct? If wrong, how is this time determined to allow calculation of driver phase response and so on?
If you could find the leading edge of the impulse with any amount of certainty, that would be a more accurate means to determine "excess". This is because the timing of the peak of the impulse is dependent on the driver's bandwidth; more specifically: high frequency response. Anyway, try to home in on the point where the measured impulse begins to rise (some estimation/judgment required) and you're good to go: that should effectively eliminate excess time.
While trying to find online references I found this blog, which may help some (I have not yet read all of it myself, but it looks relevant): Phase Alignment of Subs – Why I don’t use the impulse response
I personally don't use the phase generated by HOLMImpulse. I import the SPL response into SoundEasy, do a Hilbert-Bode transform to determine minimum phase. I then use this method: Finding Relative Acoustic Offsets Empirically to determine relative acoustic offset.
Here's some more info (you may have to log in at PartsExpress Tech Talk forum to get the file): How To Use Passive Crossover Designer To Find The Relative Acoustic Offset.
While trying to find online references I found this blog, which may help some (I have not yet read all of it myself, but it looks relevant): Phase Alignment of Subs – Why I don’t use the impulse response
I personally don't use the phase generated by HOLMImpulse. I import the SPL response into SoundEasy, do a Hilbert-Bode transform to determine minimum phase. I then use this method: Finding Relative Acoustic Offsets Empirically to determine relative acoustic offset.
Here's some more info (you may have to log in at PartsExpress Tech Talk forum to get the file): How To Use Passive Crossover Designer To Find The Relative Acoustic Offset.
Thanks for posting this info and links. I am glad I did not have to point out,
LOUDNESS=DISTANCE
as a ridiculous conclusion. The problem then turns into, without knowing the distance, the impulse mapping into response cannot be made therefore, this whole impulse testing method without accurate distance (excess time through everything) is more or less a complete waste of time whereas the results are meaningless.
LOUDNESS=DISTANCE
as a ridiculous conclusion. The problem then turns into, without knowing the distance, the impulse mapping into response cannot be made therefore, this whole impulse testing method without accurate distance (excess time through everything) is more or less a complete waste of time whereas the results are meaningless.
It is not clear why you think it might be a waste of time. Perhaps if you share what it is you are trying to achieve you will get an answer that addresses your problem more directly.
On your earlier request for HOLMImpulse software writers to respond... askbojeson did a lot of development here in the early days, but has not posted here since the software reached maturity. I am sure that here are others who understand the workings of the software well enough to answer your original question.
On your earlier request for HOLMImpulse software writers to respond... askbojeson did a lot of development here in the early days, but has not posted here since the software reached maturity. I am sure that here are others who understand the workings of the software well enough to answer your original question.
I a hobbyist with experience aligning drivers in a multiway system. I agree with Shaun that if we knew the context for the question then the answer will become easier to convey.
Your concept of relating loudness to time/distance is erroneous. The level of the sweep signal is constant across the frequency range. The IR that is calculated by the analysis of the mic measurement results in an initial large peak in the IR that closely relates to the time the highest frequencies captured arrive at the mic. This time is (mostly) independent of the level of the signal and the mic input level. As Shaun indicated, the initial rise is really the better indicator of the delay of the first arriving (highest) freqs.
The phase response and group delay of a driver is changing over the its frequency range so there is no single "acoustic center" or "delay" for all frequencies.
These are general comments and may not help answer your question, but if not, they may help you form a more detailed one. Are you trying to align drivers in a multiway speaker design? Do you have a way to introduce a set delay in your system with a speaker management box like a DCX2496? Is it just an academic question?
Holm is not very well documented, but REW, Arta, and many other speaker measurement packages are. You may get your answer by reading their docs.
Manipulation Feature in Holmimpulse
Hi all,
I am working on vinyl EQ measurements and my idea was to use the manipulation feature of HOLMimpulse to calculate the difference of a DUTs Trasfer Function from the ideal RIAA (or Columbia, Decca,...).
First I generated the RIAA transfer function (txt file, frequency, magnitude phase) in Scilab. It imports fine in HOLMimpulse using the import function. Second I measured the actual transfer function of the phono-pre. Unfortunately the manipulation (e.g. C = A/B) does not work properly. As a result I can only see a 0 degree phase line but no magnitude. It works well if I use two imported FR (e.g. Decca + RIAA) so I assume the problem is related to different sampling/frequency resolution, however I used 96kHz samplerate and 32768 FFT size for both, measurement and TF calculation in Scilab.
Any idea what the problem might be?
kind regards, Daniel
Hi all,
I am working on vinyl EQ measurements and my idea was to use the manipulation feature of HOLMimpulse to calculate the difference of a DUTs Trasfer Function from the ideal RIAA (or Columbia, Decca,...).
First I generated the RIAA transfer function (txt file, frequency, magnitude phase) in Scilab. It imports fine in HOLMimpulse using the import function. Second I measured the actual transfer function of the phono-pre. Unfortunately the manipulation (e.g. C = A/B) does not work properly. As a result I can only see a 0 degree phase line but no magnitude. It works well if I use two imported FR (e.g. Decca + RIAA) so I assume the problem is related to different sampling/frequency resolution, however I used 96kHz samplerate and 32768 FFT size for both, measurement and TF calculation in Scilab.
Any idea what the problem might be?
kind regards, Daniel
Hehe, yes. Did that 
Just tested again with loop trough measurement of soundcard (which is close to 0dB for all frequencies) and RIAA (which is 0dB at 0Hz, about -20dB at 1kHz,...) and all manipulation (A/B, A*B, A-B,...) yields the same result: no magnitude, phase 0 degree line and also no impulse response (auto zoom).
It works fine with two measurements and also with two imported transfer functions (or imported impulse responses) but not with measurement + import
Any help is appreciated since I think it's a nice application
kind regards, Daniel
Just tested again with loop trough measurement of soundcard (which is close to 0dB for all frequencies) and RIAA (which is 0dB at 0Hz, about -20dB at 1kHz,...) and all manipulation (A/B, A*B, A-B,...) yields the same result: no magnitude, phase 0 degree line and also no impulse response (auto zoom).
It works fine with two measurements and also with two imported transfer functions (or imported impulse responses) but not with measurement + import
Any help is appreciated since I think it's a nice application
kind regards, Daniel
Someone here may have a good suggestion. I can only mention a couple of basic ideas:
> Do you have the same issue in REW V5.01 Beta 20 or Arta?
> Have you exported the measurement and then imported it back in? [maybe the format will then be compatible as they are then both "imported"?]
If you want someone to take a look, I would do that. I am not the most skilled user here, but occasionally have found workarounds for problems like this. You would need to post the .zip (holm file) and also the import file.
> Do you have the same issue in REW V5.01 Beta 20 or Arta?
> Have you exported the measurement and then imported it back in? [maybe the format will then be compatible as they are then both "imported"?]
If you want someone to take a look, I would do that. I am not the most skilled user here, but occasionally have found workarounds for problems like this. You would need to post the .zip (holm file) and also the import file.
You need to make sure that the measurement and the import both have the same sample rate. If one was 48Khz and the other 96Khz you won't be able to combine them (from memory). That is something I would check as I think I experienced the same issue and this is what it turned out to be.
edit: I should read more carefully as you said you did do them at the same sampling rate. I have a suspicion that holm changes the sampling rate when it exports. I'd suggest exporting both the measured one and the imported one and then re-import both and see if it works.
Tony.
edit: I should read more carefully as you said you did do them at the same sampling rate. I have a suspicion that holm changes the sampling rate when it exports. I'd suggest exporting both the measured one and the imported one and then re-import both and see if it works.
Tony.
It's possible to record the Holm test signal and play it back with a music server if that is what you have in mind. This allows us to see the impact of a computer based PEQ or in my case the impact of a phase correction IR (RePhase IR) implemented using Foo_convolve.
I modified my sweep file to add 3 tics spaced at 1 sec intervals so I know when to initiate the Holm measurement.
I modified my sweep file to add 3 tics spaced at 1 sec intervals so I know when to initiate the Holm measurement.
That is a whole mouth full of assumptions. "The phase response and group delay of a driver is changing over the its frequency range so there is no single "acoustic center" or "delay" for all frequencies. " is a really curious notion. Group delay occurs at the end of a low pass filters bandwidth and is never "changing." The problem is these assumptions the high frequencies blah blah blah is completely erroneous to what a driver does. Holm Impulse makes so many assumptions about the behavior of drivers my goal is to point out the assumptions outweigh the data in most cases. To measure a driver the very first thing which is required it to determine the drivers center of band frequency along with bandwidth. There are many ways to do this but none include impulse testing because without the knowledge of the bandwidth and band center the impulse data cannot be decoded into useful information. After the bandwidth and band center are determined accurate measurement of total excess time can be measured using frequencies near the band center. Again there are many methods to do this with all requiring the bandwidth and band center knowledge. Since Holm makes so many assumptions the information concerning phase response is useless and so the bandwidth and band center are almost never determined correctly which leads to the rest of the results being junk information also.
Really good drivers have wide bandwidths and maintain very small variations in phase over a that wide bandwidth. Indeed, I have an off the shelf driver right here which maintains about plus and minus 10 degrees from 200Hz to 5000Hz and nearly perfect frequency response over that range. On axis and 30 degrees off axis phase and frequency response are almost identical over that frequency range.
I really do not feel like getting into some far flung argument about this. All of it is well documented in AES journals prior to 1985. My key point is to show this nice piece of software does nothing really useful. It could be fixed but that would require several more pieces of software to first measure bandwidth and band center and then measure delay. With that knowledge, impulse testing could offer some really useful date. Without these steps this is a complete waste of time.
Perhaps you should start by telling us what those assumptions are, instead of arguing with those who are trying to help you.
@sumaudioguy. Maybe useless to you (or for your specific needs), but it has allowed many diyers to make crossovers far superior than would be possible without the use of it as a tool.
I'm not sure whether you have actually read the holmimpulse documentation. I might be wrong but from memory it does not actually use an impulse response to measure the driver, the impulse response is mathematically calculated from the measurement.
I for one use the swept sine mode when I do my measurements... I'm not sure how that is an impulse response.. It seems that maybe you are making as many assumptions as you presume others are
Tony.
I'm not sure whether you have actually read the holmimpulse documentation. I might be wrong but from memory it does not actually use an impulse response to measure the driver, the impulse response is mathematically calculated from the measurement.
I for one use the swept sine mode when I do my measurements... I'm not sure how that is an impulse response.. It seems that maybe you are making as many assumptions as you presume others are
Tony.
Assuming the subject is "driver timing alignment" in a multiway system, Holm is very well suited to the task. The user control features are well designed to easily find the correct delay needed to optimize the phase tracking through the XO range. This provides a stable central lobe pointed at the LP and helps to stabilize the direct sound field around the LP.
You are correct that raw drivers have a bandpass range that is basically flat and constant delay. When IIR XO filters are added for XO, additional phase rotation is also added and the result is that it is impossible to XO between drivers without significant phase rotation and thus changing GD through the XO range.