Re: Re: Still semantics
I, too, was referring to a minimum-phase system, that of a raw driver, the orginal subject of my first post, since that was your topic at the time and the intended subject of my subsequent posts. In addition, a piston driver is minimum-phase not only in its pass-band but also its breakup. Using CALSOD, for easy confirmation, one creates a model using a base passband and a set of minimum-phase elements to model any non-linearities in FR, as could be done for your example. The detail as to how many of the points of deviation one wishes to model. I've modeled 10" drivers to beyond 20KHz to match the SPL response to every single deviation in the driver. The Hilbert-Bode generated phase matches the measured phase to the smallest detail if one does it accurately.
If one is to say that the breakup is non-linear due to the fact that it creates non-linear distortion in addition to the linear response, then one must then say that all drivers are non-linear in their entirety. All drivers exhibit deviations from flat FR (exclude the typical passband extremes and any motor-generated distortion for now). Any deviation from a flat FR generates harmonic distortion in the presence of multiple tones, regardless of whether it is below our measurement threshold or display resolution. As Linkwitz has pointed out on Page 1 of his "Issues of Loudspeaker Design":
Since all drivers generate some level of harmonic distortion even in their passband due to non-flat response, they therefore generate IMD. This is before we add the motor distortion, also prevalent in the passband. So in essence one would have to say that all drivers are non-linear throughout their full range of operation in that case, not just the breakup region.
Dave
marche said:
I, too, was referring to a minimum-phase system, that of a raw driver, the orginal subject of my first post, since that was your topic at the time and the intended subject of my subsequent posts. In addition, a piston driver is minimum-phase not only in its pass-band but also its breakup. Using CALSOD, for easy confirmation, one creates a model using a base passband and a set of minimum-phase elements to model any non-linearities in FR, as could be done for your example. The detail as to how many of the points of deviation one wishes to model. I've modeled 10" drivers to beyond 20KHz to match the SPL response to every single deviation in the driver. The Hilbert-Bode generated phase matches the measured phase to the smallest detail if one does it accurately.
If one is to say that the breakup is non-linear due to the fact that it creates non-linear distortion in addition to the linear response, then one must then say that all drivers are non-linear in their entirety. All drivers exhibit deviations from flat FR (exclude the typical passband extremes and any motor-generated distortion for now). Any deviation from a flat FR generates harmonic distortion in the presence of multiple tones, regardless of whether it is below our measurement threshold or display resolution. As Linkwitz has pointed out on Page 1 of his "Issues of Loudspeaker Design":
Since all drivers generate some level of harmonic distortion even in their passband due to non-flat response, they therefore generate IMD. This is before we add the motor distortion, also prevalent in the passband. So in essence one would have to say that all drivers are non-linear throughout their full range of operation in that case, not just the breakup region.
Dave
Re: Nonsense
I'm sorry Dave, but this is not a valid argument. I can't imagine that you've double checked all of your measurements over the last 10 years to prove that they are minimum phase. And even if you would... the fact that an observation is different from yours, doesn't imply that therefore somebody elses information is wrong by definition. In that sense your post is not blunt, it is just wrongly argumented, and discusses things like "... I have more experience than you ...". I'm not interested to be honest.
If you want to find flaws in the information presented, ask for the measurements (I will share them with any of you), and show errors based on technical arguments instead of just hand-waving. I don't mind you being suspicious, I do object conclusions being drawn based on disbelief.
For common practices, a minumum phase model of a speaker unit is perfectly valid for speaker design, because in its typical operation region a loudspeaker IS a minimum phase system.
On one hand your reasoning bluntly assumes breakup is some sort of linear property ("shows up in the same manner everywhere"), but on the other hand you combine this reasoning with non-linearity. Of course you will get a falsum with such a kind of reasoning...
Independent of lineairity or non-linearity, discussing a frequency region of a speaker that one normally avoids, is not what I consider spending my time useful...
Marc
dlr said:
I'm sorry Dave, but this is not a valid argument. I can't imagine that you've double checked all of your measurements over the last 10 years to prove that they are minimum phase. And even if you would... the fact that an observation is different from yours, doesn't imply that therefore somebody elses information is wrong by definition. In that sense your post is not blunt, it is just wrongly argumented, and discusses things like "... I have more experience than you ...". I'm not interested to be honest.
If you want to find flaws in the information presented, ask for the measurements (I will share them with any of you), and show errors based on technical arguments instead of just hand-waving. I don't mind you being suspicious, I do object conclusions being drawn based on disbelief.
For common practices, a minumum phase model of a speaker unit is perfectly valid for speaker design, because in its typical operation region a loudspeaker IS a minimum phase system.
dlr said:
On one hand your reasoning bluntly assumes breakup is some sort of linear property ("shows up in the same manner everywhere"), but on the other hand you combine this reasoning with non-linearity. Of course you will get a falsum with such a kind of reasoning...
Independent of lineairity or non-linearity, discussing a frequency region of a speaker that one normally avoids, is not what I consider spending my time useful...
Marc
marche said:
I believe this statement is not correct as written.
The defintion of non-linear being applied in this thread is not very clear (which I think goes some way to explain the constant confusion). The usual definition would be equivalent to saying that the sound pressure at a given measurement point and at a given frequency is not proportional to the input voltage at that frequency. (True for harmonic or IM distortion result from non-linearities, whereas frequency response errors present at all levels do not.*)
That is a separate issue from minimum phase which can be a feature of an entirely linear system (in the sense defined above). A way to have non-minimum phase is to have part of the sound reaching the measurement point after a time delay compared to the primary path from the main radiator. That needs energy storage (masses and springs forming resonators) or reflection of travelling waves, both of which can produce non minimum phase behaviour even in a perfectly linear model.
*Some people, for example, seem to say that because the frequency response is not a "straight line" there is non-linearity, but that is not the normal usage in a physics/engineering sense, and only confusing when the discussion is about HD or IM.
Sorry if I was taking your statement to mean something other than you seemed to write.
Ken
Re: Re: Nonsense
I understand where you're coming from on this. It's been a common misconception mostly due to errors in the models created.
I didn't bluntly assume, I came to this conclusion after years worth of measuring and modeling and always getting a match. Empirical data that showed minimum-phase behavior in the breakup region in every case could not occur if it was not minimum-phase. The software requires minimum-phase behavior to generate the phase from the HBT. I have no need to go back and check because I matched the SPL and thus the generated phase each time as I went along. There is no doubt that either your measurement is flawed or your model is incomplete, more likely the latter. What you've shown is a small range of the model, insufficient to draw a conclusion as shown. The fact that you didn't match it does not invalidate it.
It's not hand-waving, I have years worth of measurements that have all shown minimum-phase response when modeled accurately. That latter comment is critical. Models that don't show minimum-phase behavior generally have errors in the lowpass that introduce excess-phase (whether positive or negative) in the generated response or the Fc of the lowpass was badly chosen in addition. I'll see if I can get time this weekend to post some of those measurements and generated phase.
These are two sites of one person who has confirmed this himself and is in full agreement on this aspect as his experience over time was the same as mine.
John's original web site
John's commercial site
It's one of the simplest things one can do in design, yet one of the most common sources of error.
Dave
marche said:
I understand where you're coming from on this. It's been a common misconception mostly due to errors in the models created.
I didn't bluntly assume, I came to this conclusion after years worth of measuring and modeling and always getting a match. Empirical data that showed minimum-phase behavior in the breakup region in every case could not occur if it was not minimum-phase. The software requires minimum-phase behavior to generate the phase from the HBT. I have no need to go back and check because I matched the SPL and thus the generated phase each time as I went along. There is no doubt that either your measurement is flawed or your model is incomplete, more likely the latter. What you've shown is a small range of the model, insufficient to draw a conclusion as shown. The fact that you didn't match it does not invalidate it.
It's not hand-waving, I have years worth of measurements that have all shown minimum-phase response when modeled accurately. That latter comment is critical. Models that don't show minimum-phase behavior generally have errors in the lowpass that introduce excess-phase (whether positive or negative) in the generated response or the Fc of the lowpass was badly chosen in addition. I'll see if I can get time this weekend to post some of those measurements and generated phase.
These are two sites of one person who has confirmed this himself and is in full agreement on this aspect as his experience over time was the same as mine.
John's original web site
John's commercial site
It's one of the simplest things one can do in design, yet one of the most common sources of error.
Dave
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.