Speaking of Klippel.
We can make a poormans Klippels setup.
A little bit of back to the basics again to explain it;
https://en.wikipedia.org/wiki/Thiele/Small_parameters
So we have basically 5 variables, Cms (1/Kms), Mms, Rms, Re and BL.
Measuring them all isn't possible unfortunately, and therefor Klippel assumes a few things;
Since Mms is constant, we know that in our impedance measurements the frequency shift is only being created by the difference (non-linearity) of the Cms (1/Kms). Because we know that:
- Cms = 1 / ((2*pi*fs)²*Mms)
Since Mms is assumed to be constant (by Klippel), the only variable we have to measure is the fs to find the difference in Cms.
Kms is 1/Cms.
The height of our impedance peak tells us something about the BL, because we know that
BL²=(Zmax-Re)*Rms -> BL = sqrt((Zmax-Re)*Rms)
Since Re and Rms are both assumed to be constants, our only variable is the peak of the impedance (Zmax).
So in a sense we can now make a graph where you plot the voltage and/or the current vs BL or Cms (or Kms).
If we also calibrate our microphone, we can even measure SPL at the same time.
So we can make a BL(V) or BL(I) plot, as well as Kms(V) and Kms(I) plots etc etc.
The biggest catch here, is that we only measure the average difference, so we can't see any asymmetry problems etc.
Or in other words, we at least will get some relative numbers of how the BL and Kms will behave.
In very simple words; the less shift in frequency and the more constant the peak (height) of our impedance is, the more linear the motor behaves!
As long as we make sure we compare the same sound pressure levels from different loudspeakers, we have a very nice little relative comparison!
You can do the same with Le btw.
The problem is that this doesn't work well at all with the standard resistor attenuation circuit that 99% of people use for measuring the impedance.
Ideally this needs to be done with a constant voltage source*, or otherwise with a very small reference resistor like mentioned here in this thread.
A little side note on all of this, which I think is very important but not mentioned by Klippel or even guys like Novak (https://ant-novak.com/publications/ , must reads!)
We all talk about distortion, non-linearities and IMD.
However, I have come across some speakers with quite bad shift in Cms (so fs) but mostly BL.
Especially when these drivers are being used in a BR system, this is VERY audible, since it totally changes the total tuning of the entire system.
With a particular SB Acoustics 4 inch driver, it was as bad as going from about 70-80Hz to 50Hz as well as a impedance peak that just totally changed!
If you even model/simulate these differences, you will see a huge difference.
* Official AES papers even very clearly state that T/S should be measured with a constant voltage source at 0.1V
We can make a poormans Klippels setup.
A little bit of back to the basics again to explain it;
https://en.wikipedia.org/wiki/Thiele/Small_parameters
So we have basically 5 variables, Cms (1/Kms), Mms, Rms, Re and BL.
Measuring them all isn't possible unfortunately, and therefor Klippel assumes a few things;
- Mms is constant
- Re needs to be measured, but is also assumed to be constant afterwards or needs to be measured directly after.
- Rms is "constant", or better way of describing it, so much lower than BL that we assume it to be constant.
Since Mms is constant, we know that in our impedance measurements the frequency shift is only being created by the difference (non-linearity) of the Cms (1/Kms). Because we know that:
- Cms = 1 / ((2*pi*fs)²*Mms)
Since Mms is assumed to be constant (by Klippel), the only variable we have to measure is the fs to find the difference in Cms.
Kms is 1/Cms.
The height of our impedance peak tells us something about the BL, because we know that
BL²=(Zmax-Re)*Rms -> BL = sqrt((Zmax-Re)*Rms)
Since Re and Rms are both assumed to be constants, our only variable is the peak of the impedance (Zmax).
So in a sense we can now make a graph where you plot the voltage and/or the current vs BL or Cms (or Kms).
If we also calibrate our microphone, we can even measure SPL at the same time.
So we can make a BL(V) or BL(I) plot, as well as Kms(V) and Kms(I) plots etc etc.
The biggest catch here, is that we only measure the average difference, so we can't see any asymmetry problems etc.
Or in other words, we at least will get some relative numbers of how the BL and Kms will behave.
In very simple words; the less shift in frequency and the more constant the peak (height) of our impedance is, the more linear the motor behaves!
As long as we make sure we compare the same sound pressure levels from different loudspeakers, we have a very nice little relative comparison!
You can do the same with Le btw.
The problem is that this doesn't work well at all with the standard resistor attenuation circuit that 99% of people use for measuring the impedance.
Ideally this needs to be done with a constant voltage source*, or otherwise with a very small reference resistor like mentioned here in this thread.
A little side note on all of this, which I think is very important but not mentioned by Klippel or even guys like Novak (https://ant-novak.com/publications/ , must reads!)
We all talk about distortion, non-linearities and IMD.
However, I have come across some speakers with quite bad shift in Cms (so fs) but mostly BL.
Especially when these drivers are being used in a BR system, this is VERY audible, since it totally changes the total tuning of the entire system.
With a particular SB Acoustics 4 inch driver, it was as bad as going from about 70-80Hz to 50Hz as well as a impedance peak that just totally changed!
If you even model/simulate these differences, you will see a huge difference.
* Official AES papers even very clearly state that T/S should be measured with a constant voltage source at 0.1V
Last edited:
Also pretty much all Klippel content I've consumed past 24 hours indicate Bl has most effect on distortion, whole bandwidth is affected (IMD), and quite a lot.
Here is very nice power point presentation about the stuff, more hands on stuff than the papers and helps get intuition how to read the graphs: https://www.klippel.de/uploads/medi..._Signal_Testing_in_QC_and_R_D_by_R.Werner.pdf
I would think that unless there is way to find out the cone DC offset then practicality of home measurements is limited to comparing drivers to each other. This could be done faster with THD and IMD (multitone) tests done with target SPL, so no need to start plotting BL and Kms?
If DC offset was known, then we could reduce distortion some. But you say it's hard / impossible to measure. Wouldn't measuring asymmetry of current of some symmetric test signal do, no?
edit. Klippel paper on measurement https://www.klippel.de/fileadmin/_m...Measurement_of_Large-Signal_Parameters_01.pdf
Here is very nice power point presentation about the stuff, more hands on stuff than the papers and helps get intuition how to read the graphs: https://www.klippel.de/uploads/medi..._Signal_Testing_in_QC_and_R_D_by_R.Werner.pdf
Isn't this basically just plot of impedance peak value and frequency with various input voltages and the smaller the point cloud (of impedance peaks) the better the driver?
I would think that unless there is way to find out the cone DC offset then practicality of home measurements is limited to comparing drivers to each other. This could be done faster with THD and IMD (multitone) tests done with target SPL, so no need to start plotting BL and Kms?
If DC offset was known, then we could reduce distortion some. But you say it's hard / impossible to measure. Wouldn't measuring asymmetry of current of some symmetric test signal do, no?
edit. Klippel paper on measurement https://www.klippel.de/fileadmin/_m...Measurement_of_Large-Signal_Parameters_01.pdf
Last edited:
Yes, although keep in mind that it's Kms that is proportional to fs^2 and BL that is proportional to sqrt(Zmax).
So therefor it's not directly comparable.
But in a visual sense, yeah, just show say like 3 impedance plots at difference levels or so and you will get a pretty good indication how well a motor behaves!
To repeat it again, this MUST be done with a constant voltage source or otherwise (second best) with low ref resistor.
Yeah, that's what I meant with relative measurements.
Just comparing is mostly enough, although it's handy to have some numbers to it.
Mostly because it "talks" a little easier, if you know what I mean.
The second part, that's why I put that side note there.
There are many things important for different reasons.
Distortion and IMD is mostly important for mid frequencies.
IMD mostly when you want to use a driver fullrange (low end + mid range or more)
(without excursion, IMD is a lot less)
But it's also important to know how stable a motor is for low frequency behavior as well as getting a sense of it's max excursion capabilities for example.
Another example I use real Klippel measurements for (from Erin as well VC magazine), is to see how trustworthy things like xmax are.
Some manufactures say they have like 11mm xmax, yet the Cms or BL is already hitting 20% at a much lower value.
I think what he means, is that compliance isn't linear vs cone excursion.
Since cone excursion is a function of current, I guess you can say that resonance will change with a different current.
But it's a bit of a stretch to say it that way, although technically not wrong.
I am extremely disappointed by the fact that no one is taking measurements here on diyAudio. There are only theoretical heroes here who type more or less gross nonsense with their keyboard. So once again I spared no effort or expense to be the first again.
...to prove the distortion resulting from a mechanical resonance as such.
D2o is without 20R hence 0R
D2 is with 20R series resistor
Yeah there is also 2nd order distortion visible on the first post, but there it is rising towards lows, and about only common thing is the peak around 1kHz.
Why is this measurement plot different? what I should be looking at? By saying proving mechanical resonance makes distortion, do you refer to the 1kHz peak or something else? Or is the point here that 2nd order distortion is not much affected by the resistor, or something else? thanks!
Why is this measurement plot different? what I should be looking at? By saying proving mechanical resonance makes distortion, do you refer to the 1kHz peak or something else? Or is the point here that 2nd order distortion is not much affected by the resistor, or something else? thanks!
Now that's unfair. Did you ever end up testing and validating your assumption that current through a resistor is proportional to acoustic output? Maybe you should be the first to measure that?
You show very clearly how much you appreciate the effort behind my recent measurement in post #146.
...and do not miss #19 and #146 on your way back.
Hi, sorry I have to ask again, it is very cryptic to me. Could you tell more about the measurement in post #146, is it acoustic measurement of the same system as in post #1? Don't you have data for D3, or why omitting that? What does the D2 tell? I just do not understand what are you trying to show here.
Thanks!
Thanks!
You're not the only one.
Also the whole rant kinda beats me?
Ever thought about the concept that people have only so much time or not having the space (atm)?
Does that automatically mean people aren't allowed to brainstorm or just explain (or understand) the underlying physics?
Which btw, often already explains half of the problems and results.
Yeah, assuming #146 is acoustic measurement it shows that 2nd order harmonic is not affected (much) by the resistor, which means main source for 2nd order harmonic distortion is not in electrical domain, but in mechanical / acoustic domain.
Measurement in post #1 shows distortion measured in electrical domain, which shows that a series resistor reduces 2nd order harmonic through out the frequency spectrum. It shows all harmonics are reduced in electrical domain. This is due to following reason: to maintain fair measurement condition acoustic output was kept the same with or without series resistor. This means there was increased amplifier drive voltage to reach same acoustic output (cone excursion) with series resistor in place. As the cone excursion is the same with or without resistor, also the backEMF voltage is same on both cases, but the current from backEMF voltage is less due to the increased circuit impedance, the resistor.
Since distortion measured in electrical domain reduced with series resistor it means the distortion (current) was mostly form the driver, from backEMF voltage, which now makes less current in the circuit with the high impedance. If distortion did not reduce in electrical domain with series resistor, it would then mostly be from amplifier. Distortion from amplifier is small as plot in post #9 shows, so the distortion (current) in post #1 is indeed from the driver.
If hörnli posted 3rd order distortion plot measured in acoustic domain, it should show reduced 3rd order acoustic distortion with the resistor in place, because main source for 3rd order acoustic distortion seems to be in the electrical domain, in the motor, which translates to acoustic domain through current generated by the motor, which was reduced due to the resistor.
It's nice that there is measurements, but they have hardly any meaning to most without discussion / explanation. The stuff explains itself if everyone knew what the plots represent. It is very hard topic to explain understandably, and I've tried too many times everyone have their own way of thinking, which makes complex topics like this quite hard to communicate.
If my assumption about measurement in post #146 is not correct, hopefully there is explanation by hörnli how the data was generated, what it represents and what it shows. Perhaps there is more to it, I do not know what hörnli has in mind.
Measurement in post #1 shows distortion measured in electrical domain, which shows that a series resistor reduces 2nd order harmonic through out the frequency spectrum. It shows all harmonics are reduced in electrical domain. This is due to following reason: to maintain fair measurement condition acoustic output was kept the same with or without series resistor. This means there was increased amplifier drive voltage to reach same acoustic output (cone excursion) with series resistor in place. As the cone excursion is the same with or without resistor, also the backEMF voltage is same on both cases, but the current from backEMF voltage is less due to the increased circuit impedance, the resistor.
Since distortion measured in electrical domain reduced with series resistor it means the distortion (current) was mostly form the driver, from backEMF voltage, which now makes less current in the circuit with the high impedance. If distortion did not reduce in electrical domain with series resistor, it would then mostly be from amplifier. Distortion from amplifier is small as plot in post #9 shows, so the distortion (current) in post #1 is indeed from the driver.
If hörnli posted 3rd order distortion plot measured in acoustic domain, it should show reduced 3rd order acoustic distortion with the resistor in place, because main source for 3rd order acoustic distortion seems to be in the electrical domain, in the motor, which translates to acoustic domain through current generated by the motor, which was reduced due to the resistor.
It's nice that there is measurements, but they have hardly any meaning to most without discussion / explanation. The stuff explains itself if everyone knew what the plots represent. It is very hard topic to explain understandably, and I've tried too many times
everyone have their own way of thinking, which makes complex topics like this quite hard to communicate.If my assumption about measurement in post #146 is not correct, hopefully there is explanation by hörnli how the data was generated, what it represents and what it shows. Perhaps there is more to it, I do not know what hörnli has in mind.
Last edited:
Hi, thanks!
These are interesting measurements, since there seems to be peak around 1kHz in all the distortion measurements as well as in the impedance plot. Even in this acoustic measurement of 3rd harmonic it looks like the resistor doesn't help as much as elsewhere, which indicates there is relatively strong distortion source in the acoustic or mechanical domain around the 1kHz.
Which driver is it? and what are the box dimensions and damping? What do you think where the 1kHz region gets the distortion from?
I'm not expert on the details, I bet causes for harmonic distortion and distribution of the harmonics are likely explained in detail in various studies and white papers, like why there is lots of 2nd harmonic in acoustic domain, and not much higher order harmonics.
These are interesting measurements, since there seems to be peak around 1kHz in all the distortion measurements as well as in the impedance plot. Even in this acoustic measurement of 3rd harmonic it looks like the resistor doesn't help as much as elsewhere, which indicates there is relatively strong distortion source in the acoustic or mechanical domain around the 1kHz.
Which driver is it? and what are the box dimensions and damping? What do you think where the 1kHz region gets the distortion from?
For completeness, small addition to the above: by stating that 3rd order harmonic seems to be from electrical domain I mean that in all measurements I've seen, like yours, the added circuit impedance reduces the 3rd order harmonic (and higher) in acoustic domain much more than 2nd harmonic. This indicates most 3rd harmonic is from driver motor (electrical domain) and not from mechanical or acoustic domain while most of 2nd harmonic is not from electrical domain but mechanical / acoustic. I believe both 2nd, 3rd and higher harmonics are generated in electrical and in acoustic domain, but which ever makes the most dominates the graphs.
I'm not expert on the details, I bet causes for harmonic distortion and distribution of the harmonics are likely explained in detail in various studies and white papers, like why there is lots of 2nd harmonic in acoustic domain, and not much higher order harmonics.
Last edited:
I decided at the beginning of the project never to name the transducer.
Posting #81 this thread:
Elsewhere I mentioned the membrane area: Sd: 200cm²
1kHz ...
We see it in the impedance plot.
We see it in the sound pressure. (I will dump the full set of my microphone measurements below)
We see it in the D2 current
We see it in the D2 sound pressure
You found it in the D3 sound pressure (great find!)
Now i see it in the D5 sound pressure too
This statement, as simple as it may sound, is very significant.
It takes several measurements using different measuring methods to reliably identify the causal source of a fault. This was achieved here using rather ordinary methods.
The overlays are without the 20R (hence 0R)
The overlay in the graph of the fundamental is yellow
Last edited:
^Yeah, the distortion measurements indicate it is not from the motor.
This is significant concept with audio and understanding frequency response graphs in general, loudest sound dominates while everything else makes mere ripple to it.
Just yesterday geddes posted something https://www.diyaudio.com/community/...w-distortion-with-a-2-way.334757/post-7556734 which could be problem with your driver as well. Your test box is such it shouldn't show up in the measurements, so the problem is something with the driver.
Nice, this was the reason for the additional post, I'm glad it was helpful
This is significant concept with audio and understanding frequency response graphs in general, loudest sound dominates while everything else makes mere ripple to it.Just yesterday geddes posted something https://www.diyaudio.com/community/...w-distortion-with-a-2-way.334757/post-7556734 which could be problem with your driver as well. Your test box is such it shouldn't show up in the measurements, so the problem is something with the driver.
I have not read all this thread but the first 2 pages got me curious.
Here is the Dayton ND91-4 in a small ported enclosure. Amplifier is TPA3255. Measured anechoic at 50cm with Beyer MM1, I didn't think to check the SPL since I was interested in the relative differences. It was probably about 85dB 1m just guessing.
Below: Voltage Drive
Below: Amplifier has a bridged output so I added one 10R resistor in each phase.
Yellow: Voltage Drive
Pink: 10R resistors
Green: 10R resistors with EQ
Below: Distortion products with EQ'd 10R resistor drive.
Conclusion - Practically the same across the full range. I think some of the mess around 500Hz is port resonances.
I'll throw this in for more detail:
Here is the Dayton ND91-4 in a small ported enclosure. Amplifier is TPA3255. Measured anechoic at 50cm with Beyer MM1, I didn't think to check the SPL since I was interested in the relative differences. It was probably about 85dB 1m just guessing.
Below: Voltage Drive
Below: Amplifier has a bridged output so I added one 10R resistor in each phase.
Yellow: Voltage Drive
Pink: 10R resistors
Green: 10R resistors with EQ
Below: Distortion products with EQ'd 10R resistor drive.
Conclusion - Practically the same across the full range. I think some of the mess around 500Hz is port resonances.
I'll throw this in for more detail: