Member
Joined 2003
I don't know. What are you missing?
I can't really see what you're finding so difficult to understand. Take a major resonant mode, e.g. 25KHz. Divide by 3 and you will see a spike in HD3 around that frequency. Divide by 5 and you will see a spike in HD5 around that frequency. Kill the 25KHz resonance with a high impedance notch and you will see the distortion amplification at those sub-multiples heavily reduced. An example of this is show in the animated gif (slightly different frequencies as it's a different drive unit) I posted. I'm afraid that's the simplest way I can put it. Sorry.
I can't really see what you're finding so difficult to understand. Take a major resonant mode, e.g. 25KHz. Divide by 3 and you will see a spike in HD3 around that frequency. Divide by 5 and you will see a spike in HD5 around that frequency. Kill the 25KHz resonance with a high impedance notch and you will see the distortion amplification at those sub-multiples heavily reduced. An example of this is show in the animated gif (slightly different frequencies as it's a different drive unit) I posted. I'm afraid that's the simplest way I can put it. Sorry.
Member
Joined 2003
I can't hear anything at 25kHz, whether 20dB or 100dB, that's what I'm saying.
I'll say again, then move on from this dead horse - why should I care about anything that occurs at 25kHz? What benefit does the notch provide at frequencies heard by humans?
Comprehend?
Reduced with a notch - looks amazing, right?
Let's compare H3 relative to the fundamental, at the harmonic frequency. Nothing has changed.
I'll say again, then move on from this dead horse - why should I care about anything that occurs at 25kHz? What benefit does the notch provide at frequencies heard by humans?
Comprehend?
Reduced with a notch - looks amazing, right?
Let's compare H3 relative to the fundamental, at the harmonic frequency. Nothing has changed.
Last edited:
Dave, have you not heard that distortion products are "weighted" by the fundamental?
See supersonic peak in FR. Now see how there are similar peaks in the HD values at lower frequencies? THAT is why a peak at a HF is consequential, and not because of the primary effect of the SPL. But if you do not care about HD, then yes I guess "meh" would be an appropriate response.
See supersonic peak in FR. Now see how there are similar peaks in the HD values at lower frequencies? THAT is why a peak at a HF is consequential, and not because of the primary effect of the SPL. But if you do not care about HD, then yes I guess "meh" would be an appropriate response.
HD is one measure of the audio output quality, or fidelity of you prefer. So, do you care about that? You are welcome to define "audible" to match your own preferences or tastes, but nonetheless the HD is overall a measurement of how true to the original the signal reproduction from the driver will be. The loudspeaker is the worst part of the reproduction chain, unless you like some lame forms of amplification that I shall not name here. Any and all means to make the loudspeaker (and its drivers) as clean as possible is welcomed by me.
I tend to agree. But, if we can measure it, and can tame it - specially if the small value (sometimes awkward) parts in the bin are available - then why not? If nothing else, the graph will look nicer 🙂
Dead horse indeed...
Member
Joined 2003
Thank you @Draki for understanding 🙂. There can be a compulsive need to correct things to perfection without consideration for actual audible outcome. I'm all for over-engineering, it's the nature of Hi-Fi, but trying to focus that over-engineering on areas where I have some hope of hearing an improvement or difference.
DcibeL I accompany you, I wouldn't likely perceive stuff above 20kHz.
HD distortion plots are easily interpreted wrong, there is no subharmonics although it looks like it. Usually a HD graph shows peak at fundamental but the harmonic acoustic output is not at the fundamental but at it's frequency, multiple of the fundamental. Peak in HD3 graph at 8.3kHz is not sound at 8.3kHz, thats the fundamental which has a HD3 to it, which outputs at the 25kHz as sound. Attenuating the 25kHz would drop HD3 of 8.3kHz, likely inaudible change in sound because sound at 25kHz got attenuated. The graph cleans up though.
I once bought parts to try it at HF, but the notch wasn't even close where it should have been so abandoned the idea 😀 for woofers, easy to make, quite affordable still, and harmonics piling up somewhere mids must be more audible in comparison so might be worth it most of the time.
HD distortion plots are easily interpreted wrong, there is no subharmonics although it looks like it. Usually a HD graph shows peak at fundamental but the harmonic acoustic output is not at the fundamental but at it's frequency, multiple of the fundamental. Peak in HD3 graph at 8.3kHz is not sound at 8.3kHz, thats the fundamental which has a HD3 to it, which outputs at the 25kHz as sound. Attenuating the 25kHz would drop HD3 of 8.3kHz, likely inaudible change in sound because sound at 25kHz got attenuated. The graph cleans up though.
I once bought parts to try it at HF, but the notch wasn't even close where it should have been so abandoned the idea 😀 for woofers, easy to make, quite affordable still, and harmonics piling up somewhere mids must be more audible in comparison so might be worth it most of the time.
Last edited:
If they're not audible, i.e. the baseline motor distortion levels are sufficiently low that even the amplified peaks are at a low enough level not to cause issues -then they're not an issue. QED. Alas, very few tweeters have motors that good. 😉
Okay, hear me out here-
Does this mean the DXT as pictured above, which is considered robust and can xover low, should not be used below 8kHz because 2nd order rises below there when viewing harmonics at the aligned audible frequency?
It seems counterintuitive to suggest suppression of the fundamental to reduce harmonic peaks at that frequency when they are both inaudible to the human ear and at the same time more suppressed than would be audible if at -50dB.
However, it seems most take the HD2 as shown for second order rise and attempt to xover above where it rises. No one has past corrected my typical usage this way, so as such I may actually have fallen prey to this interpretation as well. Fortunately, it doesn't seem to have impacted the results of my projects. However, now I question my implementation methods.
Does this mean the DXT as pictured above, which is considered robust and can xover low, should not be used below 8kHz because 2nd order rises below there when viewing harmonics at the aligned audible frequency?
It seems counterintuitive to suggest suppression of the fundamental to reduce harmonic peaks at that frequency when they are both inaudible to the human ear and at the same time more suppressed than would be audible if at -50dB.
However, it seems most take the HD2 as shown for second order rise and attempt to xover above where it rises. No one has past corrected my typical usage this way, so as such I may actually have fallen prey to this interpretation as well. Fortunately, it doesn't seem to have impacted the results of my projects. However, now I question my implementation methods.
to rephrase it, you mean the 8K HD3 is NOT due to a 2K origin that creates a H3 peak there, But instead is a fundamental disto at 8k Hz marked H3 that will hearable at x3 so 24K hz ?
It's not clear to me why there is no H3 sound at 8K while there is a peak there ! why they just no show on the HD3 orange currve a peak at 25K hz instead ?
I'm just looking measurements looking at H3 peaks for instance, just hoping it will not polute the same fundamental and H2 frequency in the threshold of audibility (= don't care if the H3 is at -40 dB, and even less if the H5 is even less) ? Or if in the threshold of audibility at this frequenty, accept its effect in relation to the fundamental and H2 : clearly hopping it is below H2. So I was wrong there ?
So not sure I understood what you wrote.
As I understand it : ideal would be a mic that has flat H2 H3 etc, each below the above number : so H7<H6<H5<H4<H3<H2<fundamental in spl level, idem for the driver, in order the well recorded instruments exhibit their harmonic structure we need to recognize them; i.e. for illustration some instruments at some frequencies have more odd orders than even and so on.
For the reader I will attempt to clarify what is being discussed here
1) peaking/ringing of the hard dome in the ultrasound band (tail or base of arrow)
2) effects on harmonics at the frequency where they fall (tip or head arrow)
3) reducing the resonance at 25KHz down to 80dB (see: tail or base of arrow), ALSO reduces all harmonics (head or tip of arrow)
@DcibeL is not asking whether the ultrasonic notch filter is removing the peaking/ringing. Clearly it is.
He is also not asking whether the harmonics are reduced- clearly it is: all the harmonics that fall at 25KHz are also reduced when the ultrasonic notch filter is introduced.
What he is asking is this- Is there any other reason for using an ultrasonic notch?
Because if we can't hear above 20KHz, why should we care about the fundamental at 25KHz, or the harmonics that fall at 25KHz?
So I guess it depends on the position that one takes- can you or I hear above 20KHz? There is a recent to study that showed that a group of people could not reliably hear, nor the brain be measured to able to detect when ultrasonic content had been removed from the speaker.
I've had my hearing tested and I for one cannot hear above 16KHz.
For me, there's no need for tweeters that extend above 20KHz, and also no need to notch out ultrasonics.
@wolf_teeth
When determining the lower cut-off point at the tweeter, for some years it has been a habit for some of us to look at the harmonics at the frequency where they are excited.
Here at 3KHz, the fundamental is 92dB, it's 2nd harmonic is at 42dB (-50dB),
At 1KHz the fundamental is 88dB, it's harmonic is at 56dB; (-32dB), which is higher.
Based on this representation, we might consider an ideal Fc somewhere below 3KHz but above 1KHz.
Please do NOT look at the chart where the harmonics fall, that will just confuse things further.
1) peaking/ringing of the hard dome in the ultrasound band (tail or base of arrow)
2) effects on harmonics at the frequency where they fall (tip or head arrow)
3) reducing the resonance at 25KHz down to 80dB (see: tail or base of arrow), ALSO reduces all harmonics (head or tip of arrow)
@DcibeL is not asking whether the ultrasonic notch filter is removing the peaking/ringing. Clearly it is.
He is also not asking whether the harmonics are reduced- clearly it is: all the harmonics that fall at 25KHz are also reduced when the ultrasonic notch filter is introduced.
What he is asking is this- Is there any other reason for using an ultrasonic notch?
Because if we can't hear above 20KHz, why should we care about the fundamental at 25KHz, or the harmonics that fall at 25KHz?
So I guess it depends on the position that one takes- can you or I hear above 20KHz? There is a recent to study that showed that a group of people could not reliably hear, nor the brain be measured to able to detect when ultrasonic content had been removed from the speaker.
I've had my hearing tested and I for one cannot hear above 16KHz.
For me, there's no need for tweeters that extend above 20KHz, and also no need to notch out ultrasonics.
@wolf_teeth
When determining the lower cut-off point at the tweeter, for some years it has been a habit for some of us to look at the harmonics at the frequency where they are excited.
Here at 3KHz, the fundamental is 92dB, it's 2nd harmonic is at 42dB (-50dB),
At 1KHz the fundamental is 88dB, it's harmonic is at 56dB; (-32dB), which is higher.
Based on this representation, we might consider an ideal Fc somewhere below 3KHz but above 1KHz.
Please do NOT look at the chart where the harmonics fall, that will just confuse things further.
One could argue about the pros and cons of the log chirp aka exponential sine sweep to determine the Fc of the tweeter in the playback of dynamic music, but that's beyond the scope of the current topic.
Last edited:
Member
Joined 2003
The kicker in comparison of the above two graphs, is that the relative level between fundamental and harmonic level at any frequency is unchanged. In the measurement with the notch filter in place, the fundamental and the harmonics are reduced by the same amount. There never was a "harmonic" problem per se, only a peak in fundamental response due to a resonant mode. It bears repeating, this graph shows a direct comparison of relative harmonic level, using the harmonic frequency as the fundamental reference, there truly is no meaningful change with respect to harmonic levels relative to the fundamental, the notch simply allows it all to be reduced by the same amount.
If anyone knows the secrets of an ultrasonic notch filter, it would be @lrisbo and his team.
If anyone knows the secrets of an ultrasonic notch filter, it would be @lrisbo and his team.
Hi, yeah it depends how your graph settings are. As Dcibel shows above you can swith to show the HD where it actually outputs, by checking "Plot harmonic at the harmonic frequency" in REW. By default HD plots show the fundamental which causes it. If you have a "subharmonic" of a resonance in your graph, change your plot settings and you see all the "subharmonics" line up at the resonance instead.
It's a useful setting to be able to switch, you can see either what frequency causes HD, or at which frequency the HD outputs. Just be mindful which plot you are looking at.
Last edited:
This point is very easy explained and experienced - compare a 8kHz square signal with an 8kHz sine signal. Can you hear a difference?
https://onlinetonegenerator.com/
Why? These signals LOOK completely different?
That's the same reason why most people interpret impulse responses wrong 😏
A 8kHz square has it's main wave at 8kHz and the next one at 24kHz - which we can't hear. So they sound identical.
The same with THD - when you have H3 at 8kHz - you can't hear it cause the sound will happen at 24kHz.
You would need a steep 20kHz filter (or 16-18kHz more realistically) to interpret an impulse response right. AFTER the filter the sine and square wave would look exactly the same. Do we use such filters in audio tech ... mhhhh ... 😉
When comparing impulse responses of midrange/woofers you would need to put it after a filter to cut the part of the frequency range you never will use. Some would be surprised how boring these comparisons would get.
https://onlinetonegenerator.com/
Why? These signals LOOK completely different?
That's the same reason why most people interpret impulse responses wrong 😏
A 8kHz square has it's main wave at 8kHz and the next one at 24kHz - which we can't hear. So they sound identical.
The same with THD - when you have H3 at 8kHz - you can't hear it cause the sound will happen at 24kHz.
You would need a steep 20kHz filter (or 16-18kHz more realistically) to interpret an impulse response right. AFTER the filter the sine and square wave would look exactly the same. Do we use such filters in audio tech ... mhhhh ... 😉
When comparing impulse responses of midrange/woofers you would need to put it after a filter to cut the part of the frequency range you never will use. Some would be surprised how boring these comparisons would get.
Nevertheless - I really dislike the sound of these "low membrane resonance" Seas tweeters. Like in the older Genelec monitor speakers.
And I always prefer the sound of tweeters with high membrane resonance (Beryllium or modern metal tweeters) or nearly no resonance (ring radiators).
Not sure what it is but I look after pistonic behaviour of membranes in their use range and if possible to extend above it.
And I always prefer the sound of tweeters with high membrane resonance (Beryllium or modern metal tweeters) or nearly no resonance (ring radiators).
Not sure what it is but I look after pistonic behaviour of membranes in their use range and if possible to extend above it.
I am getting more and more the sense that it is just the REW behaviour wrt harmonic distortions is being discussed here. Not a very good thing.