So why this ordanary 2. order passive filter example?
We see that Z for L (in series with speaker) is rising and is equal to Z speaker already under 2 K. So according to high Z source goal, the goal is achieved already from about 2k and distortion is reduced at even lover stimus frequencies.
C is used to broaden the upper part of the pass band.
We see that Z for L (in series with speaker) is rising and is equal to Z speaker already under 2 K. So according to high Z source goal, the goal is achieved already from about 2k and distortion is reduced at even lover stimus frequencies.
C is used to broaden the upper part of the pass band.
@tmuikku @lrisbo @markbakk
Here some preliminary results from a friend you might find very interesting!
He's currently working on a loudspeaker with a Seas DXT tweeter.
So we took the opportunity to see if the passive notches would also work for tweeters.
The Seas DXT specifically has a pretty nasty peak around 26kHz.
First just the DXT tweeters as is:
Only the distortion is show in the graphs btw, don't look at the distortion numbers in the legend.
Yes, that's an awful lot of higher harmonics still around 3-4kHz, ewww
(although levels are pretty low)
Next a parallel notch in series, as well as a -6dB resistor attenuator (L-pad, to bring down noise from the amplifier in his case) and a 12uF cap in series (to protect the tweeter).
The total response is electrically compensated for the -6dB, not for the 12uF.
Which is why distortion below 2kHz is every so slightly lower.
The total level was kept fairly even.
That mountain landscape is basically gone.
But also the 3rd and 4th harmonics have been lowered significantly!
In this case the RLC contained a 220ohm resistor, so tomorrow he's gonna measure again without one to create a deeper notch.
I will also ask him to make a schematic as well (just to make it more understandable for everyone )
So I guess this trick even looks very promising for tweeters as well!
Here some preliminary results from a friend you might find very interesting!
He's currently working on a loudspeaker with a Seas DXT tweeter.
So we took the opportunity to see if the passive notches would also work for tweeters.
The Seas DXT specifically has a pretty nasty peak around 26kHz.
First just the DXT tweeters as is:
Only the distortion is show in the graphs btw, don't look at the distortion numbers in the legend.
Yes, that's an awful lot of higher harmonics still around 3-4kHz, ewww
(although levels are pretty low)
Next a parallel notch in series, as well as a -6dB resistor attenuator (L-pad, to bring down noise from the amplifier in his case) and a 12uF cap in series (to protect the tweeter).
The total response is electrically compensated for the -6dB, not for the 12uF.
Which is why distortion below 2kHz is every so slightly lower.
The total level was kept fairly even.
That mountain landscape is basically gone.
But also the 3rd and 4th harmonics have been lowered significantly!
In this case the RLC contained a 220ohm resistor, so tomorrow he's gonna measure again without one to create a deeper notch.
I will also ask him to make a schematic as well (just to make it more understandable for everyone
)So I guess this trick even looks very promising for tweeters as well!
For sure! did you try if it's audible difference compared to DSP only?
The pad might make audible difference as it reduces any hiss from electronics, but not sure about the notch/resonance, as it's so high in frequency. But, as it shows up on measurements and costs only few currency units I see no reason to omit in DIY speakers that aim high
Short story, I went to local small harbour to enjoy night sky last winter. I noticed there was weird noise that seemed to be audible only if I looled certain direction, which was weird. Of course it took my attention and got searching for it. Turned out it was some electrical box on a lamp post, on close proximity it was obviojsly the sound source, but bit further away I had to look away from it to hear it! For some reason the hogh pitched noise was very audible, when it was behind me, and about mute when it was in front. So, weird as it is, if you can't hear difference DUT in front of you on these high frequencies, turn ye head
The pad might make audible difference as it reduces any hiss from electronics, but not sure about the notch/resonance, as it's so high in frequency. But, as it shows up on measurements and costs only few currency units I see no reason to omit in DIY speakers that aim high
Short story, I went to local small harbour to enjoy night sky last winter. I noticed there was weird noise that seemed to be audible only if I looled certain direction, which was weird. Of course it took my attention and got searching for it. Turned out it was some electrical box on a lamp post, on close proximity it was obviojsly the sound source, but bit further away I had to look away from it to hear it! For some reason the hogh pitched noise was very audible, when it was behind me, and about mute when it was in front. So, weird as it is, if you can't hear difference DUT in front of you on these high frequencies, turn ye head
Last edited:
To my knowledge for human hearing to hear direction (horizontally) is in a range of ~200hz to ~ 7000hz with a strong peak between 1000 and 2000hz.
In the past we did tests with a tweeter visibly hidden and play frequencies from 6000hz and higher. Simply not locatable . One exception: if tweeter beams (old cone tweeter) one can locate based on slightly moving head causing strong intensity variations.
In the past we did tests with a tweeter visibly hidden and play frequencies from 6000hz and higher. Simply not locatable . One exception: if tweeter beams (old cone tweeter) one can locate based on slightly moving head causing strong intensity variations.
sorry, poorly written and edit time over. What I mean is that the Lpad can make great audible difference already due to noise reduction, and not sure if distortion reduction is audible in addition to that, hence asking. And that it is useful to put the passives in as performance goes up, although not sure how audible it is again.
Yeah it's all very peculiar, wavelengths and head / ear size shape location stuff, and a processor that interprets
the high pitched noise I wrote the story about was very hard to localize, I had to wonder around the harbour until finally got close enough the source so that there was some other noise audible that revealed the location. Interesting bit was, that the sound was not audible at all if I turned my head (at distance), and suddenly it was when facin certain direction, still not knowing where it actually came from. Weird stuff
Not yet, therefor I said preliminary.
Well, the main goal was just to find out if it would work for tweeters as well.
If I remember well, @lrisbo had his doubts about it.
About the audibility:
Well in that case we can open a whole other can of worms when it comes down to how useful the performance of certain drivers is.
Since this addon is so simple and cheap, it also just gives peace of mind and therefor it sounds better lol.
Offtopic,For some reason the hogh pitched noise was very audible, when it was behind me, and about mute when it was in front. So, weird as it is, if you can't hear difference DUT in front of you on these high frequencies, turn ye head
I don't know what the surroundings were, but reflections can really fool you.
Just pay attention when you're on a street with high buildings on either side and an ambulance or police car passes with its sirens on perpendicular to that street.
First you hear it coming kinda from the left, although hard to locate, than all of a sudden the direction flips because of the reflections (obviously this depends a bit where you are on that street) and it basically sounds it's coming from the right direction.
When it's on the other side this flips again. For a moment it sounds like it's coming from the left again while it's actually on the right side.
@tmuikku @lrisbo @markbakk
Some more preliminary data, this time with a full deep notch on the resonance frequency.
Any results of the peaks is basically just completely gone
Next week or so, he's gonna try to levels match everything with a DSP so we can actually compare results better.
Which also leaves a problem, this DSP can only go up to 48kHz sampling rate.
So we can't compare an active notch unfortunately, because that falls out of the passband (26kHz).
Although, on the other hand, that also means a (steep) lowpass filter often around 21-22kHz , which would do similar things.
Some more preliminary data, this time with a full deep notch on the resonance frequency.
Any results of the peaks is basically just completely gone
Next week or so, he's gonna try to levels match everything with a DSP so we can actually compare results better.
Which also leaves a problem, this DSP can only go up to 48kHz sampling rate.
So we can't compare an active notch unfortunately, because that falls out of the passband (26kHz).
Although, on the other hand, that also means a (steep) lowpass filter often around 21-22kHz , which would do similar things.
And here some results that can be compared a little better.
Unfortunately there are some reflections presents.
But that only does change the prettiness of the pictures, not the underlying fundamental results.
First both responses were equalized to the same response.
He did this with VituixCAD and optimizing with a BiQuad filter block.
DSP sampling frequency was 48kHz, therefor the big resonance peak of the Seas DXT tweeter cannot be seen.
Red = notch + L-pad + 12uF capacitor (for safety)
Blue = without anything
Distortion without filters:
Distortion with passive filters;
Question from my side;
As far as I remember correctly, most systems have a pretty steep filter inside their DAC to prevent any issues with aliasing?
If that is true (which I think it is), this also automatically shows that active filtering won't do anything with that resonance peak.
Unfortunately there are some reflections presents.
But that only does change the prettiness of the pictures, not the underlying fundamental results.
First both responses were equalized to the same response.
He did this with VituixCAD and optimizing with a BiQuad filter block.
DSP sampling frequency was 48kHz, therefor the big resonance peak of the Seas DXT tweeter cannot be seen.
Red = notch + L-pad + 12uF capacitor (for safety)
Blue = without anything
Distortion without filters:
Distortion with passive filters;
Question from my side;
As far as I remember correctly, most systems have a pretty steep filter inside their DAC to prevent any issues with aliasing?
If that is true (which I think it is), this also automatically shows that active filtering won't do anything with that resonance peak.
I realize many of us are not facebook fans, but for those that have an account, in the DIY Project Pad group, there were some good investigations of the same effect after Lars Risbo posted his whitepaper wrt passive notches reducing subharmonic distortion. A few of them here:
https://www.facebook.com/groups/DIY...PJGvdkYWrDgAXyJI-lEIIxubp5Sx7Ew&__tn__=,O,P-R
https://www.facebook.com/groups/DIY...g-y1ZoNVt3BN5LacRoq4mFs2A98J-iE&__tn__=,O,P-R
https://www.facebook.com/groups/DIY...-B4kUUc84TLVr1jDviM860UG8mJ6bc8&__tn__=,O,P-R
https://www.facebook.com/groups/DIY...PJGvdkYWrDgAXyJI-lEIIxubp5Sx7Ew&__tn__=,O,P-R
https://www.facebook.com/groups/DIY...g-y1ZoNVt3BN5LacRoq4mFs2A98J-iE&__tn__=,O,P-R
https://www.facebook.com/groups/DIY...-B4kUUc84TLVr1jDviM860UG8mJ6bc8&__tn__=,O,P-R
Sub-harmonic distortion would mean that lower frequency component(s) would be created (by the breakup), but it doesn't happen here, it's misunderstanding that keeps repeating.
A distortion plot can be drawn two ways: distortion is plotted to the root frequency whose harmonic it is, or at the frequency where it actually happens, at which frequency the microphone registered it.
If you plot distortion where it actually happens, all the peaks due to the breakup would all be at the break up frequency.
If you plot to the root frequency, the peaks look like they were subharmonics but they aren't, they are just reqular harmonics for what ever reason boosted by the cone resonance at the resonant frequency.
If you translate this to audio, all the distortion peaks hear at the breakup frequency, not below it. The breakup could add harmonics, but above it. The peaks we are talking about here, and in purifi paper, are not these, they are just what ever signal in electric domain amplified by the resonance in acoustic domain.
A distortion plot can be drawn two ways: distortion is plotted to the root frequency whose harmonic it is, or at the frequency where it actually happens, at which frequency the microphone registered it.
If you plot distortion where it actually happens, all the peaks due to the breakup would all be at the break up frequency.
If you plot to the root frequency, the peaks look like they were subharmonics but they aren't, they are just reqular harmonics for what ever reason boosted by the cone resonance at the resonant frequency.
If you translate this to audio, all the distortion peaks hear at the breakup frequency, not below it. The breakup could add harmonics, but above it. The peaks we are talking about here, and in purifi paper, are not these, they are just what ever signal in electric domain amplified by the resonance in acoustic domain.
Assumtion here is that cone breakup is one resonant frequency. But as far as my knowledge goes the one causing a peak is the second mode, before that is the first mode , usually causing some dip in the response due to cancellation.
Never the less intriguing test. What is the schema of the passive filtering?
Never the less intriguing test. What is the schema of the passive filtering?
Just add series impedance.
A notch would be parallel notch so a resistor, inductor and capacitor in parallel with each other, which is then inserted in series with the driver.
Adding series impedance drops all current through voice coil, including current (harmonics) that are generated by motor non-linearities. When you drop the current, force to cone drops, acoustic distortion drops and this is what you see on the plots. The breakup is still there as it's physical property of the assembly, it boosts all current (any movement of the cone) still, but now we have preconditioned all current for it with the series impedance, including (some) motor distortion current.
A notch would be parallel notch so a resistor, inductor and capacitor in parallel with each other, which is then inserted in series with the driver.
Adding series impedance drops all current through voice coil, including current (harmonics) that are generated by motor non-linearities. When you drop the current, force to cone drops, acoustic distortion drops and this is what you see on the plots. The breakup is still there as it's physical property of the assembly, it boosts all current (any movement of the cone) still, but now we have preconditioned all current for it with the series impedance, including (some) motor distortion current.
Last edited:
The breakup is about 27kHz, so third order distortion of about 9kHz root frequency shows a peak as it should. 5th order distortion should show peak about at ~27kHz / 5, which is roughly 5.5kHz. Seems legit.
example to previous posts: sound at 5.5kHz makes the cone (dome) move, and motor non-linearity makes 5th order harmonic voltage at ~27kHz, and if circuit impedance is low at 27kHz this appears as (maximal) current as well, which makes the dome move and then boosted even more by the breakup, thus peak in distortion.
With series impedance is added at 27kHz, the distortion voltage does not make as much current anymore, so less acoustic output, thus reduced distortion measured with mic. As the distortion plot shows harmonics at the root frequency, it appears as if "subharmonic" was reduced, when in fact it's just normal 5th order distortion of 5.5kHz got preconditioned and the breakup doesn't boost it above all 5th order harmonics at 27kHz. The distortion doesn't have notch there, because there is other distortion mechanisms that are not affected by the series impedance, which maintain the distortion level where it is.
If one makes notch with DSP to 27kHz, it doesn't reduce distortion because the distortion is caused after the DSP and caused by sound at 5.5kHz, not sound at 27kHz. If you put notch with DSP at 5.5kHz, then fifth order harmonic would drop, but so would your 5.5kHz output. This is the power here, the series impedance EQ:s current, it drops the main signal at 27kHz smoothing the main signal, and also distortion current from lower frequencies generated in the motor.
Last edited: