Improve the Seas TBFC/G?

[Quwiksilver]

Gurus:

I've built a system using the TBFC/G, and it sounds fantastic. Here's the question:

Would notching out the big ultrasonic spike at 26K help reduce the different related harmonic distortions in the audible range? For example, there's a nice 3rd order peak at around 8.5K, and a 5th order peak at 5.2K. If notching out woofer breakup helps control (but not eliminate) it's midrange harmonics, why not treat the tweeter in the same manner?

Thanks,
Scott

Graphs are stolen directly from Zaph...thanks Zaph.

[Dennis H]

Notching out the resonance doesn't stop it from being stimulated by the lower frequencies (the lower freqs aren't filtered out) so you can't reduce harmonic distortion with a notch filter. That's true of both mids and tweeters.

[Quwiksilver]

Notching out the resonance doesn't stop it from being stimulated by the lower frequencies (the lower freqs aren't filtered out) so you can't reduce harmonic distortion with a notch filter. That's true of both mids and tweeters.

Not at all?

[joecarrow]

I think that this is going to be a sticky issue for quite a while. There's controversy about whether or not the 26 kilohertz output is audible to anyone, with possible manifestations being beat frequencies with other content generating audible subharmonics, or the resonance above the audible range causing intermodulation distortion.

I think there's another way to look at the root of this (without really thinking about whether it's bad or not). This might also be a chicken/egg sort of thing. Say the motor would provide the same distortion from across the range, but the cone has resonances at multiples of frequencies within the pass band. There's no way to measure that distortion without the cone's presence, so it's going to exhibit peaks in the distortion directly related to cone behavior. Does that make sense?

5th order distortion above 5khz has questionable effects on the sound. The distortion products are above human hearing, so does it matter?

I'm probably talking in circles a bit right now, so I'll stop there. It's food for thought.

[Dennis H]

The 26kHz tone will get stimulated by the lower 1/2, 1/3, 1/4, whatever, fundamentals and there's really nothing you can do about that in the crossover. But the big question is, can you hear a 26kHz tone? I can't but I'm old and I've listened to too much loud music. Some people claim they can hear it and their dogs certainly can.

[Davey]

Scott,

A practical experiment is the best answer. Give it a try.
Set up a controlled test (very close microphone placement) with the driver driven directly and then via a 26khz notch filter. (All the test equipment must have response beyond the frequency of interest.)

See what you see and post the results.

Cheers,

Davey.

[Quwiksilver]

The 26kHz tone will get stimulated by the lower 1/2, 1/3, 1/4, whatever, fundamentals and there's really nothing you can do about that in the crossover. But the big question is, can you hear a 26kHz tone? I can't but I'm old and I've listened to too much loud music. Some people claim they can hear it and their dogs certainly can.

Oh no! I definitely wasn't claiming I can hear at 26K, but I sure can at the 8.5K, 6.5K, etc.. But if I'm understanding you correctly, notching out the ringing (even if i can't hear it directly) will NOT help with the resulting harmonic distortion peaks. Correct?

Overall, I like your suggestion Davey. As soon as I've got some measuring equipment I'll give the experiment a shot. That is, unless someone else beats me to it...

Scott

[Dennis H]

Oh no! I definitely wasn't claiming I can hear at 26K, but I sure can at the 8.5K, 6.5K, etc.. But if I'm understanding you correctly, notching out the ringing (even if i can't hear it directly) will NOT help with the resulting harmonic distortion peaks. Correct?

Right but the distortion peaks will all be at 26K. So, if you can't hear that, it's no big deal. But someone you invite over, like a young woman with exceptional hearing, might be able to hear it so it's still something to consider.

[joecarrow]

Basically the way harmonic distortion works is this: When a tone is played, a whole-number multiple of its frequency is played as well. Second harmonic of 1 khz is 2 khz. Third harmonic of 8.5 khz is 25.5 khz.

At least, that's how I understand it.

[Quwiksilver]

Right but the distortion peaks will all be at 26K. So, if you can't hear that, it's no big deal. But someone you invite over, like a young woman with exceptional hearing, might be able to hear it so it's still something to consider.

I don't follow. That doesn't seem to have anything to do with the harmonic distortion peaks that ARE in audible range. How will they all be at 26K?

[Chris7]

Be careful not to confuse linear distortion with nonlinear distortion. The amplitude peak is linear distortion. Harmonic distortion is nonlinear. You can't directly reduce nonlinear distortion by modifying the amplitude peak.

That's not to say there won't be any effect at all on nonlinear distortion if you notch out the peak, just that it will be a minor secondary effect (e.g. less excursion, therefore less HD everywhere). It's worth a try if you have measurement gear, but it won't have the major effect you might intuitively expect.

[cotdt]

you can coat the dome with some material to dampen the peak.

[JonP]

Yeah, but then you're REALLY changing things... What would be a slight film on a woofer will be a heavy thick layer to a tweeter dome. T/S would probably change a lot.. FR as well... I'd guess it would be also hard to control, and end up with two similar tweeters afterwards.

I'd like to see someone do the notch filter test, and see if it helps or reduces distortion products. AFAIK, it's possible for tones to mix in your ear, and produce IM distortion products inside... especially if it's loud. So, it might still be a problem, even if it dosen't produce IM distortion in the driver itself.

I wonder if a notch could produce some electrical damping and thereby reduce the off frequency driven ringing? (have to be a paralell network?) Would be an interesting experiment...

[capslock]

All has been said in this thread, but maybe not clearly enough.

The cone has a mechanical resonance at 26 kHz. The signal going into the driver should have no energy at 26 kHz, unless the amp produces lots of distortion, the antialias filter in your DAC sucks or you are listening to one of the rare SACD / DVD-A that actually have signal content up there. IF there is really an electrical signal with content at 26 kHz, then a notch might help.

The cone resonance does get excited my nonlinear distortion in the very motor which will take an electrical input signal at 8.7 kHz and output a mechanical oscillation at 8.7 AND 26 kHz. No notch is going to do anything about this.

What will help is reducing the distortion of the motor, either by modding the motor (undercut pole piece, add copper cap) or by using current drive instead of voltage drive. By the way, I tried to discuss suitable amp topologies both here and at diyaudio, but response was minimal.

You could also try to add damping or stiffening at strategic places of the dome to suppress the mechanical resonance. I have a hunch that adding a little extra stiffness towards the outer edge might help. You could try to attach a few balsa wood ribs from the outside. This will look ugly, but who knows, it might help!

[dlr]

Anything applied to the dome will reduce sensitivity

You could also try to add damping or stiffening at strategic places of the dome to suppress the mechanical resonance. I have a hunch that adding a little extra stiffness towards the outer edge might help. You could try to attach a few balsa wood ribs from the outside. This will look ugly, but who knows, it might help!

I haven't tried adding to a tweeter dome, but even small amounts of damping material applied to a mid-woofer quickly reduces the sensitivity due to the added mass. Tweeter domes being extremely light will be affected rather severely I would think.

The FR linearity in the primary passband will likely be altered considerably as well.

[Dennis H]

I don't follow. That doesn't seem to have anything to do with the harmonic distortion peaks that ARE in audible range. How will they all be at 26K?

The fundamentals are in the audible range but the harmonics are higher.

2nd harmonic of 13K is 26K
3rd harmonic of 8.7K is 26K
4th harmonic of 6.5K is 26K
5th harmonic of 5.2K is 26K

So, if you were looking at a plot of 3rd harmonic distortion vs. frequency, you might see a peak at 8.7K because of the tweeter's resonance at 26K.

[JonP]

All has been said in this thread, but maybe not clearly enough.

The cone has a mechanical resonance at 26 kHz. The signal going into the driver should have no energy at 26 kHz, unless the amp produces lots of distortion, the antialias filter in your DAC sucks or you are listening to one of the rare SACD / DVD-A that actually have signal content up there. IF there is really an electrical signal with content at 26 kHz, then a notch might help.

The cone resonance does get excited my nonlinear distortion in the very motor which will take an electrical input signal at 8.7 kHz and output a mechanical oscillation at 8.7 AND 26 kHz. No notch is going to do anything about this.

Could add non oversampling DAC's to the list as well..

I agree to a point, and understand what you're saying, but what I'm thinking of is that when, for whatever reason the cone gets excited at 26khz, it will produce back EMF in the motor, and whatever electrical damping in the crossover exists should damp that mechanical vibration, irregardless of how it got created. As in, if you could add a paralell notch filter, it would look like a (near) short at 26Khz, and you would get a large damping effect at that frequency. A series notch filter would not work this way, and would only block energy at 26khz from the amp.

Now, I may be wrong, or it may not be possible to have a paralell notch filter, or maybe the driver/filter interaction isn't like I understand it to be... I need to try and model this.. but it seems to be a reasonable concept. What do you think?

[dlr]

Ultrasonic resonances aren't reflected back through the motor

I agree to a point, and understand what you're saying, but what I'm thinking of is that when, for whatever reason the cone gets excited at 26khz, it will produce back EMF in the motor, and whatever electrical damping in the crossover exists should damp that mechanical vibration, irregardless of how it got created.

These resonances and almost all breakup resonances of larger drivers, not just tweeters, are not reflected back through the motor and cannot be controlled at all by the crossover.

The only mechanism by which they can be handled (as opposed to controlled) is to rolloff the crossover low enough that there is no significant signal there nor any significant distortion components that coincide with the area in question. This is not an option for tweeters, of course.

[cotdt]

how about using a cloth grill to block the upper frequencies.

[JonP]

These resonances and almost all breakup resonances of larger drivers, not just tweeters, are not reflected back through the motor and cannot be controlled at all by the crossover.

Ahhh... Guess I just demonstrated the weak point of thought experiments, and why it's always much better to get in the lab and get your hands dirty... You never are sure you haven't left out a crucial part that blows the whole theory...

So, my unspoken assumption that voice coil ALWAYS tracks the cone, falls apart when the cone is in breakup... (kind of obvious if you think about it, like I didn't) and thereby the ability to electrically damp it falls apart as well.

I still wonder if some effect might be available... since I'd guess that breakup might vary on how uncoupled with the voice coil a particular mode would be.
Fundementally it would be mostly uncoupled, but if you got lucky and a particular mode had moderate coupling, you would reduce it some. I have no real clue if these modes tend to be 70-80% decoupled, or 98-99%... so that lab time is still warrented!!

[Dennis H]

The thing is the tested curves are already done with the amp connected directly to the driver, providing a very low impedance load to back EMF. It's unlikely any passive shunt circuit would have a lower impedance at resonance so what you see is already a best case for electrical damping.

I guess the best clue to how effective electrical damping might be is the size of the impedance bobble at resonance. That's showing how much the mechanical breakup is coupling into the motor. It can be pretty big for some midwoofs but it's hardly anything for the TBFC.

[capslock]

Ahhh... Guess I just demonstrated the weak point of thought experiments, and why it's always much better to get in the lab and get your hands dirty... You never are sure you haven't left out a crucial part that blows the whole theory...

So, my unspoken assumption that voice coil ALWAYS tracks the cone, falls apart when the cone is in breakup... (kind of obvious if you think about it, like I didn't) and thereby the ability to electrically damp it falls apart as well.

I still wonder if some effect might be available... since I'd guess that breakup might vary on how uncoupled with the voice coil a particular mode would be.
Fundementally it would be mostly uncoupled, but if you got lucky and a particular mode had moderate coupling, you would reduce it some. I have no real clue if these modes tend to be 70-80% decoupled, or 98-99%... so that lab time is still warrented!!

The typical breakup node has the cone center moving out of phase with the cone edge. No way this is going to couple into the VC.