Does anybody have HD distortion sweeps...

[454Casull]

...Zaph-style, but with the driver having practical LP/HP/notch/etc. filters connected?

I would prefer to see drivers with high-Q breakup (e.g. magnesium diaphragm) but anything will work, I suppose, as long as there is a pre-filter chart for comparison.

I am curious to see if the high-sensitivity-induced distortion at Fpeak/n Hz (where n is the order of distortion) actually goes down if the peak is notched or otherwise attenuated. My theory (being blind to empirical data) is that the driver-induced distortion, that is, what we see in a regular sweep, is not affected by external filters.

[PeteMcK]

but the distortion needs an excitation voltage to appear, and if this has been suppressed by the filter...

[vacuphile]

I subscribe to your theory. Whenever the driver produces higher harmonics that coincide with the breakup F, they will hit the bell and make it ring, it would seem.

vac

[sdims]

But for what you want this kind of measurements?
Magnesium drivers have high-Q breakups 1-2 octaves higher it's reasonable working range (reasonable in terms of directivity).
If you decrease the output of the hard cone driver at high-Q breakup peak - distortion also must decrease.
My opinion that more bad things with them is:
- they often have 2-3 standalone huge peaks, so, to get perfect target slope in some designs you may need 2-3 precisely calculated&produced notches;
- they store and release energy in different way than paper/carbon/filled PP cones - this always visible at waterfall plots and impulse response.

Concerning LP/HP sections - I have some 'theory' too:
- LP-section not affect distortion at higher zone of driver's working range, because excitation tones of distortion lies in working range of driver and not affected by LP.
- HP-section decrease distortion at lower zone of driver's working range, proportionally to the order of the filter, because excitation tones lies in sloped zone of response and displacement of driver's cone definitely smaller with HP-section on than without it.

It's about THD.
But IMD also very important.

[454Casull]

Originally Posted by PeteMcK

but the distortion needs an excitation voltage to appear, and if this has been suppressed by the filter...

Obviously the magnitude of the signal is affected by the filter(s). A voltage which causes 80 dB at whatever distance will cause lower SPL (and lower associated distortion) when reduced, yes. But let's assume that we are talking about in-passband frequencies only?

[454Casull]

Originally Posted by sdims

If you decrease the output of the hard cone driver at they breakup high Q peak - distortion also must decrease.

I disagree. The way I see it, all a filter does is affect the signal going to the driver, which in itself is not the source of distortion.

For example: I have a driver which has flat response except for a massive peak at 5 kHz. The sensitivity of the driver is 90 dB/1w/1m, with a high Q peak centered at 120 dB at 5 kHz.

The breakup-related distortion peaks are:

-40 dB at 2.5 kHz for K2, 1 W -> 80 dB @ 5 kHz
-45 dB at 1.67 kHz for K3, 1 W -> 75 dB @ 5 kHz
-45 dB at 1.25 kHz for K4, 1 W -> 75 dB @ 5 kHz
-50 dB at 1 kHz for K5, 1 W -> 70 dB @ 5 kHz

At any of these frequencies, the above will happen at 1 W of input. Even if you notch out the peak, the input level at these frequencies does not change! And the filters do not seem to do anything else as the distortion is inherent to the driver. And, typically Fpeak/n will be inside the passband for 2nd through 5th order... or higher.

[454Casull]

That is to say, the distortion from an unfiltered driver will most likely only be reduced by filtering if the filtering acts to reduce the magnitude of the input signal.

[sdims]

If you see Zaph's plots, for example magnesium W18EX001, you will see that it have flat response and 10 dB relatively high-Q peak exactly at 5 kHz, huge peaks of K2-K5 exactly at 5 kHz, but nothing about relative peaks at 2.5|1.67|1.25 |1 kHz. Just because of K2 of 5 kHz excitation tone will be at 10 kHz, K3 at 15 kHz, K4 at 20 etc.

But thanks to you post - I see something new for me at K3&K5 plots - they have 10 dB peaks at 1.67|1 kHz respectively for W18EX001, but K2|K4 have nothing about those peaks. Other hard cones have similar plots of K3|K5. It's not harmonics of 5 kHz excitation tone, it looks like concentric modes of 10 dB mechanical breakup of the cone at 5 kHz. I think, that if you notch main 5 kHz breakup by 10 dB - they must lower in level. But for how much exactly - you need experiment.
PS Personally for me - such hard cones is evil. ))

[PLB]

Originally Posted by 454Casull

That is to say, the distortion from an unfiltered driver will most likely only be reduced by filtering if the filtering acts to reduce the magnitude of the input signal.

I agree, providing the impedance seen by the driver is the same.

Regards

Peter

[sdims]

454Casull, PLB
Look at this:
Nonlinear distortion testing of loudspeaker driver units

After reading this, something tells to me, that if fundamental lies in HP-region of driver's target curve with filter on - K2-K5 of those fundamental in the working range of driver will be lower than without HP. LP don't affect K2-K5 in cases if fundamental don't affected by LP. Notch will affect K2-K5 mainly at it's center frequency, but mechanical K3&K5 modes of cone breakup that lies lower in frequency must be affected too, but degree of decreasing K3&K5 depends on particular combination of drivers's breakup Q/cone material/notch/etc.
But it rather small part of the problem, because according to Klippel papers, causes of drivers nonlinearities depends on frequency, and each frequency region have different dominated causes of nonlinearities.

The only case that I can partially agree with you guys - dip in driver's response caused by impedance irregularities, with corresponded peaks in K2-K5. But if we apply small resonance circuit for flattening response at the deep - harmonics must be higher at the freq of the dip and higher frequencies after the dip, but not the same as without those resonance circuit.