Let's consider what HD is, and how HD is measured. If this driver is presented with a pure 1 kHz tone, it will produce harmonics at 2k, 3k, 4k, and 5k. That is what HD is, and in the simplist way that is how it is measured.
So the fact that there is a resonance at 5k may be related to the elevated 5th HD at 1k, but any passive or active filtering of the signal above 1k is irrelevant to the measured 5th HD.
Indeed. On the other hand with filters you don't damp the cone mode, all you are doing is levelling the axial response with regard to the rest of the response regardless of how the cone manages to produce that.. it still bends in the process.
I forgot to mention that I referred to the measurement of the driver with bandpass crossover. E.g. with the ZA14 I have the response down some 42-43dB at 9k. As far as I understand the Fourier analysis of the constructed pulse doesn't apply a frequency following filter as with old measurement methods.
If this were the case, the majority of metal cone drive units would be unusable above about 500Hz.
Yes. In the case of the W18, its main bell mode occurs at 5KHz, amplifying the innate HD3 at about 1.7KHz, HD4 at roughly 1.25KHz and HD5 at 1KHz, give or take (the L18 referred to in this context was the old model, not the H1224 he used, which has its main cone mode at a higher frequency and allowed him to use a 2KHz XO). Ergo an XO of about 1.5KHz - 1.6KHz or preferably lower is preferred with the W18.
According to the HFC measures, the W22EX has its cone mode at a similar frequency (a few hundred Hz lower), amplifying HD3 at 1.6KHz, HD4 at about 1.2KHz and HD5 at about 1KHz [nearly]. An XO around 1.4KHz is viable for many (as seen in Linkwitz's Orion) assuming a high order to help attenuate the main cone mode and reduce HD amplification. Lower would be better providing the supporting drivers can handle it, and I'm not saying it's straightforward, but it is not necessary to limit the W22EX to quite such an extreme as 600Hz.
Last edited:
For the sake of interest, here's an example of the effect of filtering (aka attenuating) stopband modes on HD lower down. This is the measured HD of the Tafal standmount speaker design, before and after the tweeter's ultrasonic breakup mode (about 26.5KHz) is suppressed in the filter.
As others have noted on occasion, some driver manufacturers have been known to present HD plots with and without filters in place.
As others have noted on occasion, some driver manufacturers have been known to present HD plots with and without filters in place.
Attachments
Last edited:
I didn’t say that. You just have to be cognizant of the actual higher order HD when you decide to use rigid-cone drivers up “high”.
As for Linkwitz (bless his soul), his work wasn’t perfect. Remember how he had to add a rear-firing tweeter to the Orion?
I would conjecture that the reduction in HD on the filtered system is because the distortion measurement was done with a broad-spectrum signal as opposed to a sine sweep. Note that the peaks still appear in the filtered graph (which is strange since you would expect the ringing to be fully botched out) which I suspect is the “actual” distortion you would see with a sine sweep, whereas the worse distortion in the unfiltered graph could be a result of some synergistic effect from the already-breaking-up diaphragm
Hope to test this soon, actually. I bet if you do a distortion sweep using a two-tone signal (one constant at the breakup frequency and one rising as typically seen), the amplified HD would be significantly worse than usual.
Last edited:
I think pretty much every post I've made on this thread indicates I'm in agreement with you on the necessity of being cognizant of HD when working with hard cone drive units. 
Yes indeed. As it happens, I preferred the immediately preceding version without the rear-firing tweeter. And I suspect the late Siegfried would have been the first to say neither he, nor any of his designs were 'perfect' and we can all point to things we would do differently (I never cared for the analogue active XO for example). They were, and are, however, a lot better than the majority.
Yes indeed. As it happens, I preferred the immediately preceding version without the rear-firing tweeter. And I suspect the late Siegfried would have been the first to say neither he, nor any of his designs were 'perfect' and we can all point to things we would do differently (I never cared for the analogue active XO for example). They were, and are, however, a lot better than the majority.
Last edited:
I wouldn't 'expect' that: it depends exactly what is being done. Either way, you have data there, giving one example of what can happen; meanwhile Accuton (to give one example) have been known to present HD data with & without filters, so there is nothing here that is news or hasn't been discussed on this forum & elsewhere on many occasions.
Scottmoose wrote
This statement bothered me a little, and I have had a day to think about it. It obviously bothers 454Casull as well...
The most accurate way to measure harmonic distortion is with a steady state sine wave signal of f1, and then measure the harmonic noise at 2*f1, 3*f1, 4*f1, 5*f1, ect. The noise at 2*f1 is expressed as a percentage, and it becomes 2nd harmonic distortion. The noise at 3*f1 is expressed as a percentage, and it becomes 3rd harmonic distortion. etc etc... A reasonable comprimise method which is much faster is to use a slowly swept sine wave so we can cover the whole spectrum quickly.
This Seas W22EX driver has a resonance at 5kHz, and it is therefore very efficient at producing 5 kHz energy. Its sensitivity at 5 kHz is about 100 dB @2.83V, which is a 12 dB increase over its nominal sensitivity.
So clearly any noise or distortion from any source (motor, spider, souround) in the 5 kHz region will be amplified by the high sensitivity of the cone at this frequency.
As Scottmoose pointed out, this is why we see a rise in HD5 at 1k, HD4 at 1.25k, HD3 at 1.7k... it all relates to the fact that this driver is very efficient at radiating 5 kHz acoustic energy.
So going back to our harmonic distortion test; If we apply a 1000 Hz continuous sinewave signal to the driver, we are going to get an HD5 measurement of -60 dB (0.1%). Any low pass filtering above 1 kHz will have no effect on HD5 because it will have no effect on our 1 kHz driving signal. Whatever is going on in the motor, suspension, or whatever to create the HD5 noise, it is still going to happen, and our resonant cone will gleefully and efficiently radiate that 5 kHz energy out into space.
Now I am not convinced that 0.1% HD5 is anything to worry about, but I am certain that applying a brickwall filter at 1.4 kHz will not change the measured HD5 distortion at 1 kHz. If a difference in HD5 was measured, I would question the test methodology. There may be many approximate methods to measure harmonic distortion, but the gold standard is what I described above: apply a steady state sinewave at f1, and then measure the response at 2x, 3x, 4x, and 5x f1.
This statement bothered me a little, and I have had a day to think about it. It obviously bothers 454Casull as well...
The most accurate way to measure harmonic distortion is with a steady state sine wave signal of f1, and then measure the harmonic noise at 2*f1, 3*f1, 4*f1, 5*f1, ect. The noise at 2*f1 is expressed as a percentage, and it becomes 2nd harmonic distortion. The noise at 3*f1 is expressed as a percentage, and it becomes 3rd harmonic distortion. etc etc... A reasonable comprimise method which is much faster is to use a slowly swept sine wave so we can cover the whole spectrum quickly.
This Seas W22EX driver has a resonance at 5kHz, and it is therefore very efficient at producing 5 kHz energy. Its sensitivity at 5 kHz is about 100 dB @2.83V, which is a 12 dB increase over its nominal sensitivity.
So clearly any noise or distortion from any source (motor, spider, souround) in the 5 kHz region will be amplified by the high sensitivity of the cone at this frequency.
As Scottmoose pointed out, this is why we see a rise in HD5 at 1k, HD4 at 1.25k, HD3 at 1.7k... it all relates to the fact that this driver is very efficient at radiating 5 kHz acoustic energy.
So going back to our harmonic distortion test; If we apply a 1000 Hz continuous sinewave signal to the driver, we are going to get an HD5 measurement of -60 dB (0.1%). Any low pass filtering above 1 kHz will have no effect on HD5 because it will have no effect on our 1 kHz driving signal. Whatever is going on in the motor, suspension, or whatever to create the HD5 noise, it is still going to happen, and our resonant cone will gleefully and efficiently radiate that 5 kHz energy out into space.
Now I am not convinced that 0.1% HD5 is anything to worry about, but I am certain that applying a brickwall filter at 1.4 kHz will not change the measured HD5 distortion at 1 kHz. If a difference in HD5 was measured, I would question the test methodology. There may be many approximate methods to measure harmonic distortion, but the gold standard is what I described above: apply a steady state sinewave at f1, and then measure the response at 2x, 3x, 4x, and 5x f1.
Last edited:
Long story short? Modern measurement systems use Fourier transform of a Dirac pulse or some derivative. A real impulse meaurement using such a pulse is highly impractical. So most measurement equipment or software apply other signals and calculate the amplitude-time relation.
The relevance of this? Only my conclusion that the output signal is measured broadband. So if a distortion peak outside the passband would be trouble, this would show up in the response graph at > -40dB.
I confess that I am not familiar with every possible approach to frequency response measurement. However, with the case of the MLS method, I believe it is not possible to attain the distortion measurements from it and so (in that specific context) it is a moot point in terms of correlating distortion to frequency response.
Regardless, the amplified HD that comes from a driver's increased sensitivity at its breakup has no relationship whatsoever to its final (filtered) frequency response in the speaker system.
Typical hard cone peaking resonance is not a real problem distortionwise, H5 at say -30dB will get totally masked when we play anything else than a sinewave at 1kHz. Peaking resonance may be a problem with music or sine at 5kHz because it typically has wider dispersion than without resonance, and that peaking energy will also get distributed "better"
Scanspeak has recently introduced line of drivers with ellipitcal coil and cone/dome "Ellipticor" Ellipticor – Scan-Speak A/S
Scanspeak has recently introduced line of drivers with ellipitcal coil and cone/dome "Ellipticor" Ellipticor – Scan-Speak A/S
Harmonic distortion at the cone breakup frequency is -well- just extra output on the cone breakup frequency. Don‘t bother if it is H2 or H5. If that extra output on the breakup frequency outside the passband measures down 40dB or more, then your total HD is < 1%. What do I fail to see? You really do not have to set your software to HD measurement mode here.
Something like this video animation shows.
Loudspeaker cone breakup (in a typical loudspeaker) 3 - YouTube
Purely for the sake of reference / completeness as the question was asked elsewhere:
https://www.diyaudio.com/forums/multi-way/317632-speaker-driver-measurements-37.html#post6440554
https://www.diyaudio.com/forums/multi-way/317632-speaker-driver-measurements-37.html#post6440554
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.