It is what the cone produces when triggered. How much possibly audible by-products is produced?
Even with 44K1 there is some signal close to the cone-breakup that could trigger it perhaps?
I do know that with or without RLC gives a difference in harmonic distortion.
Whether human hearing is capable or not, all i recall was a session in France with Michelle Reverchon and some of his engineers referring to some military study where it showed that in transients reproduction the effect of output above 20kHz was audible, whereas as tests with just a sinus tone it was not audible.
In that period we did a test in a classroom with a tweeter (AudaxTW80) producing sinus tones of i believe 30kHz. The tweeter was not viible, students were unaware of this test. But when switched on after a short while students became more nervous and agitated.
This was in the pre-CD era ;-)
that is true and with 44.1k red book we have no spectral content north of 20k (unless using NOS Dacs with poor anti image filters )
i also recall that it had been shown that the human ear mainly works like an envelope detector at these very high frequencies. a high Q peak will change the envelope for transient signals
Last edited:
Probably the products of H2, H3, ... distortion could trigger the diaphragm breakup (not necessarily in an audible way though). These would show up as increases in HD in the H2, H3, ... plots where the products are at the breakup frequency.
Successively higher orders of distortion (H4, H5, ...) are generally lower in magnitude and almost certainly wouldn't create audible breakup in 20 kHz+, I would think. And IMD probably wouldn't create anything high up enough of significant magnitude either, although I'm much less confident in that.
If i am not mistaken, our hearing sense wave front up to about 500-800hz, then the envelope front up to 6000-8000hz, and above this range sensitive for the intensity variations. As our head always moves intensity level differences (strong beaming tweeter) can be detected and can lead to automatically determining source, ie the tweeter.
With the wide directivity of the PTT1.3 i will not be surprised that i will hear nice fine defined high tones without disturbing "spotting the source" our hearing automatically does.
I get the impression that the diaphragm may get a bit wobbly wiggly as in breakup.
Even if the breakup is happening near the upper limit of human hearing. If the test sweep is a single tone sweep we will likely see the breakup in the Frequency Response, phase plot vs frequency, THD + N vs frequency plot or others. Can we hear it, who knows?
A two tone test up in the top octave may show some audible IMD side bands mixed up in the breakup.
I do not have a Purifi tweeter to test. I do have a couple of 4 inch diaphragm Compression Drivers that do breakup up in the top of the tweeter frequency range. I also have a couple of Al dome tweeters I can also take a look.
DT
It is my plan to do a two tone imd check.
The arta results , despite its limitations, are indicative.
Just responded to a post in a dutch forum with the usual bat territory , i do not hear, no music signal , extra cost, etc. Tiring but also helps on focusing.
As all speaker cones have breakup at some point on freq-range, it is logical to address this one way or the other. The cone breakup acts basically as another soundsource with ringing as one of its characteristics. For me reason enough to address it. Measurements i use to help.
The arta results , despite its limitations, are indicative.
Just responded to a post in a dutch forum with the usual bat territory , i do not hear, no music signal , extra cost, etc. Tiring but also helps on focusing.
As all speaker cones have breakup at some point on freq-range, it is logical to address this one way or the other. The cone breakup acts basically as another soundsource with ringing as one of its characteristics. For me reason enough to address it. Measurements i use to help.
Hmm some things still wonder me, looking at the Datasheet PDF.
They might even make things more complicated than how it should be but it can also be a taste thing.
With one hand, they give - and later they take it back. (Thinking of the waveguide's hump in raw SPL, tamed later with the RLC).
1. Quite low resonance freq - usually reached with ferrofluid but no mention of using it at all.
Any info on this ?
2. Page 6, the envisioned LR4 High Pass is NOT LR4 imo. I would call it LR3 if that existed (comes to the Q actually I think) but isn't this a 3rd order XO since it only contains 3 capacitor/inductor elements instead of 4 ?
Ringing is noticeable at higher orders however still okay for a passive setup (real 4th order with 2 inductors and 2 caps isn't recommended for passives) but phase shift isn't something one can easily negate, being 270° with 3rd order (so polarity reversal won't help much but nothing is black & white in audio).
3. A notch filter. And then attenuation.
Advancements (for already told reasons) made with the faceplate design & using a waveguide (a subtype of horns, still) and then the other side of the coin kicks in with the additional recommended extra circuitry to tame it. I really get it, maybe less compromise is made here by design despite the extra passive elements but I still strongly believe nothing beats a direct connected tweeter and DSP based active XO. (Maybe just 1 series cap for protection but then with a parallel inductor to make it a 180° shifting 2nd order design - still as a protection way below the real digitally-set XO point - so a polarity swapping trick might be useful).
I would never use this one for a passive design (no offense), however easy in a DSP-active chain (like most loudspeaker drivers) but then I'd rather go for planar and AMT with a ruler-flat impedance...
... because increasing impedance (due to attenuation without a proper L-Pad's parallel resistor) might also trick some tube amps, both conventional and OTL designs. I know the Purifi engineers are only looking forward and "what are tubes?" right ? But seeing this as a product on the market and maybe owning a tube amp, I wouldn't use this one with the PDF-mentioned handful of passive elements. Again, a planar/AMT has much different parameters but at the end of the day, sitting in their final places and setup, I think they are a better tradeoff than a "highly optimized" dome with a bunch of passive-element-based corrections before itself.
Sounding "raw" due to missing (needless) extra corrections always wins (I mean passive corrections like here).
Maybe I'm wrong here and there, just some thoughts. But I really respect the engineering effort put in here on physical level especially.
Thanks !
some quick answers:
1. no ferrofluid. also, i don’t think ferro fluid lowers the resonance ? why would that be other than adding some mass loading?
2. the response is LR4 acoustically.
3. the notch is to address the main dome breakup and has nothing to do with the waveguide. the notch can be left out since it operates way above the audible range.
Why would you not use it for passive ? directivity is important and cannot be fixed by filters.
cheers
Lars
some quick answers:
1. no ferrofluid. also, i don’t think ferro fluid lowers the resonance ? why would that be other than adding some mass loading?
2. the response is LR4 acoustically.
3. the notch is to address the main dome breakup and has nothing to do with the waveguide. the notch can be left out since it operates way above the audible range.
Why would you not use it for passive ? directivity is important and cannot be fixed by filters.
cheers
Lars
I used Arta to measure distortions, with different sample frequencies as only change in setup, 1,3Vrms , 340mm from baffle, Beyerdynamic M1 calibrated, Thomann t-amp, tweeter in testbox with baffleshape of my GAYA2 enclosures:
192kHz, without RLC:
192kHz with RLC:
Now 44K1 without RLC:
44k1 with RLC:
So for me it shows that if no signal above 22kHz, the conebreakup by-products are not measurable, but if there is signal the cone breakup by-products are measurable.
The distortion levels are anyhow low ;-).
Here measurements at 96kHz, with distortion in percentages:
96kHz without RLC:
96kHz with RLC:
192kHz, without RLC:
192kHz with RLC:
Now 44K1 without RLC:
44k1 with RLC:
So for me it shows that if no signal above 22kHz, the conebreakup by-products are not measurable, but if there is signal the cone breakup by-products are measurable.
The distortion levels are anyhow low ;-).
Here measurements at 96kHz, with distortion in percentages:
96kHz without RLC:
96kHz with RLC:
In my setup by the way in the okto dac8pro i did set a lo pass filter around 20kHz. The system itself does support sampling rates to 192kHz.
Practically not really, but in any mass-spring-damper system, the damper has some influence on the resonance frequency.
Since we have a under damped response (since Qm is always bigger than 0.707) we get:
Where ωd is the new resonance with damping and ω0 is the original one.
I'm not sure where you get the idea from that low fs is being created by ferrofluid?
The vast majority of tweeters these days that have low fs from other good manufactures (Wavecor, SB Acoustics, Bliesma etc etc etc) don't use ferrofluid.
Lower Fs is being created by adding the back chamber in combination with carefully chosen other mechanical and parameter properties.
As can be seen in the formula above, the contribution of the dampening has only a small effect on the Fs.
The main reason to add ferrofluid is just for cooling purposes.
Impressive looking specs and great to see such a simple crossover used for great rsults.
Maybe I missed it but my pet driver and speaker measurement is dynamic range/comprssion. Measuring the output at 2.83V and 10 dB louder. To me it's the most important measurement after the FR.
Maybe I missed it but my pet driver and speaker measurement is dynamic range/comprssion. Measuring the output at 2.83V and 10 dB louder. To me it's the most important measurement after the FR.
@eriksquires If you refer to my measurements, the level used adjusted for the short distance of 340mm .
At 1m and more i use 2.83Vrms, and as an extra change the voltage with 1dB steps, but only with filter applied.
At 1m and more i use 2.83Vrms, and as an extra change the voltage with 1dB steps, but only with filter applied.
also in theory the damping has no influence. the resonance frequency (eigen frequency or pole frequency) is only set by mass and compliance. However, the damping changes where the peak amplitude vs frequency is obtained.
Think about the driver with and without electrical shorting. here the Q goes from Qms to Qts but that does not change fs.
Where is the formula you show from and what does alpha mean?
cheers
Lars
thanks 🙏
Yes, compression data would be nice to include. I just don’t like the usual sweeps at different levels since they completely depend on the signal used and its duration etc. I think that the envelope (ie amplitude ) vs time for a step in input amplitude would be the most informative. here you can see how fast the coil heats up and compresses. Alternatively a sweep graph showing amplitude out vs in. however, that misses the time aspect. most compression comes from heating.
we have distortion be SPL plots that also tell something about the high amplitude performance.
In this case wiki, but you'll find this formula in EVERY other control theory book.
We're not looking at alpha, we are looking at the damping factor zeta = 1 / (2Q).
But if you really would like to know:
Not really important, it's just a different way of notation. (i was to lazy to photoshop it away)
The damped resonance frequency will be slightly different compared to the natural frequency.
Keep in mind that this is the underdamped response aka: when a resonator is hit with a step or dirac pulse and decays naturally, instead being forced.
Still this correction term won't be that significant, with a Qm of 3 or so, it's about 1-2% different.
But yes you're correct that the natural resonance frequency is only being determined by the mass and stiffness/compliance, or L and C if you want to think in lumped equivalents.
It was being said that damping never has an influence, but it depends how you look at the problem is all I'm trying to say.
Maybe that's where people get the idea from that ferrofluid will affect the Fs???
I guess in theory one could even make an argument that it wil raise the Fs in some cases, since it blocks the VC cavity.
I understand from the standpoint of reproducibility this may seem like unscientific, but they are also really useful. There was an online Canadian magazine, perhaps Sound and Vision? Something related, anyway until 5 years ago they published what you are describing and to me the biggest benefit was that reviewer comments seemed to correlate very closely with how and where dynamic range was limited. Whether a speaker was described as harsh, or shrill or having a lot of presence really could be taken from how very different the higher power output level was compared to the baseline.
I'm afraid I don't have the experience or education to know WHY these charts showed non-linear behavior (the addition of 20dB of voltage was not always equal to a 20dB rise in amplitude) or whether it is true that this can all be attributed to thermal compression or mechanical limits. I know talking to others in here always results in me talking to a wall about non-thermal compression, so it may be that if I had a chance to study this more in college I'd know why. And I must admit the burst and highly controlled tests that demonstrate compression in real time are fascinating... yet still I do not know that all compression is thermal and in my humble opinion the basic compression testing should be looked at more for better explaining listener experience than we give it credit for.
Thanks for indulging my limited experiences, and lack of any education in this matter besides anecdotal.
Last edited:
What can be done is apply noise to the unit whilsth measuring impedance, it will show a rise in impedance, then the time aspect can be captured. I noticed this while testing sb26adc tweeters, it can increase quite fast.