We're not talking about crap tweeters, we're talking about properly designed ones with true Be or Diamond domes.
This isn't how it works with hard dome tweeters. Music doesn't really contain any information above 20kHz, either due to sample rate limitations, or recording microphone limitations. Directly exciting the resonance is virtually impossible, especially if it's pushed up to something like 100kHz like with the B&W diamond dome.
The only way that the resonance comes into play is if the distortion products excite it and fold back into the audio range. Like with an aluminium dome with a 25kHz resonance. A tone plays through the tweeter at 8.33kHz. It generates a third harmonic at 25kHz, as the dome rings at this frequency the third harmonic gets amplified by the dome. This amplification is what creates elevated IMD products - and IMD products alone - because we cannot hear the actual 3rd harmonic, only the potentially created IMD products.
If you raise the breakup to 60kHz then the elevated 3rd harmonic is literally removed from the equation as there is very little energy at 20kHz in music. If you raise it to 100kHz, as with B&W, then you've pushed these issues out of the way up to the 5th harmonic.
Except you can predict exactly what's going to happen with the IMD products with a frequency sweep because you know where the tweeters breakup lies. It can only create additional IMD products at frequencies where the distortion products excite the resonance, so a very small set of frequencies indeed.
that's interressant...
I like when a little notch around 3 to 4 K, but not being expert myself (i don't know where the room & réflexions play a role here)
But there are Strange things ! if you look the 9900 revelator from SS, look at the bump between 2 to 5 K !!!!! It was the most aprreciated tweeter in the 90s' for many brands !
I like when a little notch around 3 to 4 K, but not being expert myself (i don't know where the room & réflexions play a role here)
But there are Strange things ! if you look the 9900 revelator from SS, look at the bump between 2 to 5 K !!!!! It was the most aprreciated tweeter in the 90s' for many brands !
I think too many generalizations are being made. I think the motor matters a lot. Like AMT tweeters, you can make a Be tweeter with very good or poor motor structure and it will show both in linearity and frequency response, probably also harmonic and IM distortion products.
Here's an example of the ScanSpeak Illuminator D3004/6040-00. A pretty good Be tweeter, if underdamped for the sake of making it as small as possible. Notice it really doesn't suffer the severe oil-can like resonance many are worried about, but has a modest resonance around 35kHz or so:
I was able to simulate it's use in a Focal Profile using only a simple second order filter plus padding resistor. Compare it's FR to the original aluminum (maybe al/mg?) in this simulation:
The blue line is the simulated Be tweeter in place of the original Focal Al tweeter. Of course, we have to accept Scanspeak's FR plots since I don't have the tweeter itself, but my experience is that I trust SS FR more than I do their T/S parameters.
Next, compare to the latest Focal Sopra No. 2, measured by Soundstage Networks:
My point to all of this again, is that the implementation matters a great deal.
Best,
Erik
Here's an example of the ScanSpeak Illuminator D3004/6040-00. A pretty good Be tweeter, if underdamped for the sake of making it as small as possible. Notice it really doesn't suffer the severe oil-can like resonance many are worried about, but has a modest resonance around 35kHz or so:
I was able to simulate it's use in a Focal Profile using only a simple second order filter plus padding resistor. Compare it's FR to the original aluminum (maybe al/mg?) in this simulation:
The blue line is the simulated Be tweeter in place of the original Focal Al tweeter. Of course, we have to accept Scanspeak's FR plots since I don't have the tweeter itself, but my experience is that I trust SS FR more than I do their T/S parameters.
Next, compare to the latest Focal Sopra No. 2, measured by Soundstage Networks:
My point to all of this again, is that the implementation matters a great deal.
Best,
Erik
roger that!
1stly poor linearity along with (secondarily) wider than needed bandwidth E.g. peaking. Or often the chief complaint with lousy metal domes in laymans terms >listeners fatigue, when complex music ( read, more than single notes ) is pushed too hard, 'the grunge' is folded back in band.
how to test linearity look at or compare spectral plots at expected peak levels ( ala Zaph or similar ), extra credit use 2 tones and move them around looking for certain troublesome products E.g. sums and differences like 2fo-1f1
comms system engineers write MATLAB program simulating these things all the time, sadly audio linearity (good motors) is NOT spec'd properly to do this. Intercept points like IP2, IP3 is a term used quantify linearity in systems.
Last edited:
Anyone with REW and a good calibrated microphone or OmniMic should be able to do their own linearity testing.
A simple example would be to measure FR at 70 dB. Increase input by 20 dB. Measure FR again. Subtract 20 dB from second plot and compare to original. AFAIK there's no one industry accepted linearity standard. You can select any distance and levels, but the plots Soundstage Networks uses, measured at the National Research Council of Canada seems to be at 2m distance, using 70 dB as a baseline. Here's a recent example, also from the Focal Sopra, with its Be tweeter, comparing the baseline to 95 dB:
The dips below 0 indicate compression. Without compression, the speaker's output should be 25 dB above it's baseline at all frequencies but you can see in this example that there is compression across the board, and about 3.5 dB of compression between 7kHz to 10 kHz. In this band the output increased only y about 21.5 dB, while the rest of the speaker fared better. Interestingly, distortion did not seem to increase very much during the measurements.
Using a completely different motor, the Magico S5 uses a Scanspeak Be tweeter, and has even broader linearity issues at 95 dB
which frankly is not a result I would have expected, given the much better motor. It does bring into question whether the money is worth the cost though. For comparison with a similarly priced $500-ish tweeter, the Mundorf AMT tweeters I use have nearly zero compression at these levels and distances using OmniMic.
Best,
Erik
A simple example would be to measure FR at 70 dB. Increase input by 20 dB. Measure FR again. Subtract 20 dB from second plot and compare to original. AFAIK there's no one industry accepted linearity standard. You can select any distance and levels, but the plots Soundstage Networks uses, measured at the National Research Council of Canada seems to be at 2m distance, using 70 dB as a baseline. Here's a recent example, also from the Focal Sopra, with its Be tweeter, comparing the baseline to 95 dB:
The dips below 0 indicate compression. Without compression, the speaker's output should be 25 dB above it's baseline at all frequencies but you can see in this example that there is compression across the board, and about 3.5 dB of compression between 7kHz to 10 kHz. In this band the output increased only y about 21.5 dB, while the rest of the speaker fared better. Interestingly, distortion did not seem to increase very much during the measurements.
Using a completely different motor, the Magico S5 uses a Scanspeak Be tweeter, and has even broader linearity issues at 95 dB
which frankly is not a result I would have expected, given the much better motor. It does bring into question whether the money is worth the cost though. For comparison with a similarly priced $500-ish tweeter, the Mundorf AMT tweeters I use have nearly zero compression at these levels and distances using OmniMic.
Best,
Erik
Last edited:
only if yer measuring stuff using the wrong rulers, I continue to claim bandwidth is not a measurement of goodness in most anything, it's a tradeoff like everything else in Engineering.
yes compression is related to linearity, but I think the testing methods can be too hard on voice coils > steady state heating, perhaps a dynamic test. E.g. look at the harmonics expected growth, when they deviate by 1dB from theory. look up IP3 testing. danger https://www.maximintegrated.com/en/app-notes/index.mvp/id/5429
Last edited:
Obviously motor linearity matters a great deal. If a motor didn't produce any harmonic distortion then all this talk about dome materials would be moot. No additional IMD products would be created, period, from the metal dome resonance. And a simple notch filter would provide complete freedom from any metal dome issues. Obviously the notch filter only matters if your music contains anything above 20khz anyway.
The fact here though is that you can get very linear motors for relatively little, the aforementioned SB Acoustics alu dome, for example, has a superb motor.
The fact here though is that you can get very linear motors for relatively little, the aforementioned SB Acoustics alu dome, for example, has a superb motor.
you keep holding on to that. anybody who works with Class D amps ought to know better. push hard and junk + sampling folds back. it's even on the TI plots!
this is mechanics at the cone, how much is an electrical notch filter going to damp the ringing when excited by products of tones>?
Last edited:
Yes because it's the only thing that matters.
You need to excite the resonance in some way otherwise it's harmless. I do wonder if you've read all that I've written because I explained this above.
The only way to excite the resonance is either directly or with harmonics.
Directly is rather difficult as 99.99% of our music contains no information above 22kHz and pretty much all metal domes break up higher than this. I am not going to discuss using equipment up to the walls (over driven class D) as another way of directly exciting the resonance as that has no place in a high fidelity system. Get a power amp with more voltage swing. And even if the resonance was being excited directly, a notch filter would work adequately to suppress this problem.
The other way is with harmonic distortion products generated by the tweeters motor itself (if your electronics are throwing out lots of crud replace them). Now if you are using drivers correctly the only distortion products being generated, in any appreciable quantity, are the 2nd and 3rd orders. If you're pushing your tweeter hard enough such that it's going to start spitting out gross quantities of 4th and 5th then, again, we're not talking hifi any more.
Nevertheless even when the tweeter isn't being pushed hard, the distortion products are going to be amplified by the domes resonance. I mentioned this in post 62, did you not read this? Raise the breakup high enough up in frequency and you're not even going to excite the resonance with the 5th harmonic with most music.
If your metal tweeter sounds great at low/moderate volumes, but not at high volumes then you need to think of why. If your electronics are capable then the only thing you need to look at is the loudspeakers distortion.
Typically 2nd and 3rd order distortion products will be quite high relative to the others and will rise, roughly, in proportion to the drive level. All other orders should be very low and remain low until the tweeter is pushed too hard, then they will shoot up. If this is the point at which your tweeters are melting your ears then it's no wonder, the raised high orders will excite the domes resonance, but at this point any dome tweeter is going to sound bad.
My personal experience differs from your findings. The TL line of tweeters provide a nice study in this. Buy the carbon and compare to ceramic, magnesium, or even Be. Measure above 20K and you will see that the different models have spikes the vary in magnitude, although all are in the 22-25k range. The carbon is the worst with about a 15dB spike and i believe it is very apparent what the effects are when listening. It is an otherwise excellent tweeter with flat response. If memory serves, JohnK dropped the Audax Gold because of a similar problem with it. I have no graphs left to support the above. Tossed them.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.