That being said, KEF and e.g. Neumann have been breaking the conventional price/quality ratio's for about a decade or so. But the DIY Community remains faithfull to its firm beliefs. When matters go pear-shaped, invariably the "ASR-Accusations" fly all over the posts...
A few years ago i tried to swap the ke25 to scan speak d3004 beryllium, i think it was third order 2500hz.
I did alot of a/b testing and to my ears they sounded very much the same, I couldn't hear any difference between them.
The DIY/startup community are blessed to access to fine products for prototyping.
For instance SB Acoustics have a tweeter with a healthy 0.6mm x-max, copper capped pole piece.
It started life as a product for Revel as part of their Performa line in 2013.
In collaboration with @fluid, here it is mounted on a custom waveguide, with a HPF for acoustic LR4 1KHz, to match a 8” midwoofer on a 24cm wide baffle.
Measured for 96dB/1m:
What seems to be missing are tweeters with compact neo motors, or interchangeable waveguides.
Reduce reuse repair recycle!
For instance SB Acoustics have a tweeter with a healthy 0.6mm x-max, copper capped pole piece.
It started life as a product for Revel as part of their Performa line in 2013.
In collaboration with @fluid, here it is mounted on a custom waveguide, with a HPF for acoustic LR4 1KHz, to match a 8” midwoofer on a 24cm wide baffle.
Measured for 96dB/1m:
What seems to be missing are tweeters with compact neo motors, or interchangeable waveguides.
Reduce reuse repair recycle!
Following my post https://www.diyaudio.com/community/threads/some-speaker-driver-measurements.317632/post-7957155 i did test on distortion effect of a parallel RLC in series with a midrange and an active filter applied:
NoRLC:
Yes RLC:
From this test it appears that it makes a difference in distortion even if the conebreakup frequency is in the stopband. In this case filter is 3.4kHz and fully (> -200dB) down 6.9kHz, cone breakup at 8.2kHz.
Aim is to do Yes/No listening tests on thursday.
NoRLC:
Yes RLC:
From this test it appears that it makes a difference in distortion even if the conebreakup frequency is in the stopband. In this case filter is 3.4kHz and fully (> -200dB) down 6.9kHz, cone breakup at 8.2kHz.
Aim is to do Yes/No listening tests on thursday.
Am I interpreting these graphs correctly if I observe in all cases the distortion remains below some 60dB?
Yet I find it difficult to conceptualize the distortion between 7 and 10Khz is actually above the acoustic output of the driver.
Yet I find it difficult to conceptualize the distortion between 7 and 10Khz is actually above the acoustic output of the driver.
Yes i agree, i also am still struggling with the interpretation of the distortion graphs. But distortion generated by the driver could be present outside of the signal driven output. Why not ?
Arta - THD approx <1%:
Acourate - THD approx < 0.1%.
Same driver, same conditions?
If so, then I find it strange that the distortion measurements differ by 20dB: -40dB for the Arta measurement, -60dB for the Acourate measurement?
Acourate - THD approx < 0.1%.
Same driver, same conditions?
If so, then I find it strange that the distortion measurements differ by 20dB: -40dB for the Arta measurement, -60dB for the Acourate measurement?
The Arta dist figures in the pictures are at the marker at 23.6Hz, but in the midrange say 1-2kHz they are more or less equal < 0.1%
Addition to the measured distortions. These include the distortions of the audio-interface pre-amplifiers and outpur amplifiers. And those are not insignificant. As is the distortion of the microphone. But the tests give an indication of changes.
Have a break:
Screenshots 1/3 show the spectrum of all k2...k5 harmonic distortion products.
Screenshots 2/4 show the amount of k2...k5 of harmonic distortion at any k1 frequency.
Look e.g. at k5 (the black line).
Screenshot 1/3: The spectrum of k5 shows a peak at approx. 12.5kHz.
Screenshot 2/4: This same peak is visible at k1 = approx. 2.5kHz (which is well inside the passband)
5 * 2.5kHz = 12.5kHz
Nearly every driver will produce some harmonic distortion products above it's passband, especially for the higher harmonics.
Now for something completely different:
You e.g. measure a metal dome which breaks up @ 30kHz. Now look at a screenshot 2/4-type of k2...k5 graph.
This graph shows seemingly peaking of k2 at 15kHz, k3 at 10kHz, k4 at 7k5, k5 at 6kHz.
Be then aware that these peaks do not represent any real k2 ...k5. It's simply the inverse of k1 * n = kn. It's 30kHz / n = kn.
If there were any harmonic distortive products of the 30kHz cone resonance/breakup, then within a spectrum of k2 at 60kHz, k3 at 90kHz ...
Last not least and nota bene:
You showed Acourate screenshots. It's one of the specialties of Acourate that there are many and also not-so-obvious bells and wistles, leading to some risk of getting irritated by it's results. In case of pending mental crash better then immediately consult Uli Brüggemann, which most of the time will gently explain what has gone wrong ...
Instead, my wrong with these ARTA graphs was to look at the delta dB between the SPL and the kx graphs which I interpreted being some -40db one from the other, erroneously relating these graphs to the leftside db scaling. I erroneously did not pay attention to the hd scaling at the right side. Sorry.
Last edited:
I have direct contact with Uli, my measurements are not wrong in the technical sense. I face some issues with my audio-interfaces which were not logical.
Images 2 and 4 are the ones i read , the 1 and 3 are just for assurance. It is a mental thing to multiply per Hn at k1.
Images 2 and 4 are the ones i read , the 1 and 3 are just for assurance. It is a mental thing to multiply per Hn at k1.
It sort of hit me, more of a Aha kind, why Uli provides the tick box SRC (Source). I already marked my results with YesSRC and NoSRC.
So YesSRC shows the source (where in the spectrum the hd is created) of the harmonic, and NoSRC shows the frequency of the harmonic, in essence:
So the interpretation is clear now. The effect of the applied filter also.
(Of course a proper spectorgram will show the time dimension as well, but this is quite readable now)
So YesSRC shows the source (where in the spectrum the hd is created) of the harmonic, and NoSRC shows the frequency of the harmonic, in essence:
So the interpretation is clear now. The effect of the applied filter also.
(Of course a proper spectorgram will show the time dimension as well, but this is quite readable now)
I do not regularly follow this forum but I've come across this thread today.
Here I have some necessary comments:
1. when we talk about the order of a harmonic distortion we talk about a factor by which a frequency gets multiplied. So for a simple example a 1 kHz sinewave will show up a 5 kHz distortion as a k5 result. This way a driver playing in combination with a bandpass crossover can produce frequencies above the given crossover.
2a. the distortion chart can show the frequencies as measured. So e.g. in the frequency response shown above ("effect") the distortions replicate the bandpass frequencies multiplied by factor 2, 3 ...
2b. the interpretation of a picture like 2a can be confusing. It is sometimes helpful to see a direct comparison between a certain frequency of a harmonic distortion and its origin = 1st order frequency. In such a case you would always need to calculate with the according factors.
Thus another representation makes sense. By applying a samplerate conversion (this is meant by SRC here!) now all the distortions are displayed in the same frequency range. In the given example (wrongly titled as "source") you can nicely see how the shape of the 1st harmonic is replicated by the higher harmonics. So you can detect a "connection". Of course you need to keep in mind that in reality the distortion frequencies are higher than displayed.
Here I have some necessary comments:
1. when we talk about the order of a harmonic distortion we talk about a factor by which a frequency gets multiplied. So for a simple example a 1 kHz sinewave will show up a 5 kHz distortion as a k5 result. This way a driver playing in combination with a bandpass crossover can produce frequencies above the given crossover.
2a. the distortion chart can show the frequencies as measured. So e.g. in the frequency response shown above ("effect") the distortions replicate the bandpass frequencies multiplied by factor 2, 3 ...
2b. the interpretation of a picture like 2a can be confusing. It is sometimes helpful to see a direct comparison between a certain frequency of a harmonic distortion and its origin = 1st order frequency. In such a case you would always need to calculate with the according factors.
Thus another representation makes sense. By applying a samplerate conversion (this is meant by SRC here!) now all the distortions are displayed in the same frequency range. In the given example (wrongly titled as "source") you can nicely see how the shape of the 1st harmonic is replicated by the higher harmonics. So you can detect a "connection". Of course you need to keep in mind that in reality the distortion frequencies are higher than displayed.
Now i understand my confusion, i knew from the wiki it was sample rate conversion, but in my mind it was stored as source. Thanks for your input @uli.brueggemann
As i indicated in post 2684, i just did listen to the midranges only with and without the RLC in place, and with the active filter in place as well.
The difference is not trivial, that is for sure. Unexpected also, as the correction is mostly above the passband.
However a visitor just arrived, so had to stop, to be continued.
The difference is not trivial, that is for sure. Unexpected also, as the correction is mostly above the passband.
However a visitor just arrived, so had to stop, to be continued.
Interesting. I'm not entirely surprised since this was something that was being discussed / worked around over 20 years ago. One of my favourite examples is actually from John Krutke. Quoting directly, with a few square-bracketed additions:
John was [is] certainly not wrong in my experience, & I've always found the HD to be quite audible under these or equivalent [depending on exactly where the breakup peak is located] conditions; this also seems to have been one of the issues that some had with the Seas Thor kit, what's now known to be some partial misunderstandings over Joe's alignment goals notwithstanding. Name your poision anyway ladies & gents & do whatever works for you. This certainly does for me, but YMMV as ever.
...drivers like the old L18 and the current Excel W18 have the breakup node in the 4-5kHz range. Observe the W18 distortion plot in the lower right corner of this spec sheet. [don't you wish Seas, Scan & everybody still did this?] A breakup node of 4.2kHz directly correlates to a peak in 3rd harmonic distortion 1.6kHz. Sharp peaks at a particular frequency will always cause a peak in the 3rd harmonic distortion plot roughly 2 octaves below. Notching out the breakup node is not enough to get rid of its effects. [John was talking here about a series LC or LCR shunted across the driver rather than the high impedance parallel stopband notch in series with the driver that Purifi & some of us other muppets sometimes use] The only way to avoid it completely is to cross over below the frequency where the peak is not excited as a harmonic. With the L18 [original H1142 version; alas now the replacement H1224 also appears to have bitten the dust] and W18, that means crossing over at 1.5kHz or lower. Many designs don't do this, and the sound is artificially colored as a result. It's most obvious with a solo piano. For a nifty test of this, set up a W18 system with a crossover frequency of 3kHz, and a perfect LR4 response. Play a recording that runs up the notes. When you get in the range of F6, F6# and G6, the piano will suddenly become completely unnatural sounding. In most recordings however, something like this is somewhat masked, and rather than sound unnatural, the music just changes character a bit. It's still noticeable however.
Now, with the L18 H1224's new higher breakup node, you can see in the distortion chart above that the 3rd order harmonic is pushed up to 2.5kHz. This is much more useful in a 2-way design. With the crossover placed at 2kHz, even the harmonic is several dB down, enough to not be noticed anymore. Now you can see why I've been so excited about this woofer.
John was [is] certainly not wrong in my experience, & I've always found the HD to be quite audible under these or equivalent [depending on exactly where the breakup peak is located] conditions; this also seems to have been one of the issues that some had with the Seas Thor kit, what's now known to be some partial misunderstandings over Joe's alignment goals notwithstanding. Name your poision anyway ladies & gents & do whatever works for you. This certainly does for me, but YMMV as ever.
Is that peak at 1.6kHz a real output at 1.6kHz? I have not found any proof of subharmonics sofar, only harmonics e.g. 2 3 4x res frequency.
I happen to bump into a in my view plausible explanation
The message:
Excellent question!
Subharmonics come into play with distortion:
If you have an acoustic resonance at 1000 Hz in a non-perfect (read: not perfectly linear) system, then you will measure a peak in the THD frequency response at the subharmonics of 1000 Hz: you will see increased distortion at 500 Hz, 333 Hz, 250 Hz etc.
(because any harmonics created through distortion at 250/333/500 Hz will be amplified if they fall into the frequency range of the resonance)
An example:
You have a headphone. The loudspeaker that is used inside the headphone has a low THD (but not zero). For the sake of simplicity let's assume that the loudspeaker produces the same 0.1% THD at all frequencies when you measure the loudspeaker on its own. Let's also assume that the loudspeaker's distortion only produces the 2nd harmonic (=k2, first overtone), that is: the harmonic distortion only consists of the 2nd harmonic. No other harmonics are produced.
Now if you put the loudspeaker into the headphone, the acoustic system will change. For example there could be an earcup resonance at 6 kHz (like on the HD800).
Now if you measure not just the loudspeaker on its own but the whole headphone, you will see that the THD has a peak at 3 kHz (the subharmonic of 6 kHz).
Because when the headphone is playing a sound at 3 kHz, the distortion will be at 6 kHz. The distortion at 6 kHz will be amplified by the earcup resonance.
So even though the loudspeaker produces the same amount of distortion, the distortion is louder (because it gets amplified by the resonance). And the frequency at which the distortion is amplified is a subharmonic of the resonance frequency.
But you can not excite a subharmonic.
An example: let's take the same headphone as before, which has an earcup resonance at 6 kHz.
If the loudspeaker plays a 12 kHz sound, the 12 kHz will not be amplified by the 6 kHz resonance. And we will also not hear anything at 6 kHz, because nothing is playing at 6 kHz.
So no, a 50 Hz sound played through a loudspeaker can not cause the room to resonate at 25 Hz.
However: some types of resonances produce not only a single resonance frequency but also resonate at harmonics.
If the room resonates at 25 Hz, it will likely also resonate at 50 Hz (second harmonic of 25 Hz).
Whether this is the case depends on the type of resonance!
helmholtz resonators resonate at only one frequency, for example.
tube resonators also resonate at harmonics (either odd or even harmonics, depending on whether it's an open-closed tube or a symmetric (open-open / closed-closed) tube.
The message:
Excellent question!
Subharmonics come into play with distortion:
If you have an acoustic resonance at 1000 Hz in a non-perfect (read: not perfectly linear) system, then you will measure a peak in the THD frequency response at the subharmonics of 1000 Hz: you will see increased distortion at 500 Hz, 333 Hz, 250 Hz etc.
(because any harmonics created through distortion at 250/333/500 Hz will be amplified if they fall into the frequency range of the resonance)
An example:
You have a headphone. The loudspeaker that is used inside the headphone has a low THD (but not zero). For the sake of simplicity let's assume that the loudspeaker produces the same 0.1% THD at all frequencies when you measure the loudspeaker on its own. Let's also assume that the loudspeaker's distortion only produces the 2nd harmonic (=k2, first overtone), that is: the harmonic distortion only consists of the 2nd harmonic. No other harmonics are produced.
Now if you put the loudspeaker into the headphone, the acoustic system will change. For example there could be an earcup resonance at 6 kHz (like on the HD800).
Now if you measure not just the loudspeaker on its own but the whole headphone, you will see that the THD has a peak at 3 kHz (the subharmonic of 6 kHz).
Because when the headphone is playing a sound at 3 kHz, the distortion will be at 6 kHz. The distortion at 6 kHz will be amplified by the earcup resonance.
So even though the loudspeaker produces the same amount of distortion, the distortion is louder (because it gets amplified by the resonance). And the frequency at which the distortion is amplified is a subharmonic of the resonance frequency.
But you can not excite a subharmonic.
An example: let's take the same headphone as before, which has an earcup resonance at 6 kHz.
If the loudspeaker plays a 12 kHz sound, the 12 kHz will not be amplified by the 6 kHz resonance. And we will also not hear anything at 6 kHz, because nothing is playing at 6 kHz.
So no, a 50 Hz sound played through a loudspeaker can not cause the room to resonate at 25 Hz.
However: some types of resonances produce not only a single resonance frequency but also resonate at harmonics.
If the room resonates at 25 Hz, it will likely also resonate at 50 Hz (second harmonic of 25 Hz).
Whether this is the case depends on the type of resonance!
helmholtz resonators resonate at only one frequency, for example.
tube resonators also resonate at harmonics (either odd or even harmonics, depending on whether it's an open-closed tube or a symmetric (open-open / closed-closed) tube.
This is interesting.... One of my favourite examples is actually from John Krutke. Quoting directly, with a few square-bracketed additions:
... With the L18 [original H1142 version; alas now the replacement H1224 also appears to have bitten the dust
Some time ago I measured some Seas Excel W22EX001 drivers with cosine shaped tone bursts. These drivers have a strong membrane resonance at about 4.8kHz. They reproduced the toneburst quite perfectly up to 1kHz, and ok at 1.25 kHz. Instead, the 1.6kHz bursts looked markedly distorted at the end, while the 2kHz burst was looking like a misery. I now guess now that it possibly was the the effect of the ringing H3.
Because of this finding, I crossed the W22 at around 800Hz to a W12. I wonder in this respect how the Linkwitz Orion acoustically behaved between 1kHz and 2kHz because of theirs crossing at 1440Hz which in this light obviously scratches at the uppermost limit for the W22, and seems rather low for the T25CF002. If he scratched the limits, then Linkwitz did well know what he did, he use shaped tone bursts to assess his drivers, and it's his website which inspired me to do so.
If these thoughts were consistent, then any Bass or Midrange driver which shows a peaking membrane resonance/breakup might be carefully measured for it's highest feasible xover frequency. Shaped tone burst seem to be quite well suited for this task, supplementing harmonic distortion measurements.