All the discussion about cone breakup and distortion is generated by people trying to use a driver outside the intended frequency range.
Doctor, it hurts when I do it. - Don't do it.
Doctor, it hurts when I do it. - Don't do it.
In trumpets, harmonics up to the 10th one are making the sound, where the fundamental frequency is actually usually missing - in trumpets ALL the sound is made from harmonics:
From this picture, the most important are H7 and H8, while H10 is probably not very audible. So if you have 10% harmonic distortion in H7, there is a major, MAJOR audible distortion and your speaker will sound like a trashcan. I verified it myself with > 100 hours of listening tests.
Most musical instruments need at least H4 or H5 to sound naturally. I suspect in most cases the H3 is the most important however I might be wrong.
More read about trumpets here and here.
This is the full trumpet spectrum, where the most fundamentals are between 100 and 800 Hz, the 1600+ peak you see in the spectrum are all H3 - H10, there is no even H2 in that area.
So if you see the speaker measurement with "low" H2 harmonic distortion it means absolutely nothing for the real sound. Check H7.
Now realize that harmonic distortion sums up. According to above the piano in the middle register has 20 - 30 harmonics.
Suppose only the first 20 are important, and your speakers have 1% distortion across the range. Then you take 1 percent 20 times, as it's all in imaginary coordinates I suspect the formula is something about multiplying them and taking a root of 20 th degrees, not sure, but anyway you will get the distortion above 10% for sure, so it will be very audible.
To understand what harmonic distortion can do to the piano read this thread about piano producing harmonics not exactly at the exact harmonic frequencies: https://music.stackexchange.com/que...harmonics-not-being-integral-multiples-of-the
As you see Fourier of the 440 Hz note up to H5 at 2640 Hz it goes exact, but the H6 is not at 3080 but at something like 3110, and so one.
So if your speaker is producing H8 distortion at 3080 exactly, it will be double note, it will play both 3080 and 3110, and it will sound horribly. Especially bad if 3110 is the area of the cone breakup, it will inject additional distortion.
Solution:
1. Your midrange woofer should be good enough that it has no H8 distortion while playing notes around 400 to 500 Hz
2. Split the drivers. Make the crossover to low pass the woofer at 2 kHz and make the 3080 to be played by the tweeter.
As you see Fourier of the 440 Hz note up to H5 at 2640 Hz it goes exact, but the H6 is not at 3080 but at something like 3110, and so one.
So if your speaker is producing H8 distortion at 3080 exactly, it will be double note, it will play both 3080 and 3110, and it will sound horribly. Especially bad if 3110 is the area of the cone breakup, it will inject additional distortion.
Solution:
1. Your midrange woofer should be good enough that it has no H8 distortion while playing notes around 400 to 500 Hz
2. Split the drivers. Make the crossover to low pass the woofer at 2 kHz and make the 3080 to be played by the tweeter.
Here is the Fourier spectrum for few more instruments. Some have non harmonic components in the spectrum like piano discussed above. And some just reach too far.
E.g. violin playing 440 Hz produces audible harmonics up to 8 kHz. So a midwoofer cone breakup at 3 kHz will interfere with this chain right in the middle and produce bad audible distortions.
The solution is the same - low pass the woofer at 2 kHz and let the harmonics in the 3 - 8 kHz area to be played by the tweeter. Splitting the drivers is your best friend.
Agreed -the only caveat I'd attach is that if HD plots aren't available, it isn't always obvious for less experienced designers to know what the optimum upper BW limit is. For e.g., the attached. From the FR alone (& ignoring the usual edge diffraction, step loss shown in Seas's measurements < 1KHz) at a stretch you might say 2.5KHz. But when you look at the distortion performance, unless you take other action, the limit for best results is more like 1.6KHz due to the cone resonance's amplification of HD3 just above that point.
Attachments
@Scottmoose on the mark! Please see what I just posted about the particular Tang Band W6 driver in the dedicated forum:
There the woofer has a cone breakup bump at 4.2 kHz and I want to cross it at 3000. It might be still too high but the tweeter has resonance at 1 kHz so I want to fade before approaching 1 kHz. It will probably be a good compromise.
@Juhazi I don't think we are disagreeing about anything, it's just I'm trying to explain complex matters with awkward words. Please read my post about piano harmonics:
So piano harmonics are not the exact multiplier of the base frequency. Look at the 440 Hz note:
The H8 and H9 are not at exact multipliers of 3080 and 3520, they are more like 3110 and 3550. If the cone playing 440 Hz has any H8 and H9 distortion at all it will be very audible because it will play in fact at...
So piano harmonics are not the exact multiplier of the base frequency. Look at the 440 Hz note:
The H8 and H9 are not at exact multipliers of 3080 and 3520, they are more like 3110 and 3550. If the cone playing 440 Hz has any H8 and H9 distortion at all it will be very audible because it will play in fact at...
There the woofer has a cone breakup bump at 4.2 kHz and I want to cross it at 3000. It might be still too high but the tweeter has resonance at 1 kHz so I want to fade before approaching 1 kHz. It will probably be a good compromise.
Probably. I'm not usually big on filters at 3KHz since it's slap bang at point of human hearing's maximum (amplitude) sensitivty, but with coax units you're working with a point source, which at least removes potential issues from separation in the x & y axis. Depends exactly how you're planning on crossing of course.
For whatever it's worth, to my eyes the above looks like a lengthy way of saying that if high-fidelity to the recording is the goal for a particular system, then it's generally better to use drive units that have lower levels of non-linear distortion (harmonic, IM, FM etc.) than higher, once the relevant filters etc. are in place. Which shows the value of individual HD plots of a drive unit (rather than THD, which as a lumped total is fairly useless), especially out to, say, HD5. So at the end of the day (and accepting there are a number of caveats), if you have a concern about a particular frequency or its harmonics, it's worth eyeballing the driver's linear distortion and its non-linear, especially the individual HD levels at those points, as this can help either pick drive units in general, or guide how you work with it in the filter.
For whatever it's worth, to my eyes the above looks like a lengthy way of saying that if high-fidelity to the recording is the goal for a particular system, then it's generally better to use drive units that have lower levels of non-linear distortion (harmonic, IM, FM etc.) than higher, once the relevant filters etc. are in place. Which shows the value of individual HD plots of a drive unit (rather than THD, which as a lumped total is fairly useless), especially out to, say, HD5. So at the end of the day (and accepting there are a number of caveats), if you have a concern about a particular frequency or its harmonics, it's worth eyeballing the driver's linear distortion and its non-linear, especially the individual HD levels at those points, as this can help either pick drive units in general, or guide how you work with it in the filter.
@Scottmoose indeed in the instruments from #165 the source H2 is always so powerful that a distortion won't be audible. H3+ matters, and then each distortion multiplied by its rank can be used to find where they converge, and that's what this driver cannot play.
Also the original harmonics get distortion harmonics, not just the fundamental. Original sounds like piano and 99% have more than on tone at once, and it all gets distorted. It's easy to calculate with sound of 100Hz fundamental, whose harmonics are 200, 300, 400Hz and so on. Now, if the sound gets distorted, all these get their own harmonic series. 2nd order harmonic for any harmonic in the original sound lands on even harmonic on the original series but take 3rd harmonic of third for exanple and it's 9th on the original. It's still at the same series and related though. Most of the tine these get lower in level with increasing frequency, but as it's easy to expand this thought experiment to x amount of of original sound sources on a recording with their harmonic series, and it's easy to see it all converts to just noise eventually.
It's not just harmonics though, low frequency sounds would make greatest excursion and driver parameters modulate with it, which affects the whole bandwidth of the driver, IMD. What ever is the low frequency sound source played back modulates it's own harmonics but all sounds that output from the same driver, related or not, which converts to a lot of noise.
Whats more audible than something else is something I don't kbow. Anyway, fun stuff to wonder. My biggest problem right now is very loud fan on an unit I'm testing, noise is the enemy, steals the attention. In general I think any sound of equipment is going to fight for attention, taking away from music , resonances being the worst 🙂
It's not just harmonics though, low frequency sounds would make greatest excursion and driver parameters modulate with it, which affects the whole bandwidth of the driver, IMD. What ever is the low frequency sound source played back modulates it's own harmonics but all sounds that output from the same driver, related or not, which converts to a lot of noise.
Whats more audible than something else is something I don't kbow. Anyway, fun stuff to wonder. My biggest problem right now is very loud fan on an unit I'm testing, noise is the enemy, steals the attention. In general I think any sound of equipment is going to fight for attention, taking away from music , resonances being the worst 🙂
I'm sorry, but it doesn't work like that. When you have a THD number, like your 1%, that is the "Total" (hence the "T" in THD) amount of error in the reproduced signal. That's across all harmonics.
And harmonics of the harmonics, that's a second order effect on something that isn't that audible to begin with.
@gedlee It's not as good and simple as you describe.
What you are saying is that the THD measurement is the HD of all orders from the primary signal, and then the HD from every harmonic the distortion generated. This is NOT what we have in the real life. Read: https://www.robvanderhaar.nl/en-us/sound-analysis-extended/
This is not distortion, this is the actual sound produced by the violin and that's what we want to hear. The main C4 (262 Hz) produces about 20 more Furier harmonics to create the violin signal.
Then every single of those harmonics played simultaneously generate distortion in the speaker. So the total distortion is:
THD = TDH(main) + C2*THD(H2(main)) + C3*TDH(H3(main)) + ... or
TDH = HD(H2(main)) + HD(H3(H2(main))) + ... + HD(H2(H2)) + HD(H3(H2)) + ... and so one for all permutations.
To measure the distortion you need to generate 20 signals, not one, and then sum up 20 distortions from every chain of distortions. So it's not exactly simple THD * 20, but probably at least 10 times measured THD from the REW sweep. Just do 10x of every distortion you measure by the sweep and you are there.
What you are saying is that the THD measurement is the HD of all orders from the primary signal, and then the HD from every harmonic the distortion generated. This is NOT what we have in the real life. Read: https://www.robvanderhaar.nl/en-us/sound-analysis-extended/
This is not distortion, this is the actual sound produced by the violin and that's what we want to hear. The main C4 (262 Hz) produces about 20 more Furier harmonics to create the violin signal.
Then every single of those harmonics played simultaneously generate distortion in the speaker. So the total distortion is:
THD = TDH(main) + C2*THD(H2(main)) + C3*TDH(H3(main)) + ... or
TDH = HD(H2(main)) + HD(H3(H2(main))) + ... + HD(H2(H2)) + HD(H3(H2)) + ... and so one for all permutations.
To measure the distortion you need to generate 20 signals, not one, and then sum up 20 distortions from every chain of distortions. So it's not exactly simple THD * 20, but probably at least 10 times measured THD from the REW sweep. Just do 10x of every distortion you measure by the sweep and you are there.
@gedlee I guess you can still do this with a REW sweep. So we need 20 harmonics and their distortions starting at 262 Hz.
Look at the chart, grab the THD at 262 Hz, then add THD at 524 Hz, then add THD at 786 Hz and so one 20 times. It will be more or less in the ballpark of 20 signals played simultaneously.
Look at the chart, grab the THD at 262 Hz, then add THD at 524 Hz, then add THD at 786 Hz and so one 20 times. It will be more or less in the ballpark of 20 signals played simultaneously.
There is an easily observable fact that different instruments are distorted at different degree, because of everything discussed above.
For example, when I slap on my speaker a crossover that doesn't sufficiently cut the cone breakup area, I still get perfectly clean sound with electric and maybe even acoustic guitars, voice and drums, so regular rock music is already good to go.
But when I listen for trumpets, saxophones, piano and symphonic orchestra in general they all sound horribly with easily observable defects. So that's what I use for the listening tests.
For example, when I slap on my speaker a crossover that doesn't sufficiently cut the cone breakup area, I still get perfectly clean sound with electric and maybe even acoustic guitars, voice and drums, so regular rock music is already good to go.
But when I listen for trumpets, saxophones, piano and symphonic orchestra in general they all sound horribly with easily observable defects. So that's what I use for the listening tests.
Sometimes I wish we could revert to the HP and B&K equipment of the old. This is going nowhere to me and as Earl Geddes already questioned (well, stated actually) what is the real life net worth of this pixel peeping? You don’t really believe you actually can notice 0,1% of 1% distortion of any live musical instrument, do you? What you might hear is other stuff or imagination, as we all suffer from that last phenomenon, don’t ever, ever rule it out.
Nothing as far as I can see, other than the fairly unremarkable notion that if fidelity to the recording happens to be the goal, then ideally you'll want to use drive units with low levels of HD, IM, FM etc. distortion over their intended operating range (once filtering is accounted for). And as Earl, Floyd Toole & others have pointed out in various ways & places, the audibility threshold for HD in particular is more forgiving than is sometimes believed. Not to say it isn't a factor, but it's possible to get a bit over-focused.
Last edited:
I want to apologize for spamming this thread with trivial stuff because I am a complete noob. I only started my first speaker building project in November and I am still only halfway in, after 100s hours of work and listening tests. I've got fascinated over how much difference in the sound some minor modifications of the box and crossover do. I am already at the second cabinet variant and 20th crossover, each having days of listening. I especially found surprising how much of the distortion difference I can get by making just a small change in XO. What stumbles me still that I can achieve either low distortion or flat response but not both. Sometimes the XO with absolutely flat curve gives me totally trashy sound with orchestra, grand piano, and a lot of wind instruments.
To me, apologies would be misplaced. And welcome to the art of speaker building 😉 your ears will lead you to endless modifications. But hey, that's life here...
+1. Apologies (very definitely) not required; we all work through similar things, so please -keep going! 🙂 👍
A new for me info I caught from GR research podcast - he claimed that a tweeter will always be far superior in quality right above the cross frequency than a woofer right below cross, at the condition that tweeter is still safely above its resonance frequency.
I am not sure this is just because a tweeter is always better. I can accept another explanation that the woofer is playing all the derived harmonic distortions generated by other sounds playing at 1/2, 1/3, etc down from the cross frequency, while the tweeter has a fresh start without this baggage.
It made me rethinking my current crossover that I originally made just 1 kHz lower than cone breakup and move according to my tweeter resonance plus 1.5 kHz for the tweeter down slope. I will need to find somewhere how much the tweeter is required to be damped at resonance, like -15 Db or something else.
I am not sure this is just because a tweeter is always better. I can accept another explanation that the woofer is playing all the derived harmonic distortions generated by other sounds playing at 1/2, 1/3, etc down from the cross frequency, while the tweeter has a fresh start without this baggage.
It made me rethinking my current crossover that I originally made just 1 kHz lower than cone breakup and move according to my tweeter resonance plus 1.5 kHz for the tweeter down slope. I will need to find somewhere how much the tweeter is required to be damped at resonance, like -15 Db or something else.