Hi guys,
Someone to explain me in simple words how "sharp peaks at a particular frequency will always cause a peak in the 3rd harmonic distortion plot roughly 2 octaves below" (quote from Zaph) as harmonics do always appear higher in frequencies ?!...
Thanks 🙂
Zaph inside:
Someone to explain me in simple words how "sharp peaks at a particular frequency will always cause a peak in the 3rd harmonic distortion plot roughly 2 octaves below" (quote from Zaph) as harmonics do always appear higher in frequencies ?!...
Thanks 🙂
Zaph inside:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Simple. At any frequency 'f', a driver having 3rd order HD will actually be producing sound at 3x f, or 4x f for 4th order, or 5x f for 5th order, etc. If the driver is more sensitive at 3x f, 4x f, 5x f, etc. this will result in higher apparent HD down low... but in practice, because peaks will be notched out, the distortion will not be as bad as it seems for an unfiltered driver.
I'm in the camp that distortion tests should be done on drivers with equalized on-axis responses, but this requires quite a bit of work.
I'm in the camp that distortion tests should be done on drivers with equalized on-axis responses, but this requires quite a bit of work.
It is real if you run the driver without any filtering. Most designers take out the cone resonance peak before doing anything else with it.
Oh, and 2nd and 3rd order HD is not nearly as harmful as 4th, 5th, and higher, except in sub applications where any HD will serve to localize the sub.
Thanks 454Casull for the explanation.
<<<< except in sub applications where any HD will serve to localize the sub.
*** interesting...and obvious.
🙂
<<<< except in sub applications where any HD will serve to localize the sub.
*** interesting...and obvious.
🙂
“Someone to explain me in simple words how "sharp peaks at a particular frequency will always cause a peak in the 3rd harmonic distortion”
Sure.
The amplitude response curve assumes that the driving force is held constant. What one sees then is that there is a broad flat area where the system is an acceleration limited system. In this region, if you were to glue a lead ring to the voice coil, you would displace that flat part of the curve downward resulting in a lower LF corner and lower efficiency (but no change in the hf response). As one goes higher in frequency, one can see the level falling a bit as the cone is no longer able to fully act like a ridged piston and is less effective.
Higher up is cone “breakup” where cone is in mechanical resonance, the first peak generally being when the outer edge is moving farther than the center / voice coil.
With higher modes, part of the cone is going one way while other parts the opposite direction.
Above this region, the bulk of the cone is mechanically decoupled from the voice coil in addition to the inductive roll off.
The bottom line is that the peaks represent acoustic gain over the surrounding area’s and so if one has a mound of “excess gain” at say 1200Hz, then one also has the situation where the motor nonlinearity is also amplified accordingly at that frequency. Remember if the signal is at 500Hz, the third harmonic produced is at 1500Hz. So, if one has a motor linearly that would produce 5% distortion at a given F and level and a gain of say 20dB at some frequency, then one knows that at frequencies of half, 1/3. 1/4 1/5, 1/6 etc of that gain, the natural harmonic content from the motor will be amplified that same 20dB or be 10X higher or 50%. I used 20 dB because it’s easy to scale being 100X in power and 10% in percentage distortion.
It might seem that that using a sharp crossover below the misbehavior would “fix” this and it does so far as the amplitude response BUT the increase in distortion remains because it is at those lower frequencies are where the distortion is produced and the mechanical system still has that gain..
As most of a loudspeakers problems get louder faster than the input signal loudness, reducing a speakers distortion for a given level often results in a higher usable loudness. At work, we make speakers for large spaces and there all of the loudness and room acoustic related problems are what shapes modern sound system design. The direction we went is to use a point source CD horn driven at multiple points based on the frequency / size of the wavelength within the horn. For the ranges below the HF section, each range is coupled to the horn through an acoustic low pass filter.
This is comprised on the air mass in the port leading into the horn and the air volume trapped between the cone and baffle board. This reduces the distortion a driver produces by attenuating the output at a second order slope above the desired upper frequency limit.
In the case of the Synergy horn an additional distortion reduction mechanism is that well above the size governed operating range, the distortion components cannot drive the horn with “gain” and instead are attenuated by a series of cancellation notches which begin when the side mounted driver is ¼ wl from the closed end.. Fwiw, It was exploration of this cancellation notch and the thought “what if I substituted a source of the opposite phase? ..which I had in the back side of the same driver” that lead to the computer model, that lead to the Tapped Horn .
The simplest band pass subwoofer alignment also has a low pass effect, it has an acoustic low pass effect and would radiate less distortion for a given SPL than the same driver as a direct radiator.
The driver’s motor linearity appears to be the dominant “thing” governing distortion over a broad band.
It is the balance between low cost, high strength to weight and internal damping which limits the Q of breakup features, that makes paper pulp popular even in 2011.
Headroom is your friend.
Best,
Tom Danley
Danley Sound Labs
Sure.
The amplitude response curve assumes that the driving force is held constant. What one sees then is that there is a broad flat area where the system is an acceleration limited system. In this region, if you were to glue a lead ring to the voice coil, you would displace that flat part of the curve downward resulting in a lower LF corner and lower efficiency (but no change in the hf response). As one goes higher in frequency, one can see the level falling a bit as the cone is no longer able to fully act like a ridged piston and is less effective.
Higher up is cone “breakup” where cone is in mechanical resonance, the first peak generally being when the outer edge is moving farther than the center / voice coil.
With higher modes, part of the cone is going one way while other parts the opposite direction.
Above this region, the bulk of the cone is mechanically decoupled from the voice coil in addition to the inductive roll off.
The bottom line is that the peaks represent acoustic gain over the surrounding area’s and so if one has a mound of “excess gain” at say 1200Hz, then one also has the situation where the motor nonlinearity is also amplified accordingly at that frequency. Remember if the signal is at 500Hz, the third harmonic produced is at 1500Hz. So, if one has a motor linearly that would produce 5% distortion at a given F and level and a gain of say 20dB at some frequency, then one knows that at frequencies of half, 1/3. 1/4 1/5, 1/6 etc of that gain, the natural harmonic content from the motor will be amplified that same 20dB or be 10X higher or 50%. I used 20 dB because it’s easy to scale being 100X in power and 10% in percentage distortion.
It might seem that that using a sharp crossover below the misbehavior would “fix” this and it does so far as the amplitude response BUT the increase in distortion remains because it is at those lower frequencies are where the distortion is produced and the mechanical system still has that gain..
As most of a loudspeakers problems get louder faster than the input signal loudness, reducing a speakers distortion for a given level often results in a higher usable loudness. At work, we make speakers for large spaces and there all of the loudness and room acoustic related problems are what shapes modern sound system design. The direction we went is to use a point source CD horn driven at multiple points based on the frequency / size of the wavelength within the horn. For the ranges below the HF section, each range is coupled to the horn through an acoustic low pass filter.
This is comprised on the air mass in the port leading into the horn and the air volume trapped between the cone and baffle board. This reduces the distortion a driver produces by attenuating the output at a second order slope above the desired upper frequency limit.
In the case of the Synergy horn an additional distortion reduction mechanism is that well above the size governed operating range, the distortion components cannot drive the horn with “gain” and instead are attenuated by a series of cancellation notches which begin when the side mounted driver is ¼ wl from the closed end.. Fwiw, It was exploration of this cancellation notch and the thought “what if I substituted a source of the opposite phase? ..which I had in the back side of the same driver” that lead to the computer model, that lead to the Tapped Horn .
The simplest band pass subwoofer alignment also has a low pass effect, it has an acoustic low pass effect and would radiate less distortion for a given SPL than the same driver as a direct radiator.
The driver’s motor linearity appears to be the dominant “thing” governing distortion over a broad band.
It is the balance between low cost, high strength to weight and internal damping which limits the Q of breakup features, that makes paper pulp popular even in 2011.
Headroom is your friend.
Best,
Tom Danley
Danley Sound Labs
Thanks Tom
Hi Tom.
As always your posts are just great! Hidden gems as well as the main point.
Thanks for taking the time to add to the forum. There are plenty of very clever doctors, dentists and scientists from all walks of life on this site, many are even have commercial audio interests and a few might actually have sold a few $ worth of their ideas..
But your CV and PROVEN theories in the $millions count for so much more...
I trust the opinion of 1 expert over a 1,000 well-meaning amateur audio professional doctors any day.
Humbly in awe
Derek.
Hi Tom.
As always your posts are just great! Hidden gems as well as the main point.
Thanks for taking the time to add to the forum. There are plenty of very clever doctors, dentists and scientists from all walks of life on this site, many are even have commercial audio interests and a few might actually have sold a few $ worth of their ideas..
But your CV and PROVEN theories in the $millions count for so much more...
I trust the opinion of 1 expert over a 1,000 well-meaning amateur audio professional doctors any day.
Humbly in awe
Derek.
Thanks a lot Tom for your input!
Does the distortion also occur at these lower frequencies?
Crazhub asks;
“Does the distortion also occur at these lower frequencies?”
Yes. Picture that when the speaker radiator moves in and out in a perfect sine wave motion, it only produces sound at that one frequency. If you did anything like touch it with your finger, you would be changing what it did in one place in it’s motion vs another and so you have added a non linearity.
The motion would then be not a pure sine wave and contains energy at other frequencies as well as the real signal.. In the real world, nothing is perfect and say with the loudspeaker, one has much greater non-linearity than with the electronic part. Non-linearity is anything / everything that causes what the system does to be different that what it is told to do. Non linearity causes “free sound” to be produced, this is energy that is not in the drive signal.
So, if one drove a speaker with a “pure tone” at 300Hz, the driver nonlinearity would be adding energy at 600Hz (2nd harmonic), 900Hz (3rd harmonic), 1200Hz (4th harmonic) and so on. It would be these harmonics that if they fell where one of those mounds of excess gain were, that would be amplified by that amount.
In Radio work, THD or Total Harmonic Distortion made perfect sense because one doesn’t want energy outside your assigned frequency and harmonics are signals produced by the nonlinearity which end up at even and odd multiples of the fundamental (intended) frequency.
In audio, back when some amplifiers had gross distortion and all of them had more similar properties (as they were all tube based) THD still made some sense.
Later work into audio compression revealed more about how we hear.
Read about perceptual masking for example, it governs if we can hear a particular sound in the presences of another sound, this plays a strong role in our hearing but one we are unaware of as it is all we know.
Psychoacoustics - Wikipedia, the free encyclopedia
The bottom line is that best case against a silent background one cannot hear a 2nd harmonic until it’s quite large but as the harmonic number rises, harmonics become easier to hear. The result is that THD in this case is not very useful as it doesn’t represent how we hear or how that changes with frequency.
Not only that but music is harmonic in structure so the even harmonic part of the spectrum doesn’t necessarily sound out of place. During the recording process since the Beatles, even order distortion is often added to make a “richer” sound. On the other hand, natural sounds are usually not harmonic and so then adding either even or odd order distortion can change the sound.
Overkill Audio said;
“Thanks for taking the time to add to the forum etc.”
You welcome but don’t get the wrong idea, anyone can have a good idea.
I write these things in part because my background is a DIY speaker builder and have loved audio stuff and electronics since childhood.
It is places like this where the seeds of tomorrows audio are sprouting / growing, the place the people who will take jobs like mine when I retire come from.
Best,
Tom Danley
Mike has posted a humorous video of a demo he did Tuesday;
The 3 speakers are the small black blob under the scoreboard and to the right. Try it with headphones.
Danley Sound Labs, Inc.'s Videos | Facebook
“Does the distortion also occur at these lower frequencies?”
Yes. Picture that when the speaker radiator moves in and out in a perfect sine wave motion, it only produces sound at that one frequency. If you did anything like touch it with your finger, you would be changing what it did in one place in it’s motion vs another and so you have added a non linearity.
The motion would then be not a pure sine wave and contains energy at other frequencies as well as the real signal.. In the real world, nothing is perfect and say with the loudspeaker, one has much greater non-linearity than with the electronic part. Non-linearity is anything / everything that causes what the system does to be different that what it is told to do. Non linearity causes “free sound” to be produced, this is energy that is not in the drive signal.
So, if one drove a speaker with a “pure tone” at 300Hz, the driver nonlinearity would be adding energy at 600Hz (2nd harmonic), 900Hz (3rd harmonic), 1200Hz (4th harmonic) and so on. It would be these harmonics that if they fell where one of those mounds of excess gain were, that would be amplified by that amount.
In Radio work, THD or Total Harmonic Distortion made perfect sense because one doesn’t want energy outside your assigned frequency and harmonics are signals produced by the nonlinearity which end up at even and odd multiples of the fundamental (intended) frequency.
In audio, back when some amplifiers had gross distortion and all of them had more similar properties (as they were all tube based) THD still made some sense.
Later work into audio compression revealed more about how we hear.
Read about perceptual masking for example, it governs if we can hear a particular sound in the presences of another sound, this plays a strong role in our hearing but one we are unaware of as it is all we know.
Psychoacoustics - Wikipedia, the free encyclopedia
The bottom line is that best case against a silent background one cannot hear a 2nd harmonic until it’s quite large but as the harmonic number rises, harmonics become easier to hear. The result is that THD in this case is not very useful as it doesn’t represent how we hear or how that changes with frequency.
Not only that but music is harmonic in structure so the even harmonic part of the spectrum doesn’t necessarily sound out of place. During the recording process since the Beatles, even order distortion is often added to make a “richer” sound. On the other hand, natural sounds are usually not harmonic and so then adding either even or odd order distortion can change the sound.
Overkill Audio said;
“Thanks for taking the time to add to the forum etc.”
You welcome but don’t get the wrong idea, anyone can have a good idea.
I write these things in part because my background is a DIY speaker builder and have loved audio stuff and electronics since childhood.
It is places like this where the seeds of tomorrows audio are sprouting / growing, the place the people who will take jobs like mine when I retire come from.
Best,
Tom Danley
Mike has posted a humorous video of a demo he did Tuesday;
The 3 speakers are the small black blob under the scoreboard and to the right. Try it with headphones.
Danley Sound Labs, Inc.'s Videos | Facebook
So, whatever you do with electronics, including rolling off before the breakup and/or notching it out, the effect on distortion remains? As the driver itself has additional output at a frequency that will at some points be a multiple of a frequency it is driven with. Makes sense to me, I'd always thought that must be the case. So, a hard cone might always exhibit some extra distortion, even if it is only at a single frequency?
Impressive coverage of the stadium in that vid btw!
Impressive coverage of the stadium in that vid btw!
Yeah, cause no matter what you do, the cone is going to resonate at some frequency(ies) and the only way to really get rid of it is to cross over low enough that the harmonics don't excite the resonance by still having the fundamental, at least that's how I understand it.
still something sticking in my head
Tom please:
<<<<< Remember if the signal is at 500Hz, the third harmonic produced is at 1500Hz. So, if one has a motor linearly that would produce 5% distortion at a given F and level and a gain of say 20dB at some frequency, then one knows that at frequencies of half, 1/3. 1/4 1/5, 1/6 etc of that gain, the natural harmonic content from the motor will be amplified that same 20dB or be 10X higher or 50%.>>>>>
I feel stupid! Please what is wrong in my thought?:
i.e. 1kHz signal. Cone break-up at 5kHz thus 5kHz level is +20db.
Second harmonic of 5kHz is 10kHz thus level of 10kHz distortion is 10x
So why "half of that gain" and not "double of that gain"?
Thanks
Tom please:
<<<<< Remember if the signal is at 500Hz, the third harmonic produced is at 1500Hz. So, if one has a motor linearly that would produce 5% distortion at a given F and level and a gain of say 20dB at some frequency, then one knows that at frequencies of half, 1/3. 1/4 1/5, 1/6 etc of that gain, the natural harmonic content from the motor will be amplified that same 20dB or be 10X higher or 50%.>>>>>
I feel stupid! Please what is wrong in my thought?:
i.e. 1kHz signal. Cone break-up at 5kHz thus 5kHz level is +20db.
Second harmonic of 5kHz is 10kHz thus level of 10kHz distortion is 10x
So why "half of that gain" and not "double of that gain"?
Thanks
With real music signal the bigger problem is intermodulation distortion which unavoidingly comes with harmonic distortion, and with speakers it comes in heavy doses.
So you end up with all the mess of sum and difference tone combinations (with a lot of time smearing as well) NOT being attenuated at all even when you take out the driver way below the trouble region, with steep slopes. Often a dirty, low resolution sound in the lower treble comes from that (mid-)woofer hash and not from a suspected tweeter or diffraction issues etc.
If you have massive boost in the reakup region, you need a very low distortion motor and suspension and design headroom to really make use of the perfect pistonic range.
- Klaus
So you end up with all the mess of sum and difference tone combinations (with a lot of time smearing as well) NOT being attenuated at all even when you take out the driver way below the trouble region, with steep slopes. Often a dirty, low resolution sound in the lower treble comes from that (mid-)woofer hash and not from a suspected tweeter or diffraction issues etc.
If you have massive boost in the reakup region, you need a very low distortion motor and suspension and design headroom to really make use of the perfect pistonic range.
- Klaus
Well is your signal 1 kHz or 5 kHz? Because 2HD of a 5 kHz will be heard as 10 kHz, whereas 2HD of 1 kHz signal will be heard as 2 kHz.
The other question - he says 50% as in resulting harmonic will be 50% instead of 5 %.
My mistake, sorry; I recall:
i.e. 0db linear signal up to 4kHz. 20db cone break-up at 5kHz.
Second harmonic of 5kHz is 10kHz thus level of 10kHz distortion should be increased. Correct?
What I don't understand is why distortion does (also?) appear at half or 1/3 or 1/4 the fr of the gain i.e. 2.5kHz, 1.66kHz, 1.25kHz ??? I could also ask this way: are 2.5, 1.66, 1.25kHz the actual distorded frequencies or would I just hear the 5kHz distortion because I didn't kill these lower frequencies?
+20db distortion / 5% distortion = 50% distortion
Hi
I am not sure I was all that clear.
What I mean is that if one had a 20dB mound of mechanical gain at 3000Hz, above the piston region looking like the example, then;
One can use a really sharp low pass filter at 2KHz and the lump of gain goes away in the response curve and problem solved?.
The wart remains however in the distortion curve because when the speaker produces a pure 1KHz tone, it invariably produces some distortion at it’s 3rd harmonic which is at 3KHz and so that amplified 20 dB or 100X by that mechanical gain.
That 20dB of gain multiplies the distortion figure by 10X (as distortion is a Voltage signal at this point and not SPL or power ).
At each sub multiple frequency of the “lump of gain”, you usually see the related harmonic amplified by that lump of gain as well.
Go to an aluminum cone like for bass guitar and you have drivers with a large “lump of gain” related to the cone break up. This gives them a sound that bites through the noise like spray auto undercoating downs a cloud of wasps. The latter being pretty good in an emergency. While a bit messy, there is no doubt, they can’t fly with even a small glob of tar on them and for the musical instrument speaker, whatever works to create the desired sound.
Best,
Tom
I am not sure I was all that clear.
What I mean is that if one had a 20dB mound of mechanical gain at 3000Hz, above the piston region looking like the example, then;
One can use a really sharp low pass filter at 2KHz and the lump of gain goes away in the response curve and problem solved?.
The wart remains however in the distortion curve because when the speaker produces a pure 1KHz tone, it invariably produces some distortion at it’s 3rd harmonic which is at 3KHz and so that amplified 20 dB or 100X by that mechanical gain.
That 20dB of gain multiplies the distortion figure by 10X (as distortion is a Voltage signal at this point and not SPL or power ).
At each sub multiple frequency of the “lump of gain”, you usually see the related harmonic amplified by that lump of gain as well.
Go to an aluminum cone like for bass guitar and you have drivers with a large “lump of gain” related to the cone break up. This gives them a sound that bites through the noise like spray auto undercoating downs a cloud of wasps. The latter being pretty good in an emergency. While a bit messy, there is no doubt, they can’t fly with even a small glob of tar on them and for the musical instrument speaker, whatever works to create the desired sound.
Best,
Tom
OK, so if on a distortion plot I see a HD3 peak situated at 1kHz, it's not the 1kHz which is distorded but the 3kHz. Correct?
- Status
- Not open for further replies.