WOW! I never thought one simple question would raise so many more... I never new speakers can have so much distortion. Another clue how little I know about audio reproduction.
Maybe I need to revisit a website I found few month ago. It has a lot of articles about amplifiers, speaker etc.
Maybe I need to revisit a website I found few month ago. It has a lot of articles about amplifiers, speaker etc.
Anti THD guys Always try to make the number meaningless by downplaying relevance with respect to amps, and to bloat it to their advantage with respect to speaker systems. Generally folks that like their amps to be sound effects boxes too, especially if they make the sound warmer. They usually are fine playing 64- or 128Kb MP3s as well, claiming no audible difference to lossless formats.
Last edited:
Harmonic distortion really isn't important at all. Intermodulation is. Mechansims which produce 2nd order harmonic distortion tend to produce 2nd order intermods. Those which produce 3rd harmonic produce 3rd order intermods. Levels of harmonic and IM distortion are usually comparable, so lowering one lowers the other.
2nd order intermod is almost indistinguishable from 2nd harmonic when the frequencies which produce them are close together. It is AM or doppler distortion when the frequencies are far apart. Both results sort of sound "natural" so you don't mind them very much. 3rd order intermods are IN YOUR FACE, as are all higher odd order intermods. This is because they fall close to the frequencies which produced them when those two frequencies are close together, and can put hash at totally unrelated frequencies when they are far apart. Your brain interprets this as noise. This makes the "staticky" sound often associated with severe crossover distortion and/or clipping.
2nd order intermod is almost indistinguishable from 2nd harmonic when the frequencies which produce them are close together. It is AM or doppler distortion when the frequencies are far apart. Both results sort of sound "natural" so you don't mind them very much. 3rd order intermods are IN YOUR FACE, as are all higher odd order intermods. This is because they fall close to the frequencies which produced them when those two frequencies are close together, and can put hash at totally unrelated frequencies when they are far apart. Your brain interprets this as noise. This makes the "staticky" sound often associated with severe crossover distortion and/or clipping.
So basically, THD is indicative of IMD as well, and as such puts it at an equal importance level to IM distortion, contrary to the opening of your paragraph =)
You can use a THD number to guage how much IM an amp wil have, even if the harmonics themselves are irrelevant. An amp will have low IM3 at/near 1KHz if the THD1k is good, and will have low IM3 at/near 10k if the THD10k is good. If the THD10k is poor, you won't hear the 22kHz produced by an 11k tone. But you will hear the 10.5 and 12k produced by an 11k and an 11.5k at the same time. If the THD10k is lower, those levels will be lower and bother you a whole lot less.
And since it is the higher frequency spectrum that describes the spatial information and separation of music content, one can imagine that THD20K becomes a sensible indication to how 'clear' and 'transparanent' an amp can sound, given high quality speakers (mids / tweeters). NOTE: As in, if THD20K is good, THD10K will be a factor X better.
Edit: But yes, how much is enough? I'd say, if it stays below 0.001% THD20K under all circumstances, reactive loads and all. Then you at least can be assured, it should not be the amp that's causing any contribution of sound on its own. But that's my opinion. Chasing any lower THD is more like a sports. One I happen to love =)
Edit: But yes, how much is enough? I'd say, if it stays below 0.001% THD20K under all circumstances, reactive loads and all. Then you at least can be assured, it should not be the amp that's causing any contribution of sound on its own. But that's my opinion. Chasing any lower THD is more like a sports. One I happen to love =)
Last edited:
But doesn't the speaker multiply the incoming distortion from further up the chain? If so, doesn't it become paramount to reduce as much as possible all contributing sources of THD?
Also, is THD necessarily spread evenly across the frequency spectrum? Theoretically can say, .02% THD be confined to a range of 0-1000hz. in one amp and predominantly in the range of say, 5000-12000hz. in another? All other frequencies with immeasurable distortion in both?
THD generally rises as frequency increases. Sometimes you'll see a THD increase in the low frequencies the lower you get, but this is because of DC blocking caps / DC servo's. This kind of distortion isn't really harmonic distortion; it's called "Wave form envelope distortion".
Generally, it depends on the distortion spectrum shape. Soft clipping low feedback tube amps usually have a much more natural distortion spectrum, so several percent doesn't seem to bother people. If it's below 1%, it's probably not the weak link of your system. How we measure distortion is controversial, and a whole different argument. Manufactures want to be able to list a very tiny number, so they measure it that way. I'd be more worried about circuit stability myself. Does it really have a good phase margin? Is there a passive Rf filter at the input?
But does this rise occur in a more or less linear fashion depending on the particular amp in question or is the curve always the same no matter the amp? I ask because I'm wondering if this is a factor in the sound variations from amp to amp and is not just in my head. Not necessarily tonally, but in terms of sibilance/transient response, timbre...
I am thinking the TDA7293 sounds better the the 3886.
As for distortion, harmonic distortion of speakers is usually orders of magnitude higher than THD of electronics. The problem in electronics is production of non-harmonically related tone. This subject is treated like a highly contagious plague with no one willing to even touch it let alone discuss it with any honesty or facts. This type of tone generation is the main reason electronics sound different and there is little correlation between numbers and sound.
As for distortion, harmonic distortion of speakers is usually orders of magnitude higher than THD of electronics. The problem in electronics is production of non-harmonically related tone. This subject is treated like a highly contagious plague with no one willing to even touch it let alone discuss it with any honesty or facts. This type of tone generation is the main reason electronics sound different and there is little correlation between numbers and sound.
Update -- quit a few newer speaker systems are better than this. Worth searching them out. The newer Quad electrostatics have been measured at or under 0.1% since they came out. And extreamly flat response. Those Quads let you hear even a small mount of additional distortion.
THx-RNMarsh
Last edited:
Check the chip datasheet and see the distortion chart. You will see that distortion will climb with power increase. Also the distortion will climb with frequency increase. So for the spec, pay attention to the frequency and power. It could be low if measured at 1kH/1W. It could be high if measured at more or less rated power (e.g. 40W) and at full bandwidth (~20kHz).
Check again your datasheet. I think 0.02% is too good to be true, even for 1W/1kHz. In my golden ear experience 😀 4152 sounds better than 4151 (same specs I believe).
partyface86 said:
partyface86 said:
partyface86 said:
Now it seems you need to give more respect to the golden ears 😀 At least no golden ears have as low as 10 years experience or less.
In a DBT may be it is difficult (even for the golden ears) to differentiate a few % THD but I believe that the human's sub conscious mind can feel it.
Problem is, THD is not the only variable, so you have to balance THD with other properties.
The rise in distortion with frequency comes mostly from the fact that audio amplifiers use global feedback, which requires some form of frequency compensation to prevent the whole system oscillating.
Semiconductors don't have infinite bandwidth so there will inevitably be phase shift caused at high frequency.
The compensation needed is such that the open loop gain is attenuated to unity before the phase reaches 180degrees. If the output has gain>1 and phase>180degrees there will be positive feedback applied at the input and oscillation occurs.
Most amplifiers will have the open loop gain already rolling off by 1KHz or so, so distortion will steadily rise from 1KHz up. As open loop gain decreases, distortion increases.
Two ways of improving this would be to use faster transistors so the compensation can be pushed up in frequency (=less open loop attenuation at low frequencies), or using a multi-pole compensation scheme whereby the attenuation increases much more steeply with frequency so less attenuation occurs at lower frequencies.
Single pole compensation will roll off open loop gain at 6dB/octave, two-pole compensation will roll off loop gain at 12dB/octave. You would therefore expect amps with single pole compensation would show a similar trend in rising distortion with frequency, while 2-pole compensated amps would show a steeper rise in distortion with frequency, but starting at a higher frequency. The fact that all things the same, both single-pole and 2-pole schemes would require OL-gain to hit unity at the same frequency which is far higher than audio frequencies means that amps with 2-pole compensation will have lower distortion through the entire audio bandwidth (up to 20KHz) than the same amp with single pole compensation.
In regards to speaker distortion, well designed speaker drivers produce almost exclusively 2nd and 3rd order distortion and not much high-order distortion. Even order distortion >2nd order is usually much lower than odd order distortion (4th order is usually non-existant compared to 5th order, same with 6th with respect to 7th, etc). Inferior drivers can produce audible amounts of 5th order. You typically won't get audible amounts of 7th order and higher harmonics unless there is a major manufacturing issue (moving parts rubbing or buzzing), or the driver is being pushed past the linear part of it's excursion capability, or the small amount of 7th+ order distortion being produced by the motor is being amplified by a severe cone breakup node.
Amplifiers on the other hand can produce much higher amounts of high order (>5th order) usually attributed to crossover distortion in class-B amps. Compared to speakers the ratio of even order to odd order distortion is much closer because of mismatches between components which produce even order distortion. The magnitude of distortion products produced by a well designed amplifier is still far lower than a well designed speaker at the same power level.
To my ears, 2nd order harmonic isn't too offensive. 3rd order and higher sound bad. YMMV.
Last edited:
Sorry to be pedantic, but 'single number' THD specs are usually THD+N specs. That means that the amount of distortion could be zero, and the THD+N number is completely noise, or contains mostly distortion, or some of both.
That said, I do use a classic distortion analyzer, but I also look at an FFT of the residual in order to separate the actual distortion spurs from the uncorrelated noise in the device's output. A single 'THD' number is pretty much useless, especially if the measurement bandwidth is not specified, because it gives absolutely no indication of the nonlinearity of a circuit. For example, a perfectly linear circuit that has a finite noise floor will have a non-zero THD+N number.
One must separate the noise from the distortion!!
That said, I do use a classic distortion analyzer, but I also look at an FFT of the residual in order to separate the actual distortion spurs from the uncorrelated noise in the device's output. A single 'THD' number is pretty much useless, especially if the measurement bandwidth is not specified, because it gives absolutely no indication of the nonlinearity of a circuit. For example, a perfectly linear circuit that has a finite noise floor will have a non-zero THD+N number.
One must separate the noise from the distortion!!
There is more information on THD perception here:
Perception
Also look up the not so technical article "Weighting Up" by Keith Howard which does some comparison of different metrics:
http://www.gedlee.com/downloads/THD_.pdf
For quality of sound perception THD seems to be overrated.
Perception
Also look up the not so technical article "Weighting Up" by Keith Howard which does some comparison of different metrics:
http://www.gedlee.com/downloads/THD_.pdf
For quality of sound perception THD seems to be overrated.
So ýou guys are looking for low THD above 10kHz?
Interesting. There was a time in the 80s when a unit that did exactly the opposite - add extra distortion to the trebles on purpose - was so much sought-after that recording and broadcast studios had to rent not buy it. the Aphex Exciter. It "will remove the curtain between listener and speaker".. lol...now everybody is going for lowest THD...
Interesting. There was a time in the 80s when a unit that did exactly the opposite - add extra distortion to the trebles on purpose - was so much sought-after that recording and broadcast studios had to rent not buy it. the Aphex Exciter. It "will remove the curtain between listener and speaker".. lol...now everybody is going for lowest THD...
That was the time when everybody was so badly informed about a good sound 😛
It couldn't trick me now or then 😀
- Status
- Not open for further replies.