When a power rating is given at 1% or 0.7% THD, what is the non-clipping maximum?

[mike7877]

So we all know that 1% THD is -40dB. (If you didn't, you do now!)

A lot of reviews Amir does, the power vs THD+n chart will show distortion falling as power is increasing, and usually, somewhere around 10-20% of full power, this straight line reaching for the bottom, starts reaching for the side (I consider full power the point at which clipping occurs. So... while the line is falling in a straight line and power is increasing, harmonic distortion is entirely masked by the grass that his hiss. When it stops falling at this constant rate, I believe harmonic distortion is beginning to poke through the noise (ie. it's not entirely masked anymore). When the THD+n value starts quickly rising to the top of the chart, that is when clipping is happening. Distortion rises fast, but it's not extremely fast. The difference of power output at 1% and 10% distortion is actually pretty huge... So say the amplifier is normal-good, and right before it begins clipping at 122 watts into 8 ohms, THD+n is -100dB, or 0.001%. Since manufacturers give often give the power rating at 0.7% THD or 1% THD, I'd like to know roughly, by what amount, does power output increase from -100dB to -40dB? Basically I want to be able to take a rating like:

140W @ 1% THD+n

And multiply it by 0.94 to get where the amp [most likely] begins clipping.

For this to be generally applicable and on the conservative side, it's the power difference going from -80dB to -40db THD+n, when the increase is from clipping.
(sentence not worded the best, but y'know what I mean! (I hope lol) )

 

[DVDdoug]

You'd have to look at the curve, if you have one.

I'm not sure how useful the information would be... Since audio levels vary, and we usually have a volume control that we frequently mess-with it's hard to set it for exactly 122W (etc.). If you are that close, you probably should design-in some headroom.

Plus, your speakers probably have more than 1% distortion.

 

[mike7877]

You'd have to look at the curve, if you have one.

I'm not sure how useful the information would be... Since audio levels vary, and we usually have a volume control that we frequently mess-with it's hard to set it for exactly 122W (etc.). If you are that close, you probably should design-in some headroom.

Plus, your speakers probably have more than 1% distortion.

My speakers are extremely low distortion, actually. Less than 0.1% THD! Most of which is second harmonic... third never breaches higher than 0.03% when the speaker is performing within specifications (which I assume means until the underhung woofer leaves its gap, which is when the cone is moving more than 12mm peak to peak (8mm coil, 20mm gap). I think >100Hz.


The value I'm asking about shouldn't change much from amp to amp, unless the amp being measured behaves oddly when it clips. Or if its distortion is really close to 1% in the last few watts before clipping begins.


To illustrate better what I'm looking for: when clipping is causing 1% distortion and the volume is then increased until distortion reaches 10%, the power increase is almost exactly 19%. So if you get an amplifier that specs its power output 100W @ 10% THD, and you want a value that's more realistic, you can multiply the 100 watts: (1/1.19) * 100 to calculate, giving 84 watts at 1% distortion.

I'm looking for theprevious sentence, but from 0.01% to 1% (0.01% being the -80db THD+n I mentioned in the OP)

 

[MaxwellsEq]

It's an interesting Idea and if doable, would allow a conversion from some ludicrous marketing value to a real-world usable amount. Given the nonlinearity around the inflection and the upward slope, it's probably not realistic to have something that can convert from 10% THD, but converting from 1% to 0.1% might be viable.

I think you would need a "model power amplifier", perhaps one derived from multiple overlays of the graphs. You would also have to make assumptions about Noise Figures, since inherently noisy amplifiers would mess up the model.

 

[solderdude]

So we all know that 1% THD is -40dB. (If you didn't, you do now!)

A lot of reviews Amir does, the power vs THD+n chart will show distortion falling as power is increasing, and usually, somewhere around 10-20% of full power, this straight line reaching for the bottom, starts reaching for the side (I consider full power the point at which clipping occurs. So... while the line is falling in a straight line and power is increasing, harmonic distortion is entirely masked by the grass that his hiss. When it stops falling at this constant rate, I believe harmonic distortion is beginning to poke through the noise (ie. it's not entirely masked anymore). When the THD+n value starts quickly rising to the top of the chart, that is when clipping is happening. Distortion rises fast, but it's not extremely fast. The difference of power output at 1% and 10% distortion is actually pretty huge... So say the amplifier is normal-good, and right before it begins clipping at 122 watts into 8 ohms, THD+n is -100dB, or 0.001%. Since manufacturers give often give the power rating at 0.7% THD or 1% THD, I'd like to know roughly, by what amount, does power output increase from -100dB to -40dB? Basically I want to be able to take a rating like:

140W @ 1% THD+n

And multiply it by 0.94 to get where the amp [most likely] begins clipping.

For this to be generally applicable and on the conservative side, it's the power difference going from -80dB to -40db THD+n, when the increase is from clipping.
(sentence not worded the best, but y'know what I mean! (I hope lol) )

This is more complex than one might think.
The plots made by Amir are continuous sine. Music consists of short impulses and usually have a bit more headroom.
Also the 'knee' can be 'sharp' or 'slow' which is highly design (and power supply) dependent.
It is also impedance dependent and as most music has the largest excursions in the bass and most speakers have a higher impedance in the bass area the output voltage of a 4ohm speaker might be higher than a plot would suggest.

Also note that in music a short peak of 10% distortion might not be audible at all when the rest of the music is far below 0.1%.

Then we have the issue of 'recovery' after clipping. This can be audible and is highly design dependent and not shown in any graphs.

Also with many power amps (not having wide range regulated power supplies) the output power is also mains voltage dependent a little.

So... while measurements can say something a dB or 2 in max. output power (factor 1.6) in practice is not very audible and depends more on the dynamic/continuous power rating, clipping recovery behavior and speaker impedance than the 'where is the knee and how steep is the rise' issue.

Numbers are fine as an indicator and say something about the design and specs under certain lab conditions. Lab conditions may differ from usage in the home under non-lab conditions.

 

[mike7877]

It's an interesting Idea and if doable, would allow a conversion from some ludicrous marketing value to a real-world usable amount. Given the nonlinearity around the inflection and the upward slope, it's probably not realistic to have something that can convert from 10% THD, but converting from 1% to 0.1% might be viable.

I think you would need a "model power amplifier", perhaps one derived from multiple overlays of the graphs. You would also have to make assumptions about Noise Figures, since inherently noisy amplifiers would mess up the model.

Yeah, noisy amps could, or ones with high harmonic content all the time (or at least in the highest 20% of power (ie. watts 40 through 50 of a 50 watt amp))

Btw, if it can be done from 1% down to 0.01%, it can be done from 10% - you just need to do the 19% formula I have in post 3 to turn the 10 to 1% before going down to 0.01%. There would be a bit more error, but I think it'd be within a percent or two

 

[mike7877]

This is more complex than one might think.
The plots made by Amir are continuous sine. Music consists of short impulses and usually have a bit more headroom.
Also the 'knee' can be 'sharp' or 'slow' which is highly design (and power supply) dependent.
It is also impedance dependent and as most music has the largest excursions in the bass and most speakers have a higher impedance in the bass area the output voltage of a 4ohm speaker might be higher than a plot would suggest.

Also note that in music a short peak of 10% distortion might not be audible at all when the rest of the music is far below 0.1%.

Then we have the issue of 'recovery' after clipping. This can be audible and is highly design dependent and not shown in any graphs.

Also with many power amps (not having wide range regulated power supplies) the output power is also mains voltage dependent a little.

So... while measurements can say something a dB or 2 in max. output power (factor 1.6) in practice is not very audible and depends more on the dynamic/continuous power rating, clipping recovery behavior and speaker impedance than the 'where is the knee and how steep is the rise' issue.

Numbers are fine as an indicator and say something about the design and specs under certain lab conditions. Lab conditions may differ from usage in the home under non-lab conditions.

You've made some good points. At first I was thinking "yeah, but I just want the difference". But then I saw all these common contributing factors and realised it's probably a pretty pointless thing to do... Since going from 1% to 10% distortion (all from clipping) is 19% more power, let's pretend going from 0.01 to 1% is 1.88% more power. This 1.88 can be diminished by 25% if the amp is noisy. Then you have supply voltage, which is actually a huge factor. Since everything's designed for 120V and the power coming in is usually between 110 and 117V, at just 114V up to 10% of output power can be sacrificed! Of course this is only with non-regulated amplifiers that are clipping against the rails and not for other reasons, but that's still going to be more than half of all ab designs... (aside: I have an auto-transformer which can give anywhere from 0 to 144V from 120V, - up to 20A - I hooked it up to my Denon X3700H and gave it +10% voltage and its amp section turned into the X4700H! lol, I'm pretty sure the 4700's amp is identical, only the transformer in it has a few more turns for the output stage)

Back on track: I think I'm just going to call it 2%. Subtract 2% for good measure from the 1% or 0.7% THD figure.

It's really disingenuous of all these manufacturers to rate their stuff at 0.1% or 0.7% or 1%... The reason they do it is so that you can't compare the actual value which is always better and often 0.04 or 0.023 or 0.0088%. I think it's a collective silent agreement because the amplifiers they make vary so much and the value doesn't always line up with the cost of the units or position in their lineup. It should, at least to an extent, and a value of 0.08% minimum THD+n (ie. at max power) would actually be an unacceptable value for, say, a Denon X1700H, but because they never print the real numbers, they could release something like that and get away with it with 95% of their customers (the less discerning or with poorer speakers etc.)

 

[Sokel]

You've made some good points. At first I was thinking "yeah, but I just want the difference". But then I saw all these common contributing factors and realised it's probably a pretty pointless thing to do... Since going from 1% to 10% distortion (all from clipping) is 19% more power, let's pretend going from 0.01 to 1% is 1.88% more power. This 1.88 can be diminished by 25% if the amp is noisy. Then you have supply voltage, which is actually a huge factor. Since everything's designed for 120V and the power coming in is usually between 110 and 117V, at just 114V up to 10% of output power can be sacrificed! Of course this is only with non-regulated amplifiers that are clipping against the rails and not for other reasons, but that's still going to be more than half of all ab designs... (aside: I have an auto-transformer which can give anywhere from 0 to 144V from 120V, - up to 20A - I hooked it up to my Denon X3700H and gave it +10% voltage and its amp section turned into the X4700H! lol, I'm pretty sure the 4700's amp is identical, only the transformer in it has a few more turns for the output stage)

Back on track: I think I'm just going to call it 2%. Subtract 2% for good measure from the 1% or 0.7% THD figure.

It's really disingenuous of all these manufacturers to rate their stuff at 0.1% or 0.7% or 1%... The reason they do it is so that you can't compare the actual value which is always better and often 0.04 or 0.023 or 0.0088%. I think it's a collective silent agreement because the amplifiers they make vary so much and the value doesn't always line up with the cost of the units or position in their lineup. It should, at least to an extent, and a value of 0.08% minimum THD+n (ie. at max power) would actually be an unacceptable value for, say, a Denon X1700H, but because they never print the real numbers, they could release something like that and get away with it with 95% of their customers (the less discerning or with poorer speakers etc.)

You should also add thermals to all the above.
Things tend to change as temp goes up,sometimes cooling is critical.

 

[nagster]

Changes in response due to load fluctuations are also likely to be unique to the amplifier.
Therefore, it is necessary to measure at least each load (2/4/8ohm).
OP's suggestion is interesting, but actual sweep measurements may be unavoidable in order to make a decision.

various knees.

 

[solderdude]

Back on track: I think I'm just going to call it 2%. Subtract 2% for good measure from the 1% or 0.7% THD figure.

2% difference in power is a bit less than 0.1dB ... to put this into perspective.

 

[mhardy6647]

My speakers are extremely low distortion, actually. Less than 0.1% THD! Most of which is second harmonic... third never breaches higher than 0.03% when the speaker is performing within specifications (which I assume means until the underhung woofer leaves its gap, which is when the cone is moving more than 12mm peak to peak (8mm coil, 20mm gap). I think >100Hz.

Oh, I'd like to see and read more about this!

 

[gwing]

I'm not sure why you would want to know at what point clipping starts (aside from general interest) but as it has been said earlier it isn't going to be possible to calculate your clipping point from the 1%distortion figure.

Short of having the actual curve to refer to I generally consider the point where distortion starts to rise as the start of minimal clipping (maybe I'm not being strictly accurate here?) and we do sometimes get that quoted in the manufacturers specs in the form of 'distortion x% at y frequency'. It would be nice if it was common/standard to also list maximum power output at say 0.1% as well as 1% and 10%.

 

[mike7877]

I'm not sure why you would want to know at what point clipping starts (aside from general interest) but as it has been said earlier it isn't going to be possible to calculate your clipping point from the 1%distortion figure.

Short of having the actual curve to refer to I generally consider the point where distortion starts to rise as the start of minimal clipping (maybe I'm not being strictly accurate here?) and we do sometimes get that quoted in the manufacturers specs in the form of 'distortion x% at y frequency'. It would be nice if it was common/standard to also list maximum power output at say 0.1% as well as 1% and 10%.

Yeah, just for general interest. If you want to know the 10% figure from 1% (if it's an AB amplifiers which behaves properly (so follows the signal and when the signal makes the output higher than the rail, the voltage remains at the rail until the input falls to a level that it doesn't) it's +19%.

To get 1% from 10 (the more likely scenario by far I think) you use the inverse.

eg.
Amplifier is rated 40WPC @ 10% THD+n:

(1/1.19) * 40

= 33.6W


Like a lot of the information large corporations give us, audio measurements seem to have been chosen and omitted over the years to say the absolute least about something as possible.
For example... how is it that home theater receivers are rated into at most two channels... The front three should be driven simultaneously to 100%, while all the remaining should be driven to 40%, and when the voltage droop on the rails causes clipping of the front channels, that should be their "home theater power". Or something pretty similar.

But no, instead they rate the HT receiver like it's a stereo receiver... it's not even close lol...

It seems like they want to make things look as good as possible using the fewest types of measurements possible.
Amplifiers which have a wider bandwidth than usual, say their -3dB point is at 170kHz... Manufacturers don't tell you things that could possibly sell their devices better

 

[mike7877]

Changes in response due to load fluctuations are also likely to be unique to the amplifier.
Therefore, it is necessary to measure at least each load (2/4/8ohm).
OP's suggestion is interesting, but actual sweep measurements may be unavoidable in order to make a decision.

various knees.

This was more meant to be generally applicable to ab amplifiers, to do entirely with what happens to power consumption as THD rises from clippping.

I think 1.5% or 2% is probably a good value to use. I was hoping someone knew the math or had an amplifier in a circuit simulator

 

[DVDdoug]

I've said this before --- I've NEVER heard ANY distortion from ANYTHING unless it was broken or overdriven. (I don't claim to have golden ears, but most people who think they do "don't believe in" blind listening tests.)

I've heard noise and I've heard frequency response variations. (Most frequency response issues I've heard were from speakers & headphones, and from records & phono cartridges back in the analog days.)

If I was buying an amplifier I wouldn't even look at the distortion spec. I'd look at the power spec, but I wouldn't fully trust it unless it's independently measured. Noise specs are pretty useless too because there are too many different measurement standards/methods. There's no reason for an amplifier not to have flat frequency response over the audible range so that's not something I worry about, but I probably would check the specs.

Of course, the ASR measurements are standardized and comparable.

 

[gwing]

Yeah, just for general interest. If you want to know the 10% figure from 1% (if it's an AB amplifiers which behaves properly (so follows the signal and when the signal makes the output higher than the rail, the voltage remains at the rail until the input falls to a level that it doesn't) it's +19%.

To get 1% from 10 (the more likely scenario by far I think) you use the inverse.

eg.
Amplifier is rated 40WPC @ 10% THD+n:

(1/1.19) * 40

= 33.6W


Like a lot of the information large corporations give us, audio measurements seem to have been chosen and omitted over the years to say the absolute least about something as possible.
For example... how is it that home theater receivers are rated into at most two channels... The front three should be driven simultaneously to 100%, while all the remaining should be driven to 40%, and when the voltage droop on the rails causes clipping of the front channels, that should be their "home theater power". Or something pretty similar.

But no, instead they rate the HT receiver like it's a stereo receiver... it's not even close lol...

It seems like they want to make things look as good as possible using the fewest types of measurements possible.
Amplifiers which have a wider bandwidth than usual, say their -3dB point is at 170kHz... Manufacturers don't tell you things that could possibly sell their devices better

Thanks for the numbers and useful example.

Extrapolating from the clipping curve is presumably OK for sensible low distortion amplifiers. If you had a really bad amp that never got below say 2% distortion you'd have a hard job extrapolating from the 10% distortion level to get the output for 1% or less distortion . Likewise if you had two amps with the same power output at 1% distortion where one was low noise but already clipping and the other where the 1% distortion was mainly/all noise then you would I imagine get quite different values for their 10% distortion levels.

 

[solderdude]

This was more meant to be generally applicable to ab amplifiers, to do entirely with what happens to power consumption as THD rises from clipping.

I think 1.5% or 2% is probably a good value to use. I was hoping someone knew the math or had an amplifier in a circuit simulator

This depends on the circuit and other circumstances so there is no single circuit to simulate.
There also cannot be a rule for this as it is design dependent.

Besides 1% is not audible nor visible on a scope and it is not the clipping level either. This is usually higher. 10% of short peaks also is not audible, just 1dB louder will be in most cases.
From 1% to clipping can take place within 0.5dB and can be a few dB depending on the circuit.

When one wants to go noticeably louder you will need at least double to quadruple the power, it is not a matter of percentage.

 

[mike7877]

This depends on the circuit and other circumstances so there is no single circuit to simulate.
There also cannot be a rule for this as it is design dependent.

Besides 1% is not audible nor visible on a scope and it is not the clipping level either. This is usually higher. 10% of short peaks also is not audible, just 1dB louder will be in most cases.
From 1% to clipping can take place within 0.5dB and can be a few dB depending on the circuit.

When one wants to go noticeably louder you will need at least double to quadruple the power, it is not a matter of percentage.

I think I met my general aim - multiply by ~0.985 to get pre-clipping power rating. AKA for most amplifiers (50-200W) the difference is no more than 4 watts, which would be difficult to discern in the best of circumstances

 

[DonH56]

Generally the difference in power from x% distortion to clipping depends upon how manufacturers rate the amp, what headroom they build into their spec, and how much feedback is present. Then you have to consider if the clipping is short-term or long-term, thermal issues, and so forth. And finally you must decide what "clipping" means to you -- is it when the waveform is completely flat-topped, some high level of distortion, or what? I do not think there can be a simple metric because designs differ greatly, as do specs. One manufacturer may rate distortion low and thus limit their power spec, whilst another may spec higher distortion (closer to clipping) to provide a higher power number. Rated full power does not always correlate to clipping or to headroom.

Clipping is usually estimated by looking at the knee of the distortion curve as power is increased. An amp with low or "no" feedback may have a very "soft" knee, making it hard to draw a line before hard clipping (flat-topped waveforms), and a rather broadly sloped distortion line as it begins to clip. An amp with high feedback will have low distortion until nearly at clipping, resulting in a very sharp knee and almost vertical line. The difference could be several dB as others have said so there is not a simple (linear) ratio for what is fundamentally a very nonlinear operation. Some manufacturers market 6 dB headroom, others almost none, again greatly affecting the power rating "before clipping". Using a single number could greatly over- or underestimate the actual performance.

 

[mike7877]

I've said this before --- I've NEVER heard ANY distortion from ANYTHING unless it was broken or overdriven. (I don't claim to have golden ears, but most people who think they do "don't believe in" blind listening tests.)

I've heard noise and I've heard frequency response variations. (Most frequency response issues I've heard were from speakers & headphones, and from records & phono cartridges back in the analog days.)

If I was buying an amplifier I wouldn't even look at the distortion spec. I'd look at the power spec, but I wouldn't fully trust it unless it's independently measured. Noise specs are pretty useless too because there are too many different measurement standards/methods. There's no reason for an amplifier not to have flat frequency response over the audible range so that's not something I worry about, but I probably would check the specs.

Of course, the ASR measurements are standardized and comparable.

This is just about watts, I knew the difference would be small - just wanted to quantify it