How Often do Amplifiers Clip?

[RayDunzl]

All three values are plotted against the left (vertical axis). The vertical axis is SPL in dB, and Voltage in volts, and Power in watts.

The SPL values show along the horizontal axis, just for a point of reference.

As SPL numbers rise linearly, an exponential increase in voltage is required (double voltage every 6dB of SPL), which produces an exponential increase in power (which follows the square of the voltage).

I just thought this display would visually indicate the rather dramatic increase in power required for small numerical increases in SPL at higher sound pressure levels..

 

[Frank Dernie]

Specific to one manufacturer but interesting nonetheless.
http://www.sanderssoundsystems.com/technical-white-papers/172-tubes-vs-transistors

 

[RayDunzl]

My little baby Krells are rated to put 138V across the speaker leads (peak)

138^2 / 4 ohms = 4,761 watts peak - not RMS. RMS rating is 700W, but may use only one leg of the differential output for that spec.

That, or even a small portion of it, is "enough".

I think I've seen 50Vpk before when measuring and playing loudly.

50^2 / 4 ohms = 625 watts peak (if there is such a measure)

 

[Sir Sanders Zingmore]

Specific to one manufacturer but interesting nonetheless.
http://www.sanderssoundsystems.com/technical-white-papers/172-tubes-vs-transistors

I like Roger Sanders. I own one of his amps. Paired with my newer, more efficient horns, I’m pretty sure I never get close to clipping.

 

[A800]

Very often.

 

[pozz]

Let's try something.

Below are wattage requirements for a single speaker with 84dB, 87dB or 90dB SPL sensitivity, and 8 ohm nominal impedance. I've included a linear scale version just for the first chart.

• For two speakers, add 3dB.
• For 4 ohm impedance, add 3dB.
• For 16 ohm impedance, subtract 3dB.

Here are continuous and burst power results for amps that @amirm measured at both 4 ohm and 8 ohm.

So, as others have said, amps will clip a lot.

 

[GeorgeBynum]

I'm comfortable with what you say, but think overDRIVEN may need to be differentiated from overLOADED. A kW amp into highly efficient speakers played at a 90dB SPL in a small den would almost certainly be negligibly loaded, but if somehow it were fed (extreme case, line levels into a microphone input as an example) with excessively high signals would clip ... but in input or intermediate stages, not the output. The same speakers connected to a 0.5W amp would overload the output devices and we'd see clipping.

Of course, an electronics nerd like myself would connect my oscilloscope to the speaker outputs and observe it.

All will clip when over-driven.

As long as one never plays loud the amp may never clip.

 

[Twelvetone]

I've been convinced for many year that all (competent) amplifiers OPERATING WITHIN THEIR LINEAR AREAS are sonically indistinguishable -- but that's a huge "if." I'm equally convinced that in the real world of typical loudspeaker sensitivities (say, 85-91 dBSPL/w/m), listening distances (2.5-4.5m), room acoustics (moderate absorption), and listening levels (ref to -10 dB), on ANY reasonably dynamic program (not pipe organ or highly-compressed pop, f.e.), many, and very likely nearly all consumer audio systems visit momentary clipping many times per typical minute (absent quiet passages). Consequently, I believe it is the behavior, upon entering and -- more critically -- exiting non-linearity, of the amplifier and its "protection" circuitry, that causes these "audible amplifier differences" we hear so much about. Therefore, I also answer: "often."

 

[solderdude]

I'm comfortable with what you say, but think overDRIVEN may need to be differentiated from overLOADED. A kW amp into highly efficient speakers played at a 90dB SPL in a small den would almost certainly be negligibly loaded, but if somehow it were fed (extreme case, line levels into a microphone input as an example) with excessively high signals would clip ... but in input or intermediate stages, not the output. The same speakers connected to a 0.5W amp would overload the output devices and we'd see clipping.

One can overdrive an amplifier (make it clip) without overloading it when you have speakers with an impedance that doesn't reach the current limit and SOA.
One can also overload an amp without overdriving it (impedance too low) where the current limit is reached yet not the voltage limit.
Of course this too is clipping but current clipping and not voltage clipping even though the effect is a clipped output voltage.

 

[DonH56]

Let's try something.

Below are wattage requirements for a single speaker with 84dB, 87dB or 90dB SPL sensitivity, and 8 ohm nominal impedance. I've included a linear scale version just for the first chart.

• For two speakers, add 3dB.
• For 4 ohm impedance, add 3dB.
• For 16 ohm impedance, subtract 3dB.

View attachment 59905
View attachment 59907
View attachment 59914
View attachment 59915
Here are continuous and burst power results for amps that @amirm measured at both 4 ohm and 8 ohm.
View attachment 59924

So, as others have said, amps will clip a lot.

Would you be willing to share/send your spreadsheet? I'd like to play with a bit for my own system out of curiousity.

Curious - Don

p.s. The AES, or IHF, or someone (forgot), ages ago put the peak-to-average ratio in music at around 17 dB. In terms of power, that means you need to have around 50x your average level to keep from clipping the peaks. I have seen studies showing ~20 dB (100x) for high-DR music and ~30 dB (1000x) for movies, though for the latter I doubt many would notice if an explosion or gunshot was clipped...

 

[Wes]

Here is the central issue - if amps are not distinguishable unless clipping then the likelihood of clipping is important.

If amps rarely clip, then it does not matter if tube amp clips "softly."

 

[pozz]

Would you be willing to share/send your spreadsheet? I'd like to play with a bit for my own system out of curiousity.

Sure, attached.

Excel can't handle logs of negative numbers, just FYI.

 

[Twelvetone]

Here is the central issue - if amps are not distinguishable unless clipping then the likelihood of clipping is important.

If amps rarely clip, then it does not matter if tube amp clips "softly."

My point exactly; without specifics it's impossible to know or even guess, but in the aggregate, where the definition of "clip" is output non-linear waveform however briefly, I'm betting all the time.

 

[DonH56]

Sure, attached.

Excel can't handle logs of negative numbers, just FYI.

Thanks! And yah, had forgotten about that restriction, used to Matlab and such.

 

[Blumlein 88]

@pozz good work on the chart.

I guess it is worth mentioning that past the critical distance the sound level doesn't drop off at 6 db for doubling. And that sound levels due to built up reflected energy will be higher than the theoretical chart at greater distances. Not sure how would could account for this. Using your charts one can be sure they have enough, and if the reverberated sound built up gives extra then you have a margin of safety.

I also think if you look at energy distribution vs frequency you could come up with a curve with less power needed for tweeters and midranges. So having 4 big subwoofers is a smart idea. By then we are complicating the issue plenty of course.

 

[Koeitje]

I am no expert, but won't the crest factor for music and especially movies cause a lot of clipping?

 

[pjug]

To those of you with scopes, what do you see when you play something dynamic turned up as loud as you can stand it? When I did this I never hit 30V (i am going from memory here, so not 100% sure). This with 86dB speakers 8-ohms. 16'x20' room, about 10' listening distance. I would say my threshold for too loud might be lower than some other people.

Maybe some of us should do this all using same piece of music and report what we see?

 

[pozz]

@pozz good work on the chart.

I guess it is worth mentioning that past the critical distance the sound level doesn't drop off at 6 db for doubling. And that sound levels due to built up reflected energy will be higher than the theoretical chart at greater distances. Not sure how would could account for this. Using your charts one can be sure they have enough, and if the reverberated sound built up gives extra then you have a margin of safety.

I also think if you look at energy distribution vs frequency you could come up with a curve with less power needed for tweeters and midranges. So having 4 big subwoofers is a smart idea. By then we are complicating the issue plenty of course.

I think I'd have to learn MATLAB to account for all those factors. Though if I did, and was actually comfortable working out the math for a specific listening position in 3D space, I could start consulting for acoustics. Ah, well.

 

[DonH56]

The simple guesstimate (rule of thumb) is add 3 dB against a wall, 6 dB in a corner. Actual in-room response depends upon the reflectivity index of the walls; amount, position, and frequency response of absorbers/diffusers (panels, furniture, etc.) in the room, effect of window wells, doors, etc.

 

[AudioJester]

Late to this thread, but I have a question about clipping.
Can software detect and prevent clipping? I sometimes use HQPlayer and the volume control will go "red" to indicate clipping - I actually dont here any distortion. Can software predict clipping? This is with convolution in play and headroom adjustment.
Roon has a similar clipping indicator.

I guess that leads to my next question. How do you work out amplifier requirements with DSP/Convolution filters. If I use Acourate to generate a filter <350Hz, I know it puts more stress on the amplifier. How do I work out how much power/current I need in an amplifier in this situation?
Its not as easy as matching speaker and amp specs?