Headphone amplifier power spec distortion cut-off

[amirm]

Hey you all...

In my last headphone amplifier review (Schiit Lyr vs Magni 3) I just pointed out where the onset of clipping occured for the two amplifiers:

That caused confusion and question as to how they relate to manufacturer's specs. And at any rate, that clipping point was different for each amplifier so you couldn't compare the two numbers against each other.

So question to the community is, what cut off point for THD+N do we want to use to provide our "official power spec" for headphone amplifiers? Picking a random 0.1% number would not even work for the tube one above as it barely gets there.

Anyone seen good criteria for this anywhere or has an opinion?

 

[pkane]

Hey you all...

In my last headphone amplifier review (Schiit Lyr vs Magni 3) I just pointed out where the onset of clipping occured for the two amplifiers:

That caused confusion and question as to how they relate to manufacturer's specs. And at any rate, that clipping point was different for each amplifier so you couldn't compare the two numbers against each other.

So question to the community is, what cut off point for THD+N do we want to use to provide our "official power spec" for headphone amplifiers? Picking a random 0.1% number would not even work for the tube one above as it barely gets there.

Anyone seen good criteria for this anywhere or has an opinion?

Stereophile seems to be using 1% THD+N as the clipping point for power amplifiers. May make sense to use the same for head amps.

 

[Dimitri]

Anyone seen good criteria for this anywhere or has an opinion?

As far as criteria go, a brief uninformal surveof published specifications will show that "most" tube amps specify a THD at full power hovering around 0.1 to 0.5 .

If you were to get fancy and differentiate measuring 2nd& 3rd order harmonics vs "all else higher" ...things get complicated (more on this here - footnote 2 : https://www.stereophile.com/content/vtlmanley-reference-350-power-amplifier )


edit - footnote 2

 

[mindbomb]

For a headphone amp, I think even .01% would function as a good cutoff for a lot of amps. Although in this case, neither amp actually gets that low.

A high end headphone amplifier, the neurochrome hp1, for example, used .005% for 32 ohms

https://www.neurochrome.com/hp-1-ultra-high-end-headphone-amp/

 

[DonH56]

For clipping I would use 1%. It is somewhat arbitrary but that is clearly around the knee of almost anything these days and allows comparison to numerous previous results from all over.

 

[mindbomb]

I think part of the beauty of 0.1% thd+n is that, psychoacoustically, even the most pernicious types of distortion like crossover distortion are still not really audible afaik.

 

[garbulky]

You got your magni power output wrong @amirm if you were going for 0.1%

 

[amirm]

You got your magni power output wrong @amirm if you were going for 0.1%

That's the point of the thread. I did NOT go by 0.1% or any other fixed number. I simply pointed out where clipping occured. To me there is no usable power after that. Problem is we can't compare different devices against each other. I will continue to report on this clipping point but like to also state where the cut off power point is.

 

[Dismayed]

That's the point of the thread. I did NOT go by 0.1% or any other fixed number. I simply pointed out where clipping occured. To me there is no usable power after that. Problem is we can't compare different devices against each other. I will continue to report on this clipping point but like to also state where the cut off power point is.

It would seem that the chart is far more informative than is a single summary statistic.

 

[Sal1950]

For clipping I would use 1%. It is somewhat arbitrary but that is clearly around the knee of almost anything these days and allows comparison to numerous previous results from all over.

+1
Also the level that was chosen by Adcom (Nelson Pass?) to trigger the "Instantaneous Distortion Alert" lights on the front panel of the GFA series of amplifiers. A arbitary choice in any case but sounds like a good point to me.

 

[SpeedyRodent]

Agreed with @DonH56 with 1%.

 

[amirm]

Thanks for the feedback everyone. I guess what I might do is to actually listen to 1% distortion with headphones and see how usable it is. I could capture it, attenuate it and listen to it that way.

 

[Dismayed]

How about THD of 0.56%. Then the signal would be 99 and 44/100% pure.

 

[DonH56]

Thanks for the feedback everyone. I guess what I might do is to actually listen to 1% distortion with headphones and see how usable it is. I could capture it, attenuate it and listen to it that way.

I made some sine wave files with various amounts of distortion a few years ago and posted them. I added it to the waveforms so you could play them at any level you liked (though if the amp added another 1% you had no way of knowing, natch). In a DBT using headphones I could pick out 1% pretty readily; 0.1% not so much. But 1% on a multitone sample was pretty hard to pick out. And of course there's always the caveat about the harmonic structure of the distortion; I added second and third-order distortion at various levels and blew off the higher-order (arguably easier to hear) components. Be interesting to see what you think.

IIRC the IHF used 1% and of course Stereophile has used that for years. I prefer it just because it is a common reference point.

 

[RayDunzl]

Where's a tool to add specified harmonic distortion to music?

 

[DonH56]

Matlab or Mathcad, don't remember which I used. Probably others, but for a simple test that is what I used.

 

[RayDunzl]

Where's a tool to add specified harmonic distortion to music?

Of course, music is distortion. Complex waveforms.

Single note on a bass guitar (measured electronically, not in the air):

Ooh... look at those spreading bases... Is my string jitterey?

 

[restorer-john]

I would like to add this as I spend a lot of time testing and repairing amplifiers.

Clipping is different to THD.

You cannot pick an arbitary THD number and declare it clipping. Clipping can be caused by various means, not just running into the supply rails. More often than not, some types of clipping are a design element in amplifiers where current limiting is applied. The level and aggressiveness of certain types of current limiting will manifest as various distortions, not necessarily harmonically related.

The bottom line is this, Stereophile are being lazy. What is most important, and is a measurement that seems to have disappeared, are the visually depicted (scope shots) effects of overload (be it transient, EIA/J toneburst or continuous clipping), their waveforms (oscillation etc) and the recovery aspects (time to stabilize in cycles etc). This was commonly tested and reported in magazine reviews in the past, both into reactive and resistive loads.

 

[restorer-john]

Where's a tool to add specified harmonic distortion to music?

Prerecorded tracks. The Sheffield Test Disc (10045-2-1)

Tracks 71-76 and tracks 77-82 cover 0/0.1/0.3/1/3/10%

Very illuminating to listen to...

 

[DonH56]

I also added clipping at I think 1%, 3%, 10%, and 50% to some files to show the harmonic spray and let folk hear what it sounds like. Not sure those are here now, but I think Amir copied the basic clipping article over... And as restorer-john says, introducing clipping to a test waveform does not necessarily (re)produce all the "badness" a real amplifier might induce, nor does it include the effect of soft-clipping circuits and HF rolloff that might reduce the impact of clipping.