Is THD+N audible for sensitive IEMs at quiet volumes?

[marksquared]

Looking at the THD+N graphs which chart THD+N percentage vs power, it seems like distortion is quite high at low volumes. Is that accurate?

I often listen with the Sennheiser IE 800S, and it has a sensitivity of 125 dB SPL/V and impedance of 16 ohms. To try and estimate the mW at my typical listening levels, I used the following:

Doubling/halving power changes the output by 3 dB, so I need to divide 1000 mW by 2^13 to get from 125 dB to 85 dB and by 2^26 to get to around 45 dB.

That gives me a typical power output between 0.2 uW and 100 uW.

Here are some typical THD+N values at the far left of the charts (usually 0.3 or 0.4 uW) at 33 ohms:

Topping DX3Pro: 0.1%
JDS Labs Atom: 0.09%
Massdrop THX AAA 789: 0.2%

In @flipflop's post about audibility, he says that 0.05% is the lenient threshold for audibility, and 0.0001% is the strict threshold. So is it correct to say that even with the best-measuring DACs and amplifiers, I would be getting audible levels of THD+N in the IE 800S?

Is this an advantage of using harder-to-drive headphones which require more power and thus operate closer to the sweet spot of the typical headphone amplifier?

 

[solderdude]

The THD+N at lower output power is noise dominated.
The noise floor doesn't change but the signal level is coming closer to the noise level.
This effectively lowers the S/R ratio.
If you are not hearing the noise (no signal playing) then you don't need to worry about it (noise nor distortion).

When you want to lower the noise floor and make more usage of the amps dynamic range use THIS or THIS

 

[Blumlein 88]

Thd+n is distortion plus noise. At lower levels noise is a bigger part of it. Actual distortion is likely going down. As noise spreads across the whole bandwidth you'll likely not notice it. So distortion is actually low at low levels.

 

[marksquared]

That's very helpful, thanks! On most gear I can't hear the noise floor, even with the IE 800S. One recent exception was the WHAMMY headphone amp which I built recently. The noise floor is inaudible with my Sennheiser HD 660S but very audible with the IE 800S.

 

[flipflop]

In @flipflop's post about audibility, he says that 0.05% is the lenient threshold for audibility, and 0.0001% is the strict threshold. So is it correct to say that even with the best-measuring DACs and amplifiers, I would be getting audible levels of THD+N in the IE 800S?

The thresholds mean that everything below them should be inaudible, not that everything above will necessary be audible.
I got rid of the THD+N vs power graphs from my thread because there are better measurements for determining distortion and noise levels. The main purpose of those graphs is to tell you how much power an amp can deliver into a given load.

 

[solderdude]

That's very helpful, thanks! On most gear I can't hear the noise floor, even with the IE 800S. One recent exception was the WHAMMY headphone amp which I built recently. The noise floor is inaudible with my Sennheiser HD 660S but very audible with the IE 800S.

THIS or THIS is the solution. As you built the WHAMMY building an attenuator will be easy.

The smaller the output signal the more 'linear' the amplification is meaning the distortion will be less. In other types of distortion plots you can actually see distortion products getting smaller in amplitude and eventually drop below the noise floor.
No need to worry about distortion at small signal levels.

 

[marksquared]

The thresholds mean that everything below them should be inaudible, not that everything above will necessary be audible.
I got rid of the THD+N vs power graphs from my thread because there are better measurements for determining distortion and noise levels. The main purpose of those graphs is to tell you how much power an amp can deliver into a given load.

Thanks @flipflop. Is the linearity graph the best one to look at for distortion levels?

 

[flipflop]

Thanks @flipflop. Is the linearity graph the best one to look at for distortion levels?

All the measurements dealing with distortion, that are outlined in my thread, are useful. I personally prefer the FFT in the dashboard view.

 

[Blumlein 88]

Thanks @flipflop. Is the linearity graph the best one to look at for distortion levels?

I don't really like THD+N. I like THD (done with a deep FFT) and then noise done separately. I understand why the THD+N is combined. And if a device has a good THD+N you know both noise and THD is good.

Normally with the exception of the ESS IMD bump, as signal level drops so does distortion. As long as you know that, you can be pretty sure that those THD+N curves that start to rise with decreasing level are rising due to noise being greater than distortion and taking on a larger part of a decreasing signal level. Just the reverse when the curve starts going up near max level. You can be pretty sure actual distortion is increasing.

For a blatant extreme example consider the following. A device has THD+N of -100 db at max level (let us say 2 volts). That -100 db is a ratio of signal level to noise plus distortion. Let us assume this is the worlds first perfect distortion free device. There is no IMD or harmonic distortion. There is a -100 db noise floor. If we lower the signal to 1 volt (a 6 db reduction) at that level THD+N has increased to -94 db. The absolute measured noise level is unchanged (it is .00002 volts in both cases), there is no distortion, but the ratio of signal to residual noise is smaller. If we keep going when the signal is down to -80 db, the noise is still .00002 volts or -100 db full scale. But the ratio of signal to noise is now only -20db THD+N. Yet in none of these circumstances was there any harmonic distortion at all with our idealized device.

By the time most real world devices are down -20 db from max, any harmonic distortion has dropped below the noise floor, and THD+N usually is only showing noise by that point. One of the reasons I don't consider it much of a measurement except at full scale. You do have to check such things though as now and again some device will work oddly and not follow that pattern.

 

[andreasmaaan]

I agree with all the comments advising that THD+N at low signal levels is dominated by noise and that if you can’t hear noise at such levels then distortion is likely nothing to worry about.

It’s also worth noting though that, generally speaking, an amplifier will produce higher distortion into a lower impedance load. So although it’s unlikely to be audible, the amount of distortion your amp is producing when feeding low-impedance IEMs is nevertheless likely to be greater than the amount of distortion it produces when feeding higher impedance headphones.

This is arguably the main potential advantage of higher impedance headphones.