How much power to headphones really need?

[Purpl3n3ss]

So, I understand that loudness is supposed to be the only thing that power dictates, not other characteristics of the sound.

However, here's something I tried:

I took my 32ohm-impedance/94db-mW-senstivity headphones (Hifiman Arya SE) and drove them directly from the 1Vrms output of a DAC (Chord Qutest, which has very low output impedance). The calculated volume is 108.9db.

Then I drove the same headphones setting the same DAC to output 3Vrms. The calculated volume is 118.5db. But, upstream from the DAC, in software, I attenuated the signal by the difference in volume - I had to use 10db instead of 9.6db - resulting in a volume of 108.5db.

With this (close) volume matching done, the second setup, which puts 282 mW into the headphones instead of 31mW in the first setup, clearly had a greater depth of soundstage.

What could be going on here?


A note about the software attenuation: It was done in HQPlayer using 15th-order noise-shaping (LNS15), while upsampling to 768Khz. My understanding is that this is about as transparent as software attenuation gets. And in any case, it was the version with the software attenuation that displayed the deeper soundstage.

 

[solderdude]

With this (close) volume matching done, the second setup, which puts 282 mW into the headphones instead of 31mW in the first setup, clearly had a greater depth of soundstage.

Except in both cases there is no 'more power', the power is the same, the output voltage is the same, the amp section reacts the same (exact same load), the headphones reacts the same.
So any perceived differences are placebo or a level difference or there is something fishy (clipping) going on without digital attenuation and there is enough digital headroom left in the 'digital attenuated' test.

One does not 'push' power into a headphone. One applies a voltage and the headphone draws current (= power).

 

[Purpl3n3ss]

Except in both cases there is no 'more power', the power is the same, the output voltage is the same, the amp section reacts the same (exact same load), the headphones reacts the same.
So any perceived differences are placebo or a level difference or there is something fishy (clipping) going on without digital attenuation and there is enough digital headroom left in the 'digital attenuated' test.

One does not 'push' power into a headphone. One applies a voltage and the headphone draws current (= power).

How is the output voltage the same though? In the first case I set it to 1Vrms and in the second to 3Vrms. Likewise, the current drawn in higher in the second case.

Here are calculations for the two cases:

1V x 31.1mA = 31.1mW:

Headphones Power Calculator

Calculate the power needed to drive your headphone based on sensitivity (db/mw or db/V), impedance (ohms) and loudness (dbSPL)

www.headphonesty.com

3V x 93.9mA = 281.7mW

Headphones Power Calculator

Calculate the power needed to drive your headphone based on sensitivity (db/mw or db/V), impedance (ohms) and loudness (dbSPL)

www.headphonesty.com

 

[solderdude]

If that were really the case and you did not use digital attenuation in the second case the SPL would be 10dB higher (twice as loud).
You clearly stated that in the second attempt you lowered the input level by 10dB.

So the amp did not output 3V but close to 1V leading to the same SPL.

 

[MaxwellsEq]

The output voltage is the same. It would only go up to 3V RMS if you fed it an unattenuated signal, but you are reducing the end to end gain until the sound level is the same, which means you are actually feeding it 1V RMS.

 

[Purpl3n3ss]

Ah, thanks, that clears up what was confusing me!

It doesn't explain the difference I'm hearing though, unless it really is all placebo. I'll do some further listening.

 

[MaxwellsEq]

Despite your level matching, if you consider the end-to-end chain from a global systemic point of view, you are listening to two different "systems". You have changed how the system creates gain, by changing the responsibility for that, and you have changed the noise floor. Designing end-to-end gain chains (e.g. in a channel of a mixing desk) to maximise operator choice whilst minimising noise requires considerable knowledge.

 

[paulrbarnard]

Ah, thanks, that clears up what was confusing me!

It doesn't explain the difference I'm hearing though, unless it really is all placebo. I'll do some further listening.

We’re you expecting the, what you thought was, higher level signal to sound better? If you did it probably would. It’s one of the amazing things about the mind and it’s ability to influence our senses.

 

[Jimbob54]

Ah, thanks, that clears up what was confusing me!

It doesn't explain the difference I'm hearing though, unless it really is all placebo. I'll do some further listening.

Have you got someone you could train as a helper? Show them how to adjust the levels properly- send you out of the room as they do(or maybe dont) make the change and you just listen. If you can tell which set up is which 9/10 times then its a fair bet there is an actual difference.

Then you need to find out why. My money would be on there still being a difference in output voltage if you tested with an accurate meter.

 

[Purpl3n3ss]

We’re you expecting the, what you thought was, higher level signal to sound better? If you did it probably would. It’s one of the amazing things about the mind and it’s ability to influence our senses.

I was expecting (what I thought was) the higher level signal to sound better, but I was hoping that it wouldn't. So, my subjective evaluation seemed to confirm my expectation, but also go against my hopes.

 

[Purpl3n3ss]

Despite your level matching, if you consider the end-to-end chain from a global systemic point of view, you are listening to two different "systems". You have changed how the system creates gain, by changing the responsibility for that, and you have changed the noise floor. Designing end-to-end gain chains (e.g. in a channel of a mixing desk) to maximise operator choice whilst minimising noise requires considerable knowledge.

Thanks. By "changed how the system creates gain", if I understand correctly, you mean that in the digitally attenuated case, the DAC "compensates" for the digital attenuation to achieve the same volume/voltage/power. As for the noise floor, just a note that in both cases I'm upsampling to 768Khz and using noise shaping (but only in the second am I applying -10db).

 

[solderdude]

Why not play the music natively and not upsample to 768kHz (it really has no benefits when using the Qtest.

 

[Purpl3n3ss]

Why not play the music natively and not upsample to 768kHz (it really has no benefits when using the Qtest.

Because when using digital volume control in HQPlayer it is apparently best to upsample and use noise-shaping at a high sample rate, as the creator of the software says here:

Death of the pre-amp and HQPlayer digital gain?

Hey @jussi_laako Let me start by saying your application is simply outstanding - listening test after listening test confirms that a properly setup HQPlayer in the media stream will deliver superior results in most setups. Well done. Last night, during a listening session, my buddy and I...

community.roonlabs.com

Rob Watts, who designed the Qutest, also seems to be saying something similar here:

Chord Electronics Qutest DAC - Official Thread

It will not be quite as good; when designing Blu 2's OP truncators, I had the 768 kHz noise shapers (guaranteeing 350 dB performance so maintaining the original precision), gaussian dither and TPDF dither; TPDF had the worst sound quality with a reduction in depth. Gaussian gave closer to the...

www.head-fi.org

 

[solderdude]

I know the theory and that those 2 persons who sell those upsamplers claim that it is 'better'.
Still... try without any upsampling.