Review and measurements of MacBook Air (2024) headphone output

[mcdn]

This is a review and measurements of the MacBook Air 2024 headphone output. It was prompted by a request from @DrSpan, as the little Mac has a very strong reputation as a headphone source. Here it is, measuring itself:

Since 2021 Apple has supported impedance detection on the headphone outputs of its laptops, meaning low impedance headphones get more current, while high impedance headphones get more voltage. When you plug in the headphones the MacBook checks the impedance, and sets the output voltage to 1.25V if it's less than 150ohms or 3V if it's up to 1kOhms/. Above that it assumes you're using it as line interface and sets the voltage to 1V. It only does this check when you physically plug in the jack, which caught me out a couple of times in testing!

So the first question is does the impedance sensing work? It absolutely does, and the voltages are superbly accurate at 1.00V (4.7kOhms) and 1.25V (30ohms), dropping a tiny bit from spec to 2.97V at 300ohms.

The second question people have is how much power it can actually deliver? I won't make you wait, it delivers exactly to spec into 300ohms at 29mW, so essentially no voltage drop. Into 30ohms I got 48mW against an ideal of 52mW, and into 15ohms the gap widened a bit more with 85mW against an ideal of 104mW.

OK, let's do some graphs! The measurements were taken with a Cosmos ADCiso and a resistive dummy load. The ADC was running at 96kHz/24bits, with the laptop outputting 44.1kHz/float32, which is the default. The laptop was connected to external power via a thunderbolt hub. The headphone output itself is a 3.5mm TRS jack, i.e. unbalanced stereo.

Here's the best SINAD I could get, which was at 0dBFS output into 30ohms, delivering a SINAD of 97.9dB:

Frequency response is ruler flat to 16KHz, then drops about 6dB by 20KHz. If you change the output sample rate in "Audio MIDI setup" then you get the same result, e.g. the drop starts around 44kHz when using a 96kHz sample rate. The good news is the filter is steep and clean with no rise in noise after. It does look quite tailored, but I didn't get into impulse tests:

Here's THD+N vas power at 15, 30 and 300ohms:

And here's CCIF IMD (19 and 20kHz tones) at 30 and 300ohms:

The multitone distortion is very clean - I've overlayed all three impedances but they are pretty much the same:

Crosstalk was about -50dB. This was measured by driving one output at 0dB while measuring the other, so they share about 1m of cable before being split out for measurement:

At 50mV into 30ohms (with the Cosmos Scaler boosting the signal to about 1V into the ADC) I was able to get a SINAD of 89.1dB, which is less than the 95dB L7 Audio Lab measured but I think their measurement may be A weighted, and they have an APx555:

In conclusion definitely 5 quokkas from me! There's simply no reason to use an external headphone amp with any MacBook made since 2021 unless you have demanding headphones. It's not nearly as good for power as a good desktop headphone amp, but it is better than popular interfaces such as the Scarlett 2i2.

 

[staticV3]

Thanks a lot for measuring!

Quick note: some of your Ohm numbers don't make sense currently

Spoiler: Example

1.00V (4.7kOhms) and 1.25V (300ohms), dropping a tiny bit from spec to 2.97V at 30ohms.

and don't match the description in the graphs themselves.

Also, that 50mV graph seems very suspect. Actual performance is likely much higher there.

 

[Andreas007]

From technical point of view I expect nothing less from Apple.
They certainly have some of the best engineers and it is good to see that they are allowed to deliver.

 

[mcdn]

Thanks a lot for measuring!

Quick note: some of your Ohm numbers don't make sense currently

"Spoiler" tags seem to break quoting, but the bit you quoted was "1.00V (4.7kOhms) and 1.25V (300ohms), dropping a tiny bit from spec to 2.97V at 30ohms.", which is the measured RMS voltage outputs at 0dBFS at each of those impedances, which match Apple's spec [edited to note, I missed the obvious typo where I swapped the 30 and 300 in that sentence. It's fixed in the review now]

Also, that 50mV graph seems very suspect. Actual performance is likely much higher there.

I dunno, a 30dB drop in SINAD when reducing output from 3V to 50mV seems reasonable to me? Remember the dBFS scale here is the 20V or so of the Cosmos ADC, not the 3V of the Macbook's output

 

[staticV3]

1.00V (4.7kOhms) and 1.25V (300ohms), dropping a tiny bit from spec to 2.97V at 30ohms.", which is the measured RMS voltage outputs at 0dBFS at each of those impedances, which match Apple's spec

Apple's spec is up to 1.25V <150Ω and 3.0V at 150-1000Ω.

2.97V at 30ohms is certainly not correct.

It also does not match the description in the graph itself, which says 300ohm which is likely the correct number:

Here's the best SINAD I could get, which was at -6dBFS output into 30ohms, delivering a SINAD of 94.8dB:

 

[mcdn]

@staticV3 thank you for sense checking this, it is quite possible I've made mistakes

 

[mcdn]

Apple's spec is up to 1.25V <150Ω and 3.0V at 150-1000Ω.

2.97V at 30ohms is certainly not correct.

It also does not match the description in the graph itself, which says 300ohm which is likely the correct number:

Doh! That's just a typo when I was editing the sentence. Fixing it now, thanks!

 

[staticV3]

I dunno, a 30dB drop in SINAD when reducing output from 3V to 50mV seems reasonable to me? Remember the dBFS scale here is the 20V or so of the Cosmos ADC, not the 3V of the Macbook's output

Ahh there's the first issue. When trying to measure a 50mV signal with the Cosmos set to 20V sensitivity, the Cosmos' self-noise dominates the result so you're not actually measuring the DUT's own SNR.

See L7's measurements where he got 95dB @50mV, which is the measurement limit of even the APx555 so the MacBook itself can do more.

With both the Cosmos ADC and APx555, an additional LNA (e.g. Cosmos APU) would be required to accurately measure the MacBook at such low levels.

 

[mcdn]

Ahh there's the first issue. When trying to measure a 50mV signal with the Cosmos set to 20V sensitivity, the Cosmos' self-noise dominates the result so you're not actually measuring the DUT's own SNR.

See L7's measurements where he got 95dB @50mV, which is the measurement limit of the APx555 so the MacBook itself can do even more.

That's possibly true, but would only account for 16dB of difference at most? I took the scaler out of the loop because I was getting too much environmental noise through it, but I can try again tomorrow.

 

[staticV3]

That's possibly true, but would only account for 16dB of difference at most?

The MacBook likely has 50mV SNR around 100dB. Your measurements show 66dB.

That is a massive 34dB discrepancy.

In Amir's ranking, that would be the difference between first place and nearly last place:

 

[mcdn]

The MacBook likely has 50mV SNR around 100dB. Your measurements show 66dB.

That is a massive 34dB discrepancy.

In Amir's ranking, that would be the difference between first place and nearly last place:
View attachment 454307

I'll remove it for now.

 

[DrSpan]

This is a review and measurements of the MacBook Air 2024 headphone output. It was prompted by a request from @DrSpan, as the little Mac has a very strong reputation as a headphone source. Here it is, measuring itself:

View attachment 454294

Since 2021 Apple has supported impedance detection on the headphone outputs of its laptops, meaning low impedance headphones get more current, while high impedance headphones get more voltage. When you plug in the headphones the MacBook checks the impedance, and sets the output voltage to 1.25V if it's less than 150ohms or 3V if it's up to 1kOhms/. Above that it assumes you're using it as line interface and sets the voltage to 1V. It only does this check when you physically plug in the jack, which caught me out a couple of times in testing!

So the first question is does the impedance sensing work? It absolutely does, and the voltages are superbly accurate at 1.00V (4.7kOhms) and 1.25V (30ohms), dropping a tiny bit from spec to 2.97V at 300ohms.

The second question people have is how much power it can actually deliver? I won't make you wait, it delivers exactly to spec into 300ohms at 29mW, so essentially no voltage drop. Into 30ohms I got 48mW against an ideal of 52mW, and into 15ohms the gap widened a bit more with 85mW against an ideal of 104mW.

OK, let's do some graphs! The measurements were taken with a Cosmos ADCiso and a resistive dummy load. The ADC was running at 96kHz/24bits, with the laptop outputting 44.1kHz/float32, which is the default. The laptop was connected to external power via a thunderbolt hub. The headphone output itself is a 3.5mm TRS jack, i.e. unbalanced stereo.

Here's the best SINAD I could get, which was at -6dBFS output into 30ohms, delivering a SINAD of 94.8dB:

View attachment 454285
Frequency response is ruler flat to 16KHz, then drops about 6dB by 20KHz. If you change the output sample rate in "Audio MIDI setup" then you get the same result, e.g. the drop starts around 44kHz when using a 96kHz sample rate. The good news is the filter is steep and clean with no rise in noise after. It does look quite tailored, but I didn't get into impulse tests:

View attachment 454290

Here's THD+N vas power at 15, 30 and 300ohms:
View attachment 454286

And here's CCIF IMD (19 and 20kHz tones) at 30 and 300ohms:
View attachment 454287
The multitone distortion is very clean - I've overlayed all three impedances but they are pretty much the same:
View attachment 454288
Crosstalk was about -50dB. This was measured by driving one output at 0dB while measuring the other, so they share about 1m of cable before being split out for measurement:
View attachment 454289
[50mV measurement removed until it can be redone, thanks to @staticV3 for catching the error]

In conclusion definitely 5 quokkas from me! There's simply no reason to use an external headphone amp with any MacBook made since 2021 unless you have exceptionally demanding headphones.

Thank you heaps!
Really appreciated.

Now that i would say we have established that it has superb headphone output, how would one go about if comparing the D/A conversion between something like this
and an external audio interface that costs as much as the whole Macbook?
I am very curious as to how to compare this but lack the technical knowledge and would not know where to start.

Would you convert a file and export it or what exactly would be the argument for an Expensive D/A converter in the first place? Not talking about the practical side
like Preamps, more In-Out etc. Only the „sound quality“ difference if any and if any, what difference.
Is it an obsolete thing?
Would love some opinions on that

 

[amirm]

The second question people have is how much power it can actually deliver? I won't make you wait, it delivers exactly to spec into 300ohms at 29mW, so essentially no voltage drop. Into 30ohms I got 48mW against an ideal of 52mW, and into 15ohms the gap widened a bit more with 85mW against an ideal of 104mW.

Thanks for the review. Promoted to the home page. On these numbers though, they are quite low. Sure, you don't need a dongle but as desktop specs go, there is not much power here, especially at 30 ohm.

 

[mcdn]

The MacBook likely has 50mV SNR around 100dB. Your measurements show 66dB.

That is a massive 34dB discrepancy.

In Amir's ranking, that would be the difference between first place and nearly last place:
View attachment 454307

Fixed now, so new 50mV chart inserted, and also the top line SINAD (30ohm 0dBFS) updated too. Cabling issue with the Cosmos Scaler was behind the problem.

50mV is still slightly behind L7's, but top line is essentially identical. I'll update the other charts over time. Thanks again!

 

[mcdn]

Thank you heaps!
Really appreciated.

Now that i would say we have established that it has superb headphone output, how would one go about if comparing the D/A conversion between something like this
and an external audio interface that costs as much as the whole Macbook?
I am very curious as to how to compare this but lack the technical knowledge and would not know where to start.

Would you convert a file and export it or what exactly would be the argument for an Expensive D/A converter in the first place? Not talking about the practical side
like Preamps, more In-Out etc. Only the „sound quality“ difference if any and if any, what difference.
Is it an obsolete thing?
Would love some opinions on that

Unless you need more power than it provides, these measurements (as with L7 audio's before) already show you can't hear any difference between this and any other headphone amp. Within its power capabilities it absolutely exceeds in accuracy any limit of human hearing. The same is true of many dongles, interfaces and amps reviewed here, but this one has the benefit of coming "free" with your laptop

 

[Drengur]

Computer audio has come a long way. I miss not needing any monitoring software on my PC, as 20 years ago I could hear what my HDD and GPU were doing via the headphone jack.

 

[DrSpan]

Unless you need more power than it provides, these measurements (as with L7 audio's before) already show you can't hear any difference between this and any other headphone amp. Within its power capabilities it absolutely exceeds in accuracy any limit of human hearing. The same is true of many dongles, interfaces and amps reviewed here, but this one has the benefit of coming "free" with your laptop

"Unless you need more power than it provides“- No i can already blow my ears off level wise just with my m1 mbp output, don`t need more power.

I am trying to define what i mean but somehow seem to be failing tbh
I am not refering to the audio quality of how good the line signal that comes out is- i don`t mean „how good is the headphone amp when i listen to the signal though headphones“
but if you work on audio in your computer -> do a mixdown->Master itb-> render the file and export it= what differences if any are there between the Laptop`s own interface D/A converter and an D/A converter costing thousands, random example a
Burl Audio B2 Bomber ADC or
Dangerous Music Convert AD+ or

Lynx Studio Hilo 2 USB. etc etc


I mean with actual measurements putting the two exported files next to each other and measuring the differences if any. The biggest advantage-argument i am aware of in expensive D/A converters in studios is how they handle hot signals and intersample peaks and S/NR but i would find it unbelievably interesting to see actual numbers through actual measuring equipment like the ones Amir and other good people bless us with on this website .
I lack the technical understanding and also equipment in order to perform such a task.

 

[Chr1]

Slightly off topic, but I am curious...

Does this mean that in this case, an external DAC would not really be necessary when outputing to a power amp (and speakers) also?

Thanks.

 

[DrSpan]

Slightly off topic, but I am curious...

Does this mean that in this case, an external DAC would not really be necessary when outputing to a power amp (and speakers) also?

Thanks.

I don`t think the output has enough omph to drive every amp there is. A dedicated external dac with dedicated Line Outputs might be better.
This could be an external audio interface for example which except some cheap ones or very minimalistic ones always have a Monitor Line Output
However i will wait for the experts to weigh in as i cant back this up with numbers

 

[AudioSceptic]

Slightly off topic, but I am curious...

Does this mean that in this case, an external DAC would not really be necessary when outputing to a power amp (and speakers) also?

Thanks.

I believe so, as long as 1.0 V is sufficient (and I think it's *on* topic).