Test of "Cirrus Hump" Distortion of Apple MacBook Air Audio Output

[jkim]

After writing my recent review, "On the Distortion of Cirrus Logic CS431xx-Based Devices: A Comparative Review", I wondered if the same anomaly exists in other electronic devices that rely on Cirrus Logic D/A converter (DAC) chips, in particular Apple MacBook products that still provide analog audio jacks. Cirrus Logic has long been an OEM partner of Apple for DAC solutions. It seems that all MacBook Air/Pro products have been using the same DAC SoC supplied by Cirrus Logic for quite a while (see here).

Apple products have long been known to provide excellent audio quality. MacBooks are no exception. Measured performance of MacBook's audio jack is really nice for a laptop, which was shown in L7 Audio Lab's measurements of 2021 MacBook M1 Max. Also, an ASR member's recent review of 2024 MacBook Air's audio output showed essentially the same great performance. However, as soon as I found out these devices are based on Cirrus Logic DAC chips, I couldn't help but test a MacBook at my home. Coincidently, it is the 2024 MacBook Air 15" M4.

Measurement Setup​

• AD converter: E1DA Cosmos ADCiso Grade A and Cosmos Scaler as a buffer.
• Software: Room EQ Wizard (v5.4 beta 85).

Results​

1 kHz SINAD is essentially the same as measured by others:

A sweep of 1 kHz sinusoids across different output levels shows excellent performance, too:

In the above sweep test, there are very interesting observations that can be seen when we focus on the devices' noise performance. Let's see the absolute RMS noise levels (not relative to fundamental tones):

For most DACs that are not based on Cirrus Logic chips, their absolute noise level does not change much depending on the signal intensity. Usually, noise variation across output levels is less than 1 dB. However, with the dynamic range enhancement (DRE) implemented in Cirrus Logic chips, noise is a function of input signal strength. Noise is relatively high when DRE is turned off with strong signals and it is low when DRE is in full action with weak signals (see my review for more information). In the case of MacBook, the effect of DRE is very strong (i.e., aggressive). When the signal is strong (1 kHz sine tone > -8 dBFS), DRE is off and hence the noise is at its highest level. But as the signal weakens below -8 dBFS, DRE increases its effect gradually (by trading analog gain for digital gain). With the help of DRE, MacBook can achieve a nearly 16 dB lower noise floor! In contrast, the JCally JM20 Max with the CS43131 chip exhibits quite low noise even without DRE and starts lowering noise with DRE from below -12 dB. The total measured reduction of noise is only 5.5 dB---this low value is likely due to ADC's measurement limit, though.

Is there any other consequence of DRE? We do not know precisely. But in my review I conjectured that the peculiar distortion behavior, dubbed "Cirrus hump," observed from CS431xx-based devices is very likely a byproduct of DRE. So, we may predict similar distortion from MacBook's audio output which is based on Cirrus Logic's OEM chip. Let's first run some multitone tests at different signal intensities:

Do you notice something wrong in the above results? The noise & distortion response of the 32-tone signal that is 11.8 dB lower (in blue) than the max level should have been placed between the response of the stronger signal (in green; 0.8 dB lower than max) and that of the weaker signal (in orange; 38.8 dB lower). In fact, it should've been much closer to the stronger signal response (green). That is not the case here, meaning the signal is corrupted to some extent.

A sweep of 32-tone tests vs. output level was performed:

As expected, MacBook's audio output is affected by the same kind of "Cirrus hump" distortion in even broader a range of signal levels.

Essentially the same story goes with much simpler dual-tone test signals (TDFD IMD in Room EQ Wizard):

Conclusion​

This review showed that 2024 MacBook Air's audio output produces the same distinctive distortion behavior as known to be associated with devices based on Cirrus Logic CS431xx DAC chips. It is suspected that MacBook products recently released (within several years) are all affected by the same artifacts. As said in my review on the distortion of CS431xx-based devices, it is tricky to determine the audibility of this "Cirrus hump" in real audio content. Until a controlled experimental study with human subjects is conducted (which is unlikely to happen), readers may make judgment based on measurements data. I would like to take some excerpts from my review that reflect my position on this issue:

It is difficult to tell how clearly audible these measured distortions would be in real audio content. But given the fact that the problem is observed even in a simple dual-tone test reported above, it is no wonder Roman at RAA was able to easily spot a movie soundtrack to demonstrate the distortion. Note that the website provides multiple recordings of the same track played at different levels, which are then level-matched for the higher % distortion to be heard easily. Other than the most audible cases, it takes trained ears familiar with distorting sound to hear it. It is not clear "clicking" or severe clipping distortion. And the original track already contains quite high background noise. The distortion sounds like occasional crunch in the midst of fluctuating noise floors. In comparison, through an unaffected device, the background noise is not much fluctuating without crunch.

Running all the tests including those in Part II has shaped my take on this issue. Of course, my perspective on this hobby has also influenced it. Listeners may not easily notice the distortion examined in this review since it is not severe clipping. However, as indicated by the testing conditions (i.e., simple multitone, even dual-tone, signals in a wide range of low amplitude levels), this distortion occurs not just in very limited, artificial situations. It should occur frequently in real audio content, although it is another matter whether a listener can hear it or not. Human auditory perception is gullible in isolated situations like hearing distortion in a controlled condition, but at the same time it is subtle and sensitive holistically. For this reason, I am still not comfortable.

 

[AnalogSteph]

In the case of MacBook, the effect of DRE is very strong (i.e., aggressive).

This may simply be a result of the DAC being a fairly low-power affair of limited dynamic range, thus making the effect of DRE a lot more obvious. If we're being realistic, a 107 dB dynamic range is perfectly adequate for human hearing.

The fundamental design of these chips is certainly sound, it's just a shame they suffer from these artifacts.

 

[jkim]

This may simply be a result of the DAC being a fairly low-power affair of limited dynamic range, thus making the effect of DRE a lot more obvious. If we're being realistic, a 107 dB dynamic range is perfectly adequate for human hearing.

The fundamental design of these chips is certainly sound, it's just a shame they suffer from these artifacts.

As the supplier of Apple mobile devices' DAC chips for a long time (since 2002 according to this news article), Cirrus Logic's primary design goal must have been low power consumption. The analog gain part of the chips must not have been low-noise circuits due to thermal noise.

DRE must have been a great solution in this case. My hunch is that Cirrus Logic designers did not expect this multitone-induced distortion when designing the chip's DRE operation logic. DRE must involve determining the amplitude of incoming signals to decide how large digital headroom exists. I guess the logic is using a very tight threshold for this, which causes the signal, when amplified digitally, to be borderline saturated. The problem might be easily solved by slightly relaxing such a threshold of signal amplitude in the DRE operation logic.

 

[AnalogSteph]

As the supplier of Apple mobile devices' DAC chips for a long time (since 2002 according to this news article), Cirrus Logic's primary design goal must have been low power consumption. The analog gain part of the chips must not have been low-noise circuits due to thermal noise.

I mean, analog stage noise is clearly low, or else DRE wouldn't work as well as it does. It's just that any DAC with a bunch of "real" instantaneous dynamic range needs to either run at relatively high analog supplies (not conductive to small process nodes), or be very "wide" when you're stuck at low supplies like 1.8 V, so you can make up for it with more current instead. Either way, more dynamic range at the DAC stage is directly reflected in higher power consumption. Hence why mobile codecs and SoCs traditionally tended to use DACs of rather outdated specs plus PGA volume control (e.g. the AS3525/43 as used in the Sandisk Sansa Clip/ClipV2/Clip+ would accept up to 48 kHz at 18 bits like it's 1994). DRE with Class H analog supplies basically is a much more advanced, automated version of that approach.

DRE must involve determining the amplitude of incoming signals to decide how large digital headroom exists. I guess the logic is using a very tight threshold for this, which causes the signal, when amplified digitally, to be borderline saturated. The problem might be easily solved by slightly relaxing such a threshold of signal amplitude in the DRE operation logic.

One would hope so. After all, AKM does something similar (AK4377/A) without such hiccups.

 

[stefanovesa]

Interesting that JCally JM20 Max is much less affected, do you think that the Cirrus Logic DAC used by Apple is a old one or is due to a different programming?

 

[jkim]

Interesting that JCally JM20 Max is much less affected, do you think that the Cirrus Logic DAC used by Apple is a old one or is due to a different programming?

We do not know. Most devices based on CS43131/CS43198 are affected. The JCally JM20 Max is one of the small number of exceptions.