On the Distortion of Cirrus Logic CS431xx-Based Devices: A Comparative Review

[jkim]

To me, the kings as of today, are these ultra cheap DSP dongles, I have a couple and are more efficient than the CS ones, got them for <10 EUR, and I couldn't care less about the extra distortion dB.
But then, a thread like this could be made about their apps. Not a problem for me, they are so cheap I can have one EQed for each headphone I own and forget forever.

Depending on the use case, this is a valid point. The Hi-Max dongle based on CB1200AU was measured at RAA. Its performance is not bad at all. The Fiio JA11 and JCally JM12 look very interesting, too.

What about Hiby FC3 non display version? it has ES9281PRO, price is close to 30€ on Ali. I know the review here only has the display version with a different chip, though I've read that actually the difference between this and ES9281ACPRO is not much in practice, but is the non display version good enough to consider it over a Cirrus Logic chip without its shenanigans? what do you think?
https://es.aliexpress.com/item/1005002003179693.html

I missed that. The FC3 (without display) was also tested at RAA. The Hiby FC3 and FC3 w/ display should be good enough and can provide healthy unbalanced output over 1 V for low-impedance headphones.

 

[mc.god]

To me, the kings as of today, are these ultra cheap DSP dongles, I have a couple and are more efficient than the CS ones, got them for <10 EUR, and I couldn't care less about the extra distortion dB.

I got 2 Hi-max CB1200 recently, they are nice toy to play with and for the incredible price, but not very good for very sensitive iems, both produce audible hissing with TRN MT1 for example. OK with other less sensitive iems.

 

[multiformous]

The Fiio JA11 and JCally JM12 look very interesting, too.

If you end up getting a JCally JM12, I'd love to see a suite of measurements for it! The interesting thing is that it can be flashed with Fiio JA11 firmware as well as from TinHiFi, but of course once this is done the stock JM12 firmware is gone. I must say, JM12 with JA11 firmware has got to be one of the noisiest dongles I've ever heard... performs absolutely horribly in all these test clips (including Windows chimes) and I swear I hear it in real content (unlike issues highlighted with CS-dongles). The on-board PEQ is neat, though.

 

[jkim]

If you end up getting a JCally JM12, I'd love to see a suite of measurements for it! The interesting thing is that it can be flashed with Fiio JA11 firmware as well as from TinHiFi, but of course once this is done the stock JM12 firmware is gone. I must say, JM12 with JA11 firmware has got to be one of the noisiest dongles I've ever heard... performs absolutely horribly in all these test clips (including Windows chimes) and I swear I hear it in real content (unlike issues highlighted with CS-dongles). The on-board PEQ is neat, though.

So, the JM12 with stock or TinHiFi firmware works fine for you without noticeable noise. Right?

 

[Fleuch]

One thing that hits the eye when looking at the spreadsheet is that six of the DACs known not to produce the "Cirrus Hump" distortion pair the CS431xx chips with the Savitech SA9312L USB bridge. Although the same USB bridge is used on three of the DACs known to produce the "Cirrus Hump" it is reasonable to suggest that the Savitech chip is the USB bridge to have better compatibility with the CS431xx chips. Also all the performant DACs do not include PEQ support. Perhaps it is the interface with the peripheral components that is accentuating a design anomaly within the CS431xx circuitry itself. Only Cirrus can offer an explanation.

 

[multiformous]

So, the JM12 with stock or TinHiFi firmware works fine for you without noticeable noise. Right?

Last I checked, the TinHifi firmware didn't have any fade in / delay when starting playback, and is much less crunchy-sounding on the polarity test clip / CS-distortion clip. But it's not perfectly clean, and I think it has a higher noise floor / audible hiss. And for whatever reason, TinHifi firmware doesn't support 44.1kHz sampling rates. JA11 firmware is definitely noisier on those test clips, but supports 44.1khz. So neither firmware is ideal (didn't test it with stock firmware, and can't without buying another), and for me JM12 is not really useable.

 

[jkim]

A poll has been created to see people's take on this issue. Some of us must be curious about that.

 

[neutroncode]

This is… interesting
These dongles are cheap thingies, but these issues are chip-related, now I wonder if the DAPs, desktop DACs, using CS431xx chips have the exact same issues?

In my earlier tests CS-based DAPs incorporating 2x CS43198 or 4x CS43131 (2 per channel) exhibited the same issue, including cheap Apple USB-C dongle. So the cost of the implementation does not matter, including the complexity, the reason is in the specific performance of CS DAC chips.

For my tests I used file with beats generated by Audacity, to emulate more or less normal audio content. And besides distortions noticed by @danadam (in the beginning of segment when sound wave gains amplitude rapidly) there is also an issue with a noise floor which is high when audio wave gained amplitude and then descending when amplitude of the audio wave fades. This behavior is depicted in this my post.

@jkim could you please try my Beats test file on CS-based devices which do not exhibit "Cirrus Hump", at least on one of them, and check the Spectrogram for both effects (distortion + noise with gradient effect).

 

[jkim]

In my earlier tests CS-based DAPs incorporating 2x CS43198 or 4x CS43131 (2 per channel) exhibited the same issue, including cheap Apple USB-C dongle. So the cost of the implementation does not matter, including the complexity, the reason is in the specific performance of CS DAC chips.

For my tests I used file with beats generated by Audacity, to emulate more or less normal audio content. And besides distortions noticed by @danadam (in the beginning of segment when sound wave gains amplitude rapidly) there is also an issue with a noise floor which is high when audio wave gained amplitude and then descending when amplitude of the audio wave fades. This behavior is depicted in this my post.

@jkim could you please try my Beats test file on CS-based devices which do not exhibit "Cirrus Hump", at least on one of them, and check the Spectrogram for both effects (distortion + noise with gradient effect).

In fact, when I tested the devices for this review, I considered using your test file but ended up not using it. The reason is that your results show not only the distortion that occurs when the signal spikes but also the CS431xx's level-dependent noise shaping effect which appears above 50 kHz. So, different effects are mixed together. I was not interested in noise shaping effects above 50 kHz. And the distortion due to DRE operation can be illustrated in better, enlightening manner using @danadam's test file.

I just ran your file on the JCally JM20 (w/o Cirrus humps) and Tanchjim Stargate (w/ Cirrus humps). The file was played at -16 dB peak and the recordings were calibrated to -1 dB peak.

JCally JM20 (with no Cirrus Hump):

Tanchjim Stargate II (with Cirrus Humps):

Except that the distortion occurs more frequently with the Stargate II than with the JM20, this test file does not as effectively show differences in their distortion behavior as the other test file used in the review.

You must want to see above 50 kHz. Right? Here you go.

JCally JM20 (with no Cirrus Hump):

Tanchjim Stargate II (with Cirrus Humps):

Interestingly, even level-dependent noise shaping seems to behave differently b/w the two (groups of) devices.

 

[VAV]

@jkim can you please check how the 4.4mm connector is wired inside devices with this connector (e.g., common ground)? Is the 4.4mm output balanced and differential on any of them?

 

[danadam]

The reason is that your results show not only the distortion that occurs when the signal spikes but also the CS431xx's level-dependent noise shaping effect which appears above 50 kHz. So, it is mixed with different effects. I am not interested in noise shaping effects above 50 kHz.

To tell the truth my file also can show noise shaping effect, it's just a matter of ADC sampling rate. Here's JM20 playing at -10 dB and ADI-2 Pro capturing at 192 kHz:

 

[jkim]

To tell the truth my file also can show noise shaping effect, it's just a matter of ADC sampling rate. Here's JM20 playing at -10 dB and ADI-2 Pro capturing at 192 kHz:
View attachment 453927

Of course, I knew that. What I meant above was about the effectiveness of your test file in showing the distortion behavior distinctively b/w the two groups of devices. And I deliberately did not show frequencies above 48 kHz. It was not just because your spectrograms showed only up to 48 kHz.

 

[jkim]

@jkim can you please check how the 4.4mm connector is wired inside devices with this connector (e.g., common ground)? Is the 4.4mm output balanced and differential on any of them?

Which device you mean? All devices with 4.4mm outputs in this test group are truly balanced because that is the only way to support output voltage close to 4 Vrms.

 

[neutroncode]

In fact, when I tested the devices for this review, I considered using your test file but ended up not using it. The reason is that your results show not only the distortion that occurs when the signal spikes but also the CS431xx's level-dependent noise shaping effect which appears above 50 kHz.

Thank you very much for capturing some results. I was interested to see whether this noise shaping behavior is consistent across CS DAC chips. To my view, the observed distortion (slight corruption of the audio wave on amplitude rise) is related to the noise shaping processing logic of the chip. 50+ kHz range is still important, although not audible, if user is using USB DAC for tasks not directly related to generic music listening, for example for some experiments. How easy >50 kHz noise can couple into audible range if not filtered by analog filter is another question but hw design using CS DACs should probably always low-pass analog output >50 kHz.

With this noise performance + distortion CS chips fall probably to the Lo-Fi segment, rather than to the modern Hi-Fi where we would like to see clear, noise free spectrogram up to at least 192 kHz (i.e. with 384 kHz PCM). It is something that you will never see with ESS or AKM DAC chips.

 

[Amperage]

With this noise performance + distortion CS chips fall probably to the Lo-Fi segment, rather than to the modern Hi-Fi where we would like to see clear, noise free spectrogram up to at least 192 kHz (i.e. with 384 kHz PCM). It is something that you will never see with ESS or AKM DAC chips.

I'm a bit conflicted by this. We're looking at a DAC chip family with previous SoA or close to SoA performance ratings achieved on several tested products after their run through the standard battery of tests, that is apparently now falling into the "Lo-Fi" category not because of a moving threshold in technology, but because of a specific behavior that was previously undetected - both by earlier tests and ears.

As these chips are present in large variety of products in different price ranges which have been in the market for a while, with positive reviews & feedback from a (potentially) large customer base; I'm very curious as to what's next.

Once again, thank you to all those involved in investigating and discovering this. And I really hope this has a fix.

 

[jkim]

To my view, the observed distortion (slight corruption of the audio wave on amplitude rise) is related to the noise shaping processing logic of the chip.

In my view, the distortion and noise shaper are not directly related. My hunch is that to achieve their design goal---high, measured performance and low power consumption for use in portable devices---, the designers had to make some compromises within given limits:

• Aggressive noise shaping was incorporated in order to compensate for relatively high-noise analog circuits (due to thermal noise).
• Low master clock rate support resulted in its essentially 96 kHz sample rate---choosing any FSR beyond this rate, even if supported, has no practical effect due to the > 50 kHz noise.
• Aggressive noise shaping was not very suitable for low-level signals because its induced ultrasonic noise was too high relative to input signals.
• DRE was utilized to deal with low-level signals while reducing the intensity of noise shaping.
• The unwanted (probably expected but considered fine) distortion was a byproduct of DRE operation.

After all, the design must have been somewhat complex. I believe the designer may have thought the end product would be an optimal tradeoff for the chip's proposed applications.

How easy >50 kHz noise can couple into audible range if not filtered by analog filter is another question but hw design using CS DACs should probably always low-pass analog output >50 kHz.

A good point.

With this noise performance + distortion CS chips fall probably to the Lo-Fi segment, rather than to the modern Hi-Fi where we would like to see clear, noise free spectrogram up to at least 192 kHz (i.e. with 384 kHz PCM). It is something that you will never see with ESS or AKM DAC chips.

Calling it "Lo-Fi" is too harsh. I think the end result is audibly fine. I would still prefer the devices with mitigated distortion from less aggressive DRE, though, like the JCally JM20 or JM20 Max.

 

[Amperage]

Should this be considered enough to warrant at least an acknowledgement and review from OEM's and chip designer?

 

[MCH]

In these cases is when I whish more IC manufacturers had a forum like the ones of Texas Instruments or Analog Devices. The secretiveness of the audio world is ridiculous, like if they were doing guided missiles or something.

I think a OEM that sells products that barely cost more than the BOM to unknown or semi unknown companies like jcally et al., could not care less.

 

[Amperage]

In these cases is when I whish more IC manufacturers had a forum like the ones of Texas Instruments or Analog Devices. The secretiveness of the audio world is ridiculous, like if they were doing guided missiles or something.

I think a OEM that sells products that barely cost more than the BOM to unknown or semi unknown companies like jcally et al., could not care less.

Right. But then there are companies with products above those prices with the promise of SoA performance that could be and are affected.

In their defense, they apparently comply with the information provided by data sheets and prior tests that didn't uncover this issue. But now that it's been discovered, their response and Cirrus Logic's will be telling.

 

[HissingFree]

"This DAC contained in our product has a huge flaw that is in the hardware itself and we cannot fix it, sorry guys"