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

[jkim]

Looks clean:
View attachment 453366

I suspect it has something to do with changing the amplitude of the signal, the fade-in. In attachment there is a similar file, but this time there's 2 second steady state:
View attachment 453368

The distortion happens only at the beginning of each segment. Here's JM20 at 0 dB, -10 dB and -20 dB:
View attachment 453370, View attachment 453372, View attachment 453373

and Samsung dongle at 0 dB and -10 dB:
View attachment 453374, View attachment 453375


Done already half a year ago :

JCALLY JM20 Headphone Dongle Review

CB5100 certainly yields interesting results if you want to build your own dongle or make a custom USB-C DSP cable for headphones. The tinkerer in me has awak.. oh, 40 bucks, nevermind.

www.audiosciencereview.com

Thanks! I will test the dongles that I have using your files when I get a chance. Basically, your tests indicate that the JM20 still exhibits a similar kind of distortion when a multitone signal's level is monotonically increasing. But based on my results, it can be said that other affected devices with "Cirrus humps" should produce distortion whenever they meet the signal (in a certain range of levels) including the case of the signal level being steady. Right?

In fact, your previous tests using the soundtrack in the RAA article (Dune Sketchbook Soundtrack - Song Of The Sister, Hans Zimmer) suggest what I just said. See your results below (copied from your previous posts).

Tanchjim Space (full, -6 dB, -12 dB, -18 dB):

JCally JM20 (full, -6 dB, -12 dB, -18 dB):

Samsung dongle (full, -6 dB, -12 dB, -18 dB):

Careful, comparative inspection of these three outcomes tells us:

• It is very likely that the Tanchjim Space produces the same "Cirrus Hump" distortion that my review showed.
• The JCally JM20 still shows hints of similar distortion, but it is much lower and less frequent.
• The Samsung does not produce the same kind of distortion but, as Amir's test showed, its noise performance is substantially worse.

Based on these tests alone, I would still rate the JM20's overall performance higher than the other two devices'.

I will post my recording tests of the dongles using these tracks.

BTW, what software did you use for recording and drawing spectrograms? ALSA on Linux?

 

[danadam]

But based on my results, it can be said that other affected devices with "Cirrus humps" should produce distortion whenever they meet the signal (in a certain range of levels) including the case of the signal level being steady. Right?

Yes, Tanchjim Space produces distortions throughout the file:
, ,

The Samsung does not produce the same kind of distortion but, as Amir's test showed, its noise performance is substantially worse.

For me it's not as bad. I guess I don't have the problem with truncation to 16-bits:

But it has other problems:

Some strange clipping with Samsung dongle

I'm posting this more like a curiosity rather than a problem to solve. In the attachment there are 2 files, 44.1 Khz sample rate, with a series of pulses in groups of 4: In "bw.sweep" each group has bandwidth 500 Hz wider than the previous one, starting from 500 Hz. All pulses are normalized...

www.audiosciencereview.com

BTW, what software did you use for recording and drawing spectrograms? ALSA on Linux?

ALSA (aplay) for playing and recording, SoX for generating the files and spectrograms.

 

[jkim]

Yes, Tanchjim Space produces distortions throughout the file:
View attachment 453434, View attachment 453435, View attachment 453436

Excellent. By the way, I have a question. The reason why you came up with this test as you posted in #37 was because of my review. Right? In particular, if these distinct characteristics hold b/w all affected and unaffected devices, the findings strongly suggest that DRE is behind the distortion. See what I said about the difference in DRE operation of the unaffected devices:

If the signal weakens incrementally, the noise level stays the same as at the starting point, meaning that the DRE logic does not increase its effect, until the signal finally reaches below -50 dBFS. But if the signal strengthens, the noise level adapts to it meaning that DRE decreases its effect (by trading digital gain for analog gain). This implies that DRE in unaffected devices makes less aggressive adjustments of its gain structure (combination of digital & analog gain).

The distortion may occur when the signal level increases and the digital gain needs to be reduced.

For me it's not as bad. I guess I don't have the problem with truncation to 16-bits:
View attachment 453437

What I meant was not Amir's Windows ASIO driver problem but the Samsung dongle's higher noise level shown in your spectrograms.

But it has other problems:

Some strange clipping with Samsung dongle

I'm posting this more like a curiosity rather than a problem to solve. In the attachment there are 2 files, 44.1 Khz sample rate, with a series of pulses in groups of 4: In "bw.sweep" each group has bandwidth 500 Hz wider than the previous one, starting from 500 Hz. All pulses are normalized...

www.audiosciencereview.com

Interesting. I guess all these low-cost dongles have issues due to various reasons.

ALSA (aplay) for playing and recording, SoX for generating the files and spectrograms.

Thanks for the info.

 

[danadam]

By the way, I have a question. The reason why you came up with this test as you posted in #37 was because of my review. Right?

Yes. I remembered that JM20 still showed some distortion on the Dune track. At another time someone also suggested the rumble polarity test from audiocheck.net and the distortion was there too. The rumble seemed like unsophisticated signal, though with <20Hz content, so I started to wonder if I can create a simple signal with only >20Hz content that would also show some distortion. Those are the files linked earlier.

the findings strongly suggest that DRE is behind the distortion.

Yes, sounds plausible.

What I meant was not Amir's Windows ASIO driver problem but the Samsung dongle's higher noise level shown in your spectrograms.

The reason might be it has max level <1V while the CS dongles have 2V. With CS dongles at -8 dB the signal levels are the same. Other than wider skirt around the signal, the difference in noise floor is only a few dB.
, ,

 

[Amperage]

Thanks @jkim for the thorough, detailed tests!

I understand all devices tested are dongle dacs. But I wonder: if this issue is related to the model's DRE implementation, then could it be potentially present in other devices with CS43131 and CS43198 chips that aren't dongle dacs as well?

If not, would it also have to do with some form of power management functionalities and/or instructions in the firmware implementation?

 

[jkim]

Yes. I remembered that JM20 still showed some distortion on the Dune track. At another time someone also suggested the rumble polarity test from audiocheck.net and the distortion was there too. The rumble seemed like unsophisticated signal, though with <20Hz content, so I started to wonder if I can create a simple signal with only >20Hz content that would also show some distortion. Those are the files linked earlier.

Great. May I use your test signal when I put additional contents in my review?

The reason might be it has max level <1V while the CS dongles have 2V. With CS dongles at -8 dB the signal levels are the same. Other than wider skirt around the signal, the difference in noise floor is only a few dB.
View attachment 453475, View attachment 453476, View attachment 453477

The Samsung's wide skirt around the signal is indicative of its jitter components, especially low-frequency random jitter.

 

[jkim]

Thanks @jkim for the thorough, detailed tests!

I understand all devices tested are dongle dacs. But I wonder: if this issue is related to the model's DRE implementation, then could it be potentially present in other devices with CS43131 and CS43198 chips that aren't dongle dacs as well?

If not, would it also have to do with some form of power management functionalities and/or instructions in the firmware implementation?

As discussed with @danadam, the cause of this distortion is very likely to be how DRE is implemented in the chip. So, it will occur whether it's a dongle or a desktop DAC. The reason why this issue is mostly associated with portable devices is simply because the CS431xx chips are frequently adopted in those device types.

 

[danadam]

Great. May I use your test signal when I put additional contents in my review?

Sure, go ahead.

 

[mc.god]

The Samsung's wide skirt around the signal is indicative of its jitter components, especially low-frequency random jitter.

This is likely due to the device working in usb sync mode, also CS43131 devices behave like this when paired wth an usb bridge working in sync mode, like the Ugreen HiFi Pro just to say

 

[jkim]

This is likely due to the device working in usb sync mode, also CS43131 devices behave like this when paired wth an usb bridge working in sync mode, like the Ugreen HiFi Pro just to say

Wow, in this case, it must involve severe jitter (both periodic and random types). Not sure in what mode the Samsung dongle was working in @danadam's measurements.

 

[mc.god]

Wow, in this case, it must involve severe jitter (both periodic and random types). Not sure in what mode the Samsung dongle was working in @danadam's measurements.

I suspect many super cheap dongles work in sync or adaptive mode, for example CX-Pro and Hi-Max. Don't know the hardware used in Samsung dongle, Apple dongle is surely sync from Amir review and given that it uses the CS46L41, the same chip used as bridge in the Ugreen.

 

[danadam]

I suspect many super cheap dongles work in sync or adaptive mode

From lsusb output:

262a:1048   12M    Adaptive       ODACrevB

05ac:110a   12M    Synchronous    Apple, Inc.                          USB-C to 3.5mm Headphone Jack Adapter
04e8:a04b   480M   Synchronous    Samsung Electronics                  Samsung Type-C to 3.5pi gender adapter
06cb:1595   480M   Synchronous    Synaptics                            CX31993+MAX97220 HiFiAudio --> JCally JM6

21b4:0082   12M    Asynchronous   AudioQuest                           AudioQuest DragonFly Red v1.0
2972:0047   480M   Asynchronous   GuangZhou FiiO Electronics Co.,Ltd   FiiO K3
05ac:1710   12M    Asynchronous   Apple, Inc.                          USB-C to Lightning Adapter
262a:18c8   480M   Asynchronous   Shenzhen CBHT Technology Co., Ltd    CS4131 HIFI Audio --> JCally JM20
2fc6:f070   480M   Asynchronous   TANCHJIM                             TANCHJIM-SPACE
2972:0102   480M   Asynchronous   FIIO                                 JadeAudio JA11

 

[jkim]

Thanks to @danadam's contribution, I updated the review with new tests by adding "Part II" there.

I believe we now know quite a lot about the distortion phenomenon of CS431xx-based devices!

 

[mc.god]

Thanks to @danadam's contribution, I updated the review with new tests by adding "Part II" there.

I believe we now know quite a lot about the distortion phenomenon of CS431xx-based devices!

Thanks for all the contributions, great addition to the original analysis!
What to say, I never had audible issues with all my CS dongles and they'll still maintain my favor, but having already 4 of them. given these results, If I will ever need to buy a new DAC i think I'll go with an AK or ESS based one

 

[Amperage]

Thanks for all the contributions, great addition to the original analysis!
What to say, I never had audible issues with all my CS dongles and they'll still maintain my favor, but having already 4 of them. given these results, If I will ever need to buy a new DAC i think I'll go with an AK or ESS based one

This is exactly where I'm at right now - except I only have one CS43198 device with which I'm also quite happy with (KA15). Given a future need, it's back to ESS/AKM and even TI, until we can see that this issue can be fixed (and is fixed) through firmware.

Is this a design feature from Cirrus that's failing to meet proper implementation at the manufacturer level? An issue with the literature or documentation for these chips, maybe?

 

[mc.god]

Is this a design feature from Cirrus that's failing to meet proper implementation at the manufacturer level? An issue with the literature or documentation for these chips, maybe?

Unless someone get in touch with Cirrus and their engineers I think it's unlikely we will ever know the real reasons, anyway the big strength of these CS chip is to make possible devices with generally great measurements and a more than adequate headphone driving capability at super affordable prices, and just for this they'll deserve my eternal gratitude.

 

[CedarX]

Unless someone get in touch with Cirrus and their engineers I think it's unlikely we will ever know the real reasons, anyway the big strength of these CS chip is to make possible devices with generally great measurements and a more than adequate headphone driving capability at super affordable prices, and just for this they'll deserve my eternal gratitude.

Agreed, we should not forget that in typical use, without specially crafted test tracks, these CS issues are most likely inaudible.
But that’s not an excuse for the OEMs who select and implement the chips: unless there is some tradeoff we’re not aware of, they should know better, and if an “advanced user” can find these issues using “amateur” measurement equipment (no negative connotation whatsoever), what excuses do the OEMs have?

 

[jkim]

anyway the big strength of these CS chip is to make possible devices with generally great measurements and a more than adequate headphone driving capability at super affordable prices, and just for this they'll deserve my eternal gratitude.

Agreed, we should not forget that in typical use, without specially crafted test tracks, these CS issues are most likely inaudible.
But that’s not an excuse for the OEMs who select and implement the chips: unless there is some tradeoff we’re not aware of, they should know better, and if an “advanced user” can find these issues using “amateur” measurement equipment (no negative connotation whatsoever), what excuses do the OEMs have?

I think what I wrote after adding Part II would serve as my reply to your comments:

Additional Remarks after writing Part II​

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.

But perhaps a more important consideration is that hi-fi consumers of these devices would enjoy listening to music through them believing that they are exceptionally transparent by modern standards, when in fact not. That is robbing them of big fun in this hobby!

Another serious problem is that these CS chips are dominating the portable DAC market. So much so that it is difficult to find a portable DAC/HP amp that does not use CS431xx and is at the same time nice on paper. Alternatives are rare. There are some products that adopt ES9039Q2M or ES9069Q (or previous gen ES9038Q2M) or AK4493SEQ. But they are either significantly more expensive (Fiio BTR17 or KA17) or with limited usability (E1DA 9039S supporting only 2.5mm balanced output), or paired with RT6863/SGM8261 op-amps which are not very good (Shanling UA2 Plus, UA3, UA5). If someone concerned with the distortion of CS431xx asks to recommend a reasonably priced portable DAC/HP amp, then ES9219-based devices come to mind (for now):

• Qudelix 5K: Now that the problem of CS431xx is known better, the Qudelix 5K is even more appealing. Sure, it's more expensive but still reasonably priced and supports Bluetooth as well as on-device parametric EQ (PEQ). What more would you ask for?
• Hiby FC4: If one does not need Bluetooth or PEQ, this seems to be a good choice.
• FIIO BTR15: A good device for the same reason as the Qudelix 5K is. But given the superior user interface and functionality of the 5K, I would not choose the BTR15 over the 5K.

 

[MCH]

I think what I wrote after adding Part II would serve as my reply to your comments:

Enter power consumption, which is very important imo, and everything becomes even more complicated.

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.

 

[HissingFree]

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