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

[tccalvin]

did I do this correctly?

In order to get comparable results with the CMaj test, you need frequency on the y axis and time on the x axis. In the case of Audacity, amplitude is color-coded.

Graphs like the one you posted only show a single moment in time and are not well suited to capture quick transients like clicks. The phenomenon is almost impossible to capture with a screenshot. In a video, you would detect the clicks as the noise in the FFT graph "bobbing" or "jumping" up and down. But even then, you would be limited by the software's frame rate.

Watch what happens here at 0:25.

AFAIU the other test is a level sweep of a 32-tone multitone. Not 100% sure how to set it up.

 

[IsaacOscar]

In order to get comparable results with the CMaj test, you need frequency on the y axis and time on the x axis. In the case of Audacity, amplitude is color-coded.

Graphs like the one you posted only show a single moment in time and are not well suited to capture quick transients like clicks. The phenomenon is almost impossible to capture with a screenshot. In a video, you would detect the clicks as the noise in the FFT graph "bobbing" or "jumping" up and down. But even then, you would be limited by the software's frame rate.

Watch what happens here at 0:25.

AFAIU the other test is a level sweep of a 32-tone multitone. Not 100% sure how to set it up.

Yeah I new I did it wrong, I was just hoping the spectrum results were still useful (as they don't show spikes that the Tanchjim had).
Thanks, I've used a different program that was much easier to use now (albeit it is a beta version with some bugs...) and updated my post.

 

[tccalvin]

definitely no hump

Hard to tell since the noise level seems to be very high in those measurements.

However, I can see faint "click" lines in the CMaj test that indicate a pattern inconsistent with the non-hump variation of the Cirrus DACs.

My bet would be on yes-hump, unfortunately. I hope I'm wrong though!

 

[IsaacOscar]

Hard to tell since the noise level seems to be very high in those measurements.

So maybe my DAC has been too noizy for the hump to make a difference

However, I can see faint "click" lines in the CMaj test that indicate a pattern inconsistent with the non-hump variation of the Cirrus DACs.

Using the same software, here is what the "Tanchjim_Stargate2_JCally_JM20Pro_CMaj" file looks like, notice the clear vertical lines that my spectrogram doesn't have.

My bet would be on yes-hump, unfortunately. I hope I'm wrong though!

I've already ordered a new DAC (my current one doesn't have PEQ or enough power), so I only measured it out of curiosity (has all the music I've been listening to been wrong ???)

I'm happy to gift my DAC to someone who is able to do proper testing and can pay for postage from Australia....

 

[tccalvin]

notice the clear vertical lines that my spectrogram doesn't have.

Yep, but you will also notice that the signal in that file is much louder (see the reddish color of the spectrogram as opposed to the orangey one in the Xiaomi measurement?).

This means that a portion of the signal is getting buried in noise. Hard to tell if the bottleneck is the DAC or the ADC.

To make sure, increase the DAC volume as much as possible and adjust the ADC gain (if available) so that you can get close to 0dBFS without clipping. The lines (clicks) will be much more apparent that way.

Edit: actually, now that you posted a comparison graph I can just barely make out the vertical stripes on your own graph as well! (Cirrus Hump confirmed)

has all the music I've been listening to been wrong ???

Of course not! We're talking about barely audible defects here anyway. People used to listen to cassette tapes which had something like 6 bits of dynamic range and they were absolutely fine with it. You don't need state-of-the-art performance to enjoy music. Even critical listening will be just fine.

Then again, if you're obsessive like me you might drop another 200 just for the feeling of knowing that you're listening to a well-measuring device. lol

 

[Sokel]

So maybe my DAC has been too noizy for the hump to make a difference

Using the same software, here is what the "Tanchjim_Stargate2_JCally_JM20Pro_CMaj" file looks like, notice the clear vertical lines that my spectrogram doesn't have.View attachment 506557


I've already ordered a new DAC (my current one doesn't have PEQ or enough power), so I only measured it out of curiosity (has all the music I've been listening to been wrong ???)

I'm happy to gift my DAC to someone who is able to do proper testing and can pay for postage from Australia....

We're lucky to have the author and a very big thread about Multitone Analyzer here:

Beta Test: Multitone Loopback Analyzer software

This is a software beta-test for IMD and THD aficionados with an intense desire for multi-tone audio testing :) The start of the documentation and the download website is here: https://distortaudio.org/multitone.html (for 64-bit Windows) Multi-tone Analyzer (MA) was more of an experiment than...

www.audiosciencereview.com

Your measurements seem to have a high noise floor which makes sense since it's through the line out of the PC.
It would (maybe) help if there was ASIO drivers or if you used WASAPI exclusive.

 

[IsaacOscar]

This means that a portion of the signal is getting buried in noise. Hard to tell if the bottleneck is the DAC or the ADC.

Probably the ADC (it's just the line in port on my B760M DS3H AX, which has an ALC897 and ES9260Q; I deliberately bought a cheap Motherboard as a good external DAC is cheaper than a Motherboard with a good internal one...). I'll try it with other ADCs

To make sure, increase the DAC volume as much as possible and adjust the ADC gain (if available) so that you can get close to 0dBFS without clipping. The lines (clicks) will be much more apparent that way.

I thought the file had to be played at -15dBFs? (I had the DAC at 100%, but I can try to increase the ADC volume, higher volumes where producing clipping.

Of course not! We're talking about barely audible defects here anyway. People used to listen to cassette tapes which had something like 6 bits of dynamic range and they were absolutely fine with it.

I very much did not like listening to cassette tapes....

 

[IsaacOscar]

Your measurements seem to have a high noise floor which makes sense since it's through the line out of the PC.
It would (maybe) help if there was ASIO drivers or if you used WASAPI exclusive.

Sadly I have no ASIO drivers, but I did choose the WASAPI option. I think my Motherboard just sucks (which is why I use an external DAC in tje first place)

 

[MaxwellsEq]

I've already ordered a new DAC (my current one doesn't have PEQ or enough power), so I only measured it out of curiosity (has all the music I've been listening to been wrong ???)

You probably shouldn't worry too much.

I suspect both you and I would struggle to tell the difference in a blind test between a good 16 bit DAC and a good 14 bit DAC. When DACs are linear and using proper reconstruction filters, they sound mostly the same. The problem with this Cirrus Logic CS431xx "design mistake", is it is audible to everyone, unless DRE is disabled.

 

[pioterd]

@jkim please add to list of devices free from these distortion effects (I did many tests according to this site, and no problems):

• Shanling UA Mini (1xCS43131 + 1xSGM 8262-2)
• Questyle M15C (1xCS43198 + SiP CMA amplifiers)

 

[scruffy1]

i've recently procured an smsl dl100 and am here after tccalvin's comment in the measurement thread for that device to try and understand the issue

having done some other reading, the "distortion" seems related to dle, which on this device is intrinsic and cannot be "turned off"

other than numerous comments about the "c major test" i'm none the wiser and wonder if i'm not looking at an oscilloscope, or listening specifically to an endless c major, what is the impact in the real world listening to complex nuanced music with a truck tonne of augmented and suspended chords with vocals and percussion : should i expect to have my pleasure diminished ? (see what i did there ?)

as a previous computer nerd, i understand the e-peen of maximal performance, but as a less sophisticated sound technician, is it really of huge consequence if there is distortion that i can't notice, especially employing 67yo ears with quite reasonable acuity for age, but >14000 is likely a mystery, and >12000 ain't what it was

 

[tccalvin]

should i expect to have my pleasure diminished ? (see what i did there ?)

You'll be just fine. (Saw what you did there)

 

[IsaacOscar]

what is the impact in the real world listening to complex nuanced music

The only actual piece of music anyone on this thread has heard a problem with is the "Song of the sisters" from the Dune soundtrack. So if you avoid listening to that, you should be good.

 

[scruffy1]

You'll be just fine. (Saw what you did there)

so, i presume this is more of a technical distress than an audible disaster

as said, my technical understanding is crap, but i am a discerning listener... does the use of multiple chips actually reduce this issue (hence them being applied in the dl100) because i was somewhat aware of the fact that the e70 uses 8 channel chip and the tone 1 is stereo, and there is some magic in resampling stereo input through 4 channels each - or i may have that all wrong

so the quad chips in the dl100 makes "sense" to improve final output, but again, via a process that is to all practical understanding, a miracle

The only actual piece of music anyone on this thread has heard a problem with is the "Song of the sisters" from the Dune soundtrack. So if you avoid listening to that, you should be good.

i'll do my best, although now that i know i will be compelled to see (hear) what happens

 

[tccalvin]

so, i presume this is more of a technical distress than an audible disaster

For sure.

does the use of multiple chips actually reduce this issue

We don't know yet. So far most measurements have been on single-chip dongles.

 

[Sokel]

as said, my technical understanding is crap, but i am a discerning listener... does the use of multiple chips actually reduce this issue

Always be suspicious when you hear about cascaded components like DACs and OPAs.
Their measurable performance (specially noise) can be very well "played" the same way we measure with cross-correlation averages for example which can have a staggering effect at the (measured) noise.

It depends on the design of course.

 

[pangcukaipang]

The only actual piece of music anyone on this thread has heard a problem with is the "Song of the sisters" from the Dune soundtrack. So if you avoid listening to that, you should be good.

Lol, I just bought Roseselsa RZ550 which uses ES9039Q2M. And guess what, it still couldn't play the intro "well" on my usual volume. The percussion sounds very clear on low volume and distorted if I up the volume.

The exact same thing with my DAP and my other DAC (both using dual CS43131).

So, I agree with you here, just avoid listening to that, and I'm good

 

[scruffy1]

Lol, I just bought Roseselsa RZ550 which uses ES9039Q2M. And guess what, it still couldn't play the intro "well" on my usual volume. The percussion sounds very clear on low volume and distorted if I up the volume.

The exact same thing with my DAP and my other DAC (both using dual CS43131).

So, I agree with you here, just avoid listening to that, and I'm good

wrong chip? (on the RZ550)

so maybe distortion is a perfect storm of chain components, and absent for other players ? i'm certain i wouldn't listen to "song of the sisters" on high rotation, but ymmv

 

[pioterd]

As for iFi products, I truly wonder how iFi as a brand is still surviving. The brand is not recognized as a reputable company here at ASR.

That's part of the reason why iFi is still doing well. ASR is only in small part objective, and science presented here have still many limitations. No need to take it too seriously in terms of "sound" and experience people have with it - very often it has nothing to do with "science", and that's good I think.

 

[scruffy1]

That's part of the reason why iFi is still doing well. ASR is only in small part objective, and science presented here have still many limitations. No need to take it too seriously in terms of "sound" and experience people have with it - very often it has nothing to do with "science", and that's good I think.

my 2c is that the secret with "hifi" is to get something that you enjoy, and then try hard not to hear anything much better (because after a certain point, the roi for "better sound" becomes much worse very fast)