Neutron HiFi DAC V1 Review

[dmitrykos]

but if they've been measured here on ASR and they measure well then you're likely just imagining the differences

Unless USB DAC is based on nowadays popular Cirrus Logic DAC chip (CS43198, CS43131,...) with enforced Dynamic Range Enhancement (DRE) technique (using ESS name for it here) which fools test for the dynamic range and shows astonishing result. So basically instrumentally test is passed with excellent mark but in reality audio performance on real music medium can be average.

Human ear is an advanced multi-frequency sensor and our brain is a super-high speed analyzer. We do not have numeric representation of the sound when we hear it but instead we feel it. This feeling is the way we perceive sound and can feel differences based on perceiving of the whole musical content. We can like the result delivered on the basis of accumulated feelings of listened content or not. It is the basis of the listening tests and we should not underestimate them because brain can catch nuances which are not covered by the tests.

Back to CS case and dynamic content. There is an article in RAA (https://reference-audio-analyzer.pro/en/demo-cs131.php#gsc.tab=0) showing the audio defects in the Spectrogram on the recordings from DACs with CS43198 and CS43131 chips. I gave it a try and reproduced the same result, including with Apple USB-C 3.5mm dongle:



I then took DAP with 2xCS43198 and got similar result, as well as another DAP with 4xCS43131 (2 per channel). ESS DAC chip shows clean spectrum, as well as DAC based on AKM AK4490:

Neutron HiFi DAC V1



AK4490-based DAC



I created a simple rhythm track with beat sound using Audacity and played with CS-based DACs. The result was similar to RAA test, CS DAC chips always show the distortion.

Here is Spectrogram of the original material:



And here is the result of the recording from CS-based DAC:



Closer look at distortion revealed that dynamic transitions of analog and digital gains of CS chip are causing high-frequency distortions:



Here is the same wave produced by Neutron HiFi DAC V1 based on ES9219 DAC chip without DRE employed:



Although CS-based DACs have good measurement results, all of them are affected by CS-enforced DRE technique which is trying to improve the dynamic range. Measurement is done by 1 kHz Sine which is signal with constant power, therefore this technique will adapt DAC's output to it and improve dynamic range measurement result. Although it measured well, in reality on dynamic music content the result will be, to my personal view, average because technically it is a kind of DSP employing digital representation of signal and hardware. Volume transitions, although done fast by the chip, are causing high-frequency artifacts. In my example distortion is present on every beat, so it will affect music and will be perceived by the human ear when you hear complex music content which is full of frequencies, amplitude and phase transitions.

 

[jkim]

Unless USB DAC is based on nowadays popular Cirrus Logic DAC chip (CS43198, CS43131,...) with enforced Dynamic Range Enhancement (DRE) technique (using ESS name for it here) which fools test for the dynamic range and shows astonishing result. So basically instrumentally test is passed with excellent mark but in reality audio performance on real music medium can be average.

Human ear is an advanced multi-frequency sensor and our brain is a super-high speed analyzer. We do not have numeric representation of the sound when we hear it but instead we feel it. This feeling is the way we perceive sound and can feel differences based on perceiving of the whole musical content. We can like the result delivered on the basis of accumulated feelings of listened content or not. It is the basis of the listening tests and we should not underestimate them because brain can catch nuances which are not covered by the tests.

Back to CS case and dynamic content. There is an article in RAA (https://reference-audio-analyzer.pro/en/demo-cs131.php#gsc.tab=0) showing the audio defects in the Spectrogram on the recordings from DACs with CS43198 and CS43131 chips. I gave it a try and reproduced the same result, including with Apple USB-C 3.5mm dongle:

View attachment 447013

I then took DAP with 2xCS43198 and got similar result, as well as another DAP with 4xCS43131 (2 per channel). ESS DAC chip shows clean spectrum, as well as DAC based on AKM AK4490:

Neutron HiFi DAC V1

View attachment 447015

AK4490-based DAC

View attachment 447014

I created a simple rhythm track with beat sound using Audacity and played with CS-based DACs. The result was similar to RAA test, CS DAC chips always show the distortion.

Here is Spectrogram of the original material:

View attachment 447025

And here is the result of the recording from CS-based DAC:

View attachment 447024

Closer look at distortion revealed that dynamic transitions of analog and digital gains of CS chip are causing high-frequency distortions:

View attachment 447017

Here is the same wave produced by Neutron HiFi DAC V1 based on ES9219 DAC chip without DRE employed:

View attachment 447018

Although CS-based DACs have good measurement results, all of them are affected by CS-enforced DRE technique which is trying to improve the dynamic range. Measurement is done by 1 kHz Sine which is signal with constant power, therefore this technique will adapt DAC's output to it and improve dynamic range measurement result. Although it measured well, in reality on dynamic music content the result will be, to my personal view, average because technically it is a kind of DSP employing digital representation of signal and hardware. Volume transitions, although done fast by the chip, are causing high-frequency artifacts. In my example distortion is present on every beat, so it will affect music and will be perceived by the human ear when you hear complex music content which is full of frequencies, amplitude and phase transitions.

Thanks for sharing your test results. The behavior of CS431xx has been discussed quite a lot and measured by some, like Roman the RAA founder.

BUT exactly what mechanism in the chip design causes this behavior is still a mystery. You attribute it to DRE---as you said CS never gave a name to this technique. How CS designed it, however, is not identical to DRE. In my view, artificially increasing AES17 DR measurements is not its sole purpose. That is why I would rather call this technique "Adaptive Signal-to-Noise Ratio Enhancement." See my measurements here (see the "Additional Analysis" in particular) and discussion here.

Anyway, not being the designer of the chips, we can only make an inference about what mechanism would cause what we observe. The RAA article author states the Class H operation is responsible for the anomaly, and you say DRE must be the cause. Although the transition b/w the H and AB modes can be related to the transition b/w DRE levels in their specific implementation, they are essentially two different things. Digital noise shaping, which also adapts to signals in CS431xx, is yet another element that can affect the phenomena we observe.

Here is an interesting example. The JCally JM20 MAX adopting CS43131 (my review here) is believed to operate only in Class AB mode---Class H mode can be disabled as an option according to the CS43131 datasheet. With this USB dongle, I was not able to reproduce the "crunchy" behavior in multitone tests suggested by the RAA article. But the sign of DRE operation is clear in this device as shown in my measurements.

Does this mean we can conclude that Class H mode operation---here I do not mean just the H/AB mode "transition" but also the implementation of the H mode itself---is solely responsible for the crunchy clipping? I do not think so. Again, this is just an inductive inference or reverse engineering instance based on limited observations. Still, even the JM20 MAX exhibits drastically different noise floors if artificial ultrasonic tones are included in test signals (see here).

Based on these measurements, can we say ALL CS431xx-based DACs should be audibly bad performers? I don't think so, either. As you know, measurements and perception are two different matters. To me, at least the JCALLY JM20 MAX sounds great. I do not hear any SQ difference between it and the current state-of-the-art, ES9039Q2M-based DACs, or any other ESS chip-based DACs (like the Qudelix 5K or Fiio BTR15 in which ES9219 is used).

 

[Eddie]

Again, this is just an inductive inference or reverse engineering instance based on limited observations.

At this point, I simply have to ask a blasphemous question:
wouldn't it be worth asking the engineers at Cirrus Logic about this issue? Perhaps there really is a flaw in the chip that they're not even aware of.

 

[dmitrykos]

we can only make an inference about what mechanism would cause what we observe. The RAA article author states the Class H operation is responsible for the anomaly, and you say DRE must be the cause. Although the transition b/w the H and AB modes can be related to the transition b/w DRE levels in their specific implementation, they are essentially two different things. Digital noise shaping, which can be made adaptive to signals, is yet another element that can affect the phenomena we observe.

I think there is no complex thing behind the observed behavior of CS DAC chips. To my view CS is really doing just the DRE thing which was pretty well explained by @IVX. In my previous post I showed Spectrogram of distortions recorded from Apple USB-C 3.5mm dongle which does not have additional buffer stage but the recording of the DAC with 2xCS43131 + output buffer stage with OPA shows more interesting result and confirms DRE in action.

Here is Spectrogram of beats with additional 100, 250, 500, 1500 Hz Sine waves which add some power to the sound and without those tones, just beats:



You can observe the stripped noise pattern. It is does not exists on DACs with ESS or AKM chips without DRE employed but with all CS DACs (CS43198, CS43131) I have it is there. Here is a closer look of just several beats with tone and without:



Besides already mentioned distortions we can observe the CS's DRE in action. Additional output buffer stage after CS DAC chip amplifies noise floor, so we can easily see how it works due to the specific gradient view of the noise:

1. When beat starts DRE logic sets analog gain to the maximum compensated by the digital gain to the opposite direction: this transition is causing observed hi-frequency distortion because the change from silent part was rapid
2. When beat fades DRE logic gradually lowers analog gain and compensates it with digital gain but in this case no distortion is happening, we see that the noise floor is gradually decreasing (fading out)

So with this test CS's SNR is like fade-in -> fade-out, fade-in -> fade-out, ...

ES9038 + output buffer stage or Neutron's ES9219 do not exhibit anything similar, the noise spectrum is stable and clean, beats do not affect either noise floor nor causing any distortions:



AKM AK4490 based USB DAC gives similar to ESS DAC chips good result.

To my view any periodic changes affect the quality of sound and the perception of the music. Moreover rapid transitions are causing high-frequency distortions which act like a parasitic DSP effect by adding some coloration to the music. We can't hear the effect of EQ when it is applied it just adds some colorization to the range of frequencies, so it works similarly to unintended DSP - it adds some specific colorization. While this DRE technique can be ok for some technical use for the consumer music application it is a no-go technology to my view as it affects audio content periodically on constant basis, periodicity depends solely on the audio content, so with depending on the music style and its dynamics the effect can be minor or major.

At this point, I simply have to ask a blasphemous question:
wouldn't it be worth asking the engineers at Cirrus Logic about this issue? Perhaps there really is a flaw in the chip that they're not even aware of.

To my view it is all about marketing and competition. Cirrus Logic wants to sell more chips, this technology allows them to do so. CS DAC chips give better measurement results and what we see? Majority of producers turned to CS DACs in order to win the race of measurements. In my measurements ESS and AKM DAC chips are far more superior in terms of noise performance, they give stable signal, low noise floor, no high-frequency distortions. ES9219 supports DRE but it is optional and off by default. CS are around just because of DRE and would measure badly having high noise floor without it.

ESS knows about issues with DRE that is impossible to make volume transitions absolutely transparent therefore ESS's DRE is optional and is Off by default on ES9219:
"ESS’ True DNR performance does not rely on Dynamic Range Enhancement (DRE) technology that shows higher DNR but is known to cause pop noise in some applications.". DRE is not needed for ESS DAC chips anyway because their noise floor is low already, so it won't give any benefit except the additional distortions.

Test file with beats used for measurements in case you will be interested to record and demonstrate Spectrogram of your USB DACs with CS chips inside. I used Audacity, set project and input device to 192000 Hz, recorded one channel. Before recording you need to calibrate USB DAC by playing 1kHz Sine at full swing (1...-1), record Spectrogram and adjust USB DAC's volume in case sine wave is clipped (overloaded), recorded sine must have as full swing as possible.

 

[Music1969]

I used Audacity, set project and input device to 192000 Hz, recorded one channel.

what do you use to record the DAC output? Cosmos ADC?

 

[dmitrykos]

what do you use to record the DAC output? Cosmos ADC?

Nope, Gigabyte B550 GAMING X V2 motherboard built-in Realtek which is max 16/192000. Sufficient for this quick analysis.

 

[jkim]

I think there is no complex thing behind the observed behavior of CS DAC chips. To my view CS is really doing just the DRE thing which was pretty well explained by @IVX. In my previous post I showed Spectrogram of distortions recorded from Apple USB-C 3.5mm dongle which does not have additional buffer stage but the recording of the DAC with 2xCS43131 + output buffer stage with OPA shows more interesting result and confirms DRE in action.

Here is Spectrogram of beats with additional 100, 250, 500, 1500 Hz Sine waves which add some power to the sound and without those tones, just beats:

View attachment 447141

You can observe the stripped noise pattern. It is does not exists on DACs with ESS or AKM chips without DRE employed but with all CS DACs (CS43198, CS43131) I have it is there. Here is a closer look of just several beats with tone and without:

View attachment 447142

Besides already mentioned distortions we can observe the CS's DRE in action. Additional output buffer stage after CS DAC chip amplifies noise floor, so we can easily see how it works due to the specific gradient view of the noise:

1. When beat starts DRE logic sets analog gain to the maximum compensated by the digital gain to the opposite direction: this transition is causing observed hi-frequency distortion because the change from silent part was rapid
2. When beat fades DRE logic gradually lowers analog gain and compensates it with digital gain but in this case no distortion is happening, we see that the noise floor is gradually decreasing (fading out)

So with this test CS's SNR is like fade-in -> fade-out, fade-in -> fade-out, ...

ES9038 + output buffer stage or Neutron's ES9219 do not exhibit anything similar, the noise spectrum is stable and clean, beats do not affect either noise floor nor causing any distortions:

View attachment 447146 View attachment 447147 View attachment 447151 View attachment 447152

AKM AK4490 based USB DAC gives similar to ESS DAC chips good result.

To my view any periodic changes affect the quality of sound and the perception of the music. Moreover rapid transitions are causing high-frequency distortions which act like a parasitic DSP effect by adding some coloration to the music. We can't hear the effect of EQ when it is applied it just adds some colorization to the range of frequencies, so it works similarly to unintended DSP - it adds some specific colorization. While this DRE technique can be ok for some technical use for the consumer music application it is a no-go technology to my view as it affects audio content periodically on constant basis, periodicity depends solely on the audio content, so with depending on the music style and its dynamics the effect can be minor or major.



To my view it is all about marketing and competition. Cirrus Logic wants to sell more chips, this technology allows them to do so. CS DAC chips give better measurement results and what we see? Majority of producers turned to CS DACs in order to win the race of measurements. In my measurements ESS and AKM DAC chips are far more superior in terms of noise performance, they give stable signal, low noise floor, no high-frequency distortions. ES9219 supports DRE but it is optional and off by default. CS are around just because of DRE and would measure badly having high noise floor without it.

ESS knows about issues with DRE that is impossible to make volume transitions absolutely transparent therefore ESS's DRE is optional and is Off by default on ES9219:
"ESS’ True DNR performance does not rely on Dynamic Range Enhancement (DRE) technology that shows higher DNR but is known to cause pop noise in some applications.". DRE is not needed for ESS DAC chips anyway because their noise floor is low already, so it won't give any benefit except the additional distortions.

Test file with beats used for measurements in case you will be interested to record and demonstrate Spectrogram of your USB DACs with CS chips inside. I used Audacity, set project and input device to 192000 Hz, recorded one channel. Before recording you need to calibrate USB DAC by playing 1kHz Sine at full swing (1...-1), record Spectrogram and adjust USB DAC's volume in case sine wave is clipped (overloaded), recorded sine must have as full swing as possible.

Thanks for the detailed explanation and data.

Your inference focuses solely on DRE as the underlying, data-generating process. But DRE is not the only mechanism in CS431xx that adaptively responds to signal levels: Class H operation by construction also does that as well as digital noise shaping in CS431xx. If you're willing to, get the JCALLY JM20 MAX and test it for yourself. You will see clear signs of DRE in action, but not be able to spot the same distortion behavior that was observed from many CS431xx-based devices (tested by Roman at RAA), like here (low-level multitone tests). Or the distortion in low-level 8 Hz sinusoid tests as shown here, which does not exist in the JM20 MAX, either. If you don't believe it, I suggest checking it for yourself.

 

[dmitrykos]

JCALLY JM20 MAX

Could you record its output based on my test file (end of previous post)? I got too many of DACs already from Neutron Player development field

 

[jkim]

Could you record its output based on my test file (end of previous post)? I got too many of DACs already from Neutron Player development field

I will do that when I get a chance.

By the way, in addition to distortions in the audible range, the spectrogram data you presented also show the effects of noise shaping being adaptive to signal levels, which can be seen in the following wideband noise level measurements reported by another ASR member here:

Note. The noise level is relative to fundamental tones. Noise shaping adapts to fundamental tones b/w -36 and -12 dBFS.

Noise shaping is at its max state (reported here):

This behavior is also exhibited by the JM20 MAX in exactly the same manner. But it passes the low-level multitone tests with flying colors.

 

[dmitrykos]

I will do that when I get a chance.

That would be great. The time between beats is 0.5s in the Beats.wav, so according to CS chip specification the noise gradient effect can't be due to Class H operation as specification clearly states 5.5s hysteresis for the voltage changes and volume is not faded that smoothly. Therefore operation in Class AB mode (ADPT_PWR = 001 or 010) would have no effect on the observed gradient effect.

I ran Beat test as DSD256 with Neutron Player (it supports PCM to DSD mode) and all previously observed issues are still there. So DSD can't bypass this processing and according to CS specification the search is limited to Multibit Modulator and circuit after it:

data you presented also show the effects of noise shaping being adaptive to signal levels

Maybe, the gradient effect of the noise. My raw guess - noise generated by the noise shaping is fading-in/out depending on signal strength which can be done by RMS analyzer wired to the gain control of the noise shaper (done in digital domain, there is no transition effect according to Spectrogram, the change is very smooth), it is like the Automatic Gain Control (or DRC in some TI ADCs) DSP for the mic but in opposite direction. The purpose of it to lower noise on quite places and get rid of hissing. The noise floor will also me measured well as a consequence.

Noise shaping is at its max state

I believe that noise shaping gain control is a workaround for such strong noise. And it also corrupts audio wave by adding high frequency noise when audio signal is changed too rapidly from quite to strong signal.

The better approach for this is to design chip with modulator working at higher frequency (like 2 or 4 times more) that allows to shift noise produced by the noise shaping further to the right (that is why DSD512 is better than DSD64). But this change will make chip more complicated due to stricter timing constraints, will increase power consumption, will increase die space needed. ESS and AKM DAC chips are doing much better in this respect and do not have any of these issues with modulator and noise shaping.

ES9219 supports external clock max 49.152MHz (mentioned AK4490 is also max 49.152MHz) while CS DAC chips support max 24.576Mhz that means ESS/AKM DAC chip has more advanced timing design and is able to run Modulator at higher frequency that results in better noise shaping where noise is pushed further to the right of the spectrum. Thus both ESS and AKM DACs do not exhibit this high-frequency noise visible in Spectrogram which is common to CS DAC chips. Can we say that currently offered CS DACs chips are really HiFi? It is very questionable, especially having ESS or AKM DACs chips without any of these issues.

 

[Robbo99999]

Unless USB DAC is based on nowadays popular Cirrus Logic DAC chip (CS43198, CS43131,...) with enforced Dynamic Range Enhancement (DRE) technique (using ESS name for it here) which fools test for the dynamic range and shows astonishing result. So basically instrumentally test is passed with excellent mark but in reality audio performance on real music medium can be average.

Human ear is an advanced multi-frequency sensor and our brain is a super-high speed analyzer. We do not have numeric representation of the sound when we hear it but instead we feel it. This feeling is the way we perceive sound and can feel differences based on perceiving of the whole musical content. We can like the result delivered on the basis of accumulated feelings of listened content or not. It is the basis of the listening tests and we should not underestimate them because brain can catch nuances which are not covered by the tests.

Back to CS case and dynamic content. There is an article in RAA (https://reference-audio-analyzer.pro/en/demo-cs131.php#gsc.tab=0) showing the audio defects in the Spectrogram on the recordings from DACs with CS43198 and CS43131 chips. I gave it a try and reproduced the same result, including with Apple USB-C 3.5mm dongle:

View attachment 447013

I then took DAP with 2xCS43198 and got similar result, as well as another DAP with 4xCS43131 (2 per channel). ESS DAC chip shows clean spectrum, as well as DAC based on AKM AK4490:

Neutron HiFi DAC V1

View attachment 447015

AK4490-based DAC

View attachment 447014

I created a simple rhythm track with beat sound using Audacity and played with CS-based DACs. The result was similar to RAA test, CS DAC chips always show the distortion.

Here is Spectrogram of the original material:

View attachment 447025

And here is the result of the recording from CS-based DAC:

View attachment 447024

Closer look at distortion revealed that dynamic transitions of analog and digital gains of CS chip are causing high-frequency distortions:

View attachment 447017

Here is the same wave produced by Neutron HiFi DAC V1 based on ES9219 DAC chip without DRE employed:

View attachment 447018

Although CS-based DACs have good measurement results, all of them are affected by CS-enforced DRE technique which is trying to improve the dynamic range. Measurement is done by 1 kHz Sine which is signal with constant power, therefore this technique will adapt DAC's output to it and improve dynamic range measurement result. Although it measured well, in reality on dynamic music content the result will be, to my personal view, average because technically it is a kind of DSP employing digital representation of signal and hardware. Volume transitions, although done fast by the chip, are causing high-frequency artifacts. In my example distortion is present on every beat, so it will affect music and will be perceived by the human ear when you hear complex music content which is full of frequencies, amplitude and phase transitions.

You responded to my post with that message, but I'll let you & @jkim have the discussion because it's above my pay grade!

 

[dmitrykos]

@Robbo99999 it's always fun to dig some technical quirks, sometimes interesting questions motivate for that

 

[dmitrykos]

The new update of Firmware 56 and NConfigurator 1.8.3 brought new DAC V1 operational mode - Voltage mode. By default DAC V1 operates in Current Mode (difference is explained in ES9219 datasheet) but in this mode analog volume is unavailable. New Voltage Mode switch in NConfigurator allows to activate Voltage Mode and release analog volume control for the [0...-24] dB range to the user. New option is beneficial for users with sensitive IEMs and headphones due to less current flowing with lower volume.

Firmware 56 improves SNR of DAC V1 by fine-tuning THD Compensation for both Current Mode and new Voltage Mode - 2nd and 3rd harmonics are much lower now (measured with 16-bit ADC):

Firmware 55 and older:


Firmware 56: Current Mode (default):


Firmware 56: Voltage Mode (new Voltage Mode):

 

[Phorize]

Unless USB DAC is based on nowadays popular Cirrus Logic DAC chip (CS43198, CS43131,...) with enforced Dynamic Range Enhancement (DRE) technique (using ESS name for it here) which fools test for the dynamic range and shows astonishing result. So basically instrumentally test is passed with excellent mark but in reality audio performance on real music medium can be average.

Human ear is an advanced multi-frequency sensor and our brain is a super-high speed analyzer. We do not have numeric representation of the sound when we hear it but instead we feel it. This feeling is the way we perceive sound and can feel differences based on perceiving of the whole musical content. We can like the result delivered on the basis of accumulated feelings of listened content or not. It is the basis of the listening tests and we should not underestimate them because brain can catch nuances which are not covered by the tests.

Back to CS case and dynamic content. There is an article in RAA (https://reference-audio-analyzer.pro/en/demo-cs131.php#gsc.tab=0) showing the audio defects in the Spectrogram on the recordings from DACs with CS43198 and CS43131 chips. I gave it a try and reproduced the same result, including with Apple USB-C 3.5mm dongle:

View attachment 447013

I then took DAP with 2xCS43198 and got similar result, as well as another DAP with 4xCS43131 (2 per channel). ESS DAC chip shows clean spectrum, as well as DAC based on AKM AK4490:

Neutron HiFi DAC V1

View attachment 447015

AK4490-based DAC

View attachment 447014

I created a simple rhythm track with beat sound using Audacity and played with CS-based DACs. The result was similar to RAA test, CS DAC chips always show the distortion.

Here is Spectrogram of the original material:

View attachment 447025

And here is the result of the recording from CS-based DAC:

View attachment 447024

Closer look at distortion revealed that dynamic transitions of analog and digital gains of CS chip are causing high-frequency distortions:

View attachment 447017

Here is the same wave produced by Neutron HiFi DAC V1 based on ES9219 DAC chip without DRE employed:

View attachment 447018

Although CS-based DACs have good measurement results, all of them are affected by CS-enforced DRE technique which is trying to improve the dynamic range. Measurement is done by 1 kHz Sine which is signal with constant power, therefore this technique will adapt DAC's output to it and improve dynamic range measurement result. Although it measured well, in reality on dynamic music content the result will be, to my personal view, average because technically it is a kind of DSP employing digital representation of signal and hardware. Volume transitions, although done fast by the chip, are causing high-frequency artifacts. In my example distortion is present on every beat, so it will affect music and will be perceived by the human ear when you hear complex music content which is full of frequencies, amplitude and phase transitions.

The consensus in audiological science is that hearing related cognition is a top down, expectations based system, and therefore a terrible analyser. Just sayin'.

 

[jkim]

The new update of Firmware 56 and NConfigurator 1.8.3 brought new DAC V1 operational mode - Voltage mode. By default DAC V1 operates in Current Mode (difference is explained in ES9219 datasheet) but in this mode analog volume is unavailable. New Voltage Mode switch in NConfigurator allows to activate Voltage Mode and release analog volume control for the [0...-24] dB range to the user. New option is beneficial for users with sensitive IEMs and headphones due to less current flowing with lower volume.

Firmware 56 improves SNR of DAC V1 by fine-tuning THD Compensation for both Current Mode and new Voltage Mode - 2nd and 3rd harmonics are much lower now (measured with 16-bit ADC):

Firmware 55 and older:
View attachment 447703

Firmware 56: Current Mode (default):
View attachment 447703

Firmware 56: Voltage Mode (new Voltage Mode):
View attachment 447705

This is good to know. That is a huge difference from the measurement point of view.

In fact, I was wondering WHY my measurements of the Qudelix 5K's THD+N are so different from (much higher than) the Fiio BTR15's when the two devices use the same integrated DAC/amp chip, the ES9219.

Of course, circuit design (e.g., power supply) can also affect their performance, but I highly doubted such a factor would be at play for a product like the 5K which has been refined over multiple iterations (and considering the designer's competence). Apparently Qudelix is more interested in its useability and functionality, not much in its technical, measurable sonic performance.

 

[dmitrykos]

@jkim being primarily an audio device its major role - digital to analog audio conversion, sonic performance shall go first, then applied functionality, imho Of course, it takes more development effort to add and support applied functionality as development can be spread over time and developer bears constant cost in its relation. It is more complex task of course than simply the development of hardware and throwing it to the market. Therefore we can see many audio devices without applied functionality and software, just dumb devices - they are simpler, require much less development effort and thus cheaper.

WHY my measurements of the Qudelix 5K's THD+N are so different from (much higher than) the Fiio BTR15's when the two devices use the same integrated DAC/amp chip, the ES9219.

THD Compensation is not the only factor. It can't fix hw design flaws and rather compensates inequality between resistors of the DAC chip. If for example clock jitter is too high then THD won't help, or when there is high-frequency noise then again THD will not do anything good. ES9219 is very versatile DAC peripheral and allows many different scenarios, so even though 2 devices use the same DAC chip the quality results can be very different: who is master/slave, what is clock source, how power is provided to external clock and DAC chip, is there enough power compensation, is it ultra-low noise, is PCB layout correct and etc..

 

[Plz virtual Surround 7.1]

When 7.1 virtual Surround please?

 

[dmitrykos]

Perhaps after fine-tuning of THD Compensation is completed. THD needs to be done for at least [0, -24] dB gain steps and there is possibility for algorithmic approach. So after Firmware 57 is released with THD Compensation for 2Vrms and 1Vrms Amp modes perhaps will take this task. There was interesting proposal in the thread to try adding Convolver DSP, so it was on agenda before 7.1.

 

[Plz virtual Surround 7.1]

Perhaps after fine-tuning of THD Compensation is completed. THD needs to be done for at least [0, -24] dB gain steps and there is possibility for algorithmic approach. So after Firmware 57 is released with THD Compensation for 2Vrms and 1Vrms Amp modes perhaps will take this task. There was interesting proposal in the thread to try adding Convolver DSP, so it was on agenda before 7.1.

Understood thank you very much for your hardwork so far!

 

[CedarX]

There was interesting proposal in the thread to try adding Convolver DSP, so it was on agenda before 7.1.

Just curious: a convolver is not too CPU-intensive for V1?