Topping E30 DAC Review

[ShinMolina]

That is the drawback of the usual delta sigmas in audio DAC’s. Plenty of ringing from the filters.

Try ‘em out unfettered if you can accept the ~ -3dB at 20kHz and like the sound (assuming you got good cans and/or near field monitors).

As for the mirroring in the (ultrasonic) higher bands and aliasing. There is only one way of knowing. Slap it in “NOS/unfettered” mode and listen; can you hear it?

Some dithering never hurt any electromechanical device, stiction, friction and all that. If the input stages of your amped speakers got some LP/bandpass filtering going on, you won’t notice the mirrored HF/ultrasonic bands anyway.

IMO.

I don't hear the difference between filters. I just usually set it to the one that has most adecuate response.

 

[redshift]

I don't hear the difference between filters. I just usually set it to the one that has most adecuate response.

Reasonable approach. What cans and near field monitors do you sport?

 

[redshift]

I don't hear the difference between filters. I just usually set it to the one that has most adecuate response.

Try creating a couple of minutes of square pulses and impulses, say with some 2s delay between the pulses, in the audio waveform generator favorite software of yours and then play it back while changing the filters. Any differences?

 

[ShinMolina]

Reasonable approach. What cans and near field monitors do you sport?

The ones I have in my signature. The ones I use the most are a pair of modded Fostex T50RP Mk3 as headphones, a pair of Tin P1 as in ear monitors and a pair of Adam Audio T5V as desktop speakers. I use the Dali Spektor 2 only with my turntable.

Try creating a couple of minutes of square pulses and impulses, say with some 2s delay between the pulses, in the audio waveform generator favorite software of yours and then play it back while changing the filters. Any differences?

I will try that tomorrow. After doing all these tests I could post everything on an independent post. Although, I can't comment much on step and impulse responses.

 

[ShinMolina]

Now imagine 16 bits/65,536 voltage steps when you crank the DAC to 0dB (full volume). Do you think those “steps” can be heard? After all, they arrive at 44.1kHz as the output voltage changes with the digitized music signal (assuming 16bit/44.1kHz)

If you’re running 24bit 192kHz+ hires songs. The default should be unfettered mode.
IMO.

I don't think those steps could be heard. It shows that the implementation is not "perfect", but having some carrier waveform at high frequencies is not likely to affect the sound you hear.

Why are those steps even there? Seems like @amirm measurements of the RME ADI-2 didn't show anything like that in the 1 kHz single tone test.

I'm 28 years old and can't hear anything above 14-16 kHz when I'm in a good and relaxed mood. When comparing my Topping DX3 Pro with my Xduoo MT-602 I can't always tell the apart. Maybe I'm not a good example for hearing artifacts, but I sure can measure things!

If you’re running 24bit 192kHz+ hires songs. The default should be unfettered mode.
IMO.

I mostly listen to Spotify at 320 kbps. Long time has passed since I listened to my full FLAC library, most of it is at 16 bit 44.1 kHz though.

 

[redshift]

I don't think those steps could be heard. It shows that the implementation is not "perfect", but having some carrier waveform at high frequencies is not likely to affect the sound you hear.

Why are those steps even there? Seems like @amirm measurements of the RME ADI-2 didn't show anything like that in the 1 kHz single tone test.

I'm 28 years old and can't hear anything above 14-16 kHz when I'm in a good and relaxed mood. When comparing my Topping DX3 Pro with my Xduoo MT-602 I can't always tell the apart. Maybe I'm not a good example for hearing artifacts, but I sure can measure things!


I mostly listen to Spotify at 320 kbps. Long time has passed since I listened to my full FLAC library, most of it is at 16 bit 44.1 kHz though.

I listen to YT music AAC VBR codec with my Sennheiser HD650, B&O BT cans, PSI near field monitors and it sounds good enough for me despite the compression. I don’t even bother with my living room rig, even though it is pretty decent. Can’t beat cans and near fielders when the ultimate crispness and fidelity is at stake.

I guess for better mixed/mastered/hires/bit perfect content the filters won’t matter, but if there is a hint of compression artifacts and sharp transients, I’m sure as hell can hear that ringing as “not nice”. It could also be the roll off from the NOS setting in the DAC.

I don’t know. Whatever. It makes me happy.

 

[redshift]

I don't think those steps could be heard. It shows that the implementation is not "perfect", but having some carrier waveform at high frequencies is not likely to affect the sound you hear.

Why are those steps even there? Seems like @amirm measurements of the RME ADI-2 didn't show anything like that in the 1 kHz single tone test.

I'm 28 years old and can't hear anything above 14-16 kHz when I'm in a good and relaxed mood. When comparing my Topping DX3 Pro with my Xduoo MT-602 I can't always tell the apart. Maybe I'm not a good example for hearing artifacts, but I sure can measure things!


I mostly listen to Spotify at 320 kbps. Long time has passed since I listened to my full FLAC library, most of it is at 16 bit 44.1 kHz though.

Those steps are the reality of digitized audio. Going one step up in voltage, say with 2V/65536 volts per increment according to a sine wave function plus some digital filtering will smoothen out the discrete steps. That’s what’s in Amir’s measurements, and for all DAC’s with filters switched on by default. RME gave us the opportunity to switch it off in their DAC’s because the AK DAC chip supports running in unfettered mode.

With the filter switched off in the case of the RME DAC’s NOS setting enabled, you are surely going to see the steps in the waveform since RME knows how to design competent output stages. For sure you’re not going to hear those steps when in full scale 16 bits data and certainly not with 24 bit full scale.

 

[ShinMolina]

Those steps are the reality of digitized audio. Going one step up in voltage, say with 2V/65536 volts per increment according to a sine wave function plus some digital filtering will smoothen out the discrete steps. That’s what’s in Amir’s measurements, and for all DAC’s with filters switched on by default. RME gave us the opportunity to switch it off in their DAC’s because the AK DAC chip supports running in unfettered mode.

With the filter switched off in the case of the RME DAC’s NOS setting enabled, you are surely going to see the steps in the waveform since RME knows how to design competent output stages. For sure you’re not going to hear those steps when in full scale 16 bits data and certainly not with 24 bit full scale.

Ah! I see. Didn't understand those steps initially. Of course filtering is needed in both R2R and delta-sigma DACs. Be it a RC filter or a digital one, something is needed.

 

[redshift]

Ah! I see. Didn't understand those steps initially. Of course filtering is needed in both R2R and delta-sigma DACs. Be it a RC filter or a digital one, something is needed.

Why? You won’t hear those steps anyway. However when recording audio using a ADC converter, those filters are essential.

 

[ShinMolina]

Why? You won’t hear those steps anyway. However when recording audio using a ADC converter, those filters are essential.

It's not that those filters are needed, but it's because of the same principle that makes the ADC have those stages. If you play something through a DAC you need a filter at half the sample rate because of antialiasing reasons.

It just happens that those filters eliminate those "steps" that appear in the ADI. If the frequencies above half the sample rates were wiped by the ADC, then the DAC should not reproduce any signal above that same rate.

 

[redshift]

It's not that those filters are needed, but it's because of the same principle that makes the ADC have those stages. If you play something through a DAC you need a filter at half the sample rate because of antialiasing reasons.

It just happens that those filters eliminate those "steps" that appear in the ADI. If the frequencies above half the sample rates were wiped by the ADC, then the DAC should not reproduce any signal above that same rate.

Why is that?

There is no inherent need for LP filtering for DAC that isn’t noise shaped. The reason for those filters is because the noise is shaped to higher frequencies and has absolutely zero with aliasing to do. Your analog gear post DAC doesn’t do any “sampling” of the signal.

For the noise shaped ones, there is even less need for that since the mirroring will be that high in frequency for it to matter anyway. Your preamp/amp and loudspeakers/cans won’t care about those frequencies anyway.

Just go unfettered and DSP/EQ those + 6dB with an upward sloping shelf. Or just do nothing if your loudspeakers and cans are relatively flat.

 

[solderdude]

The RME NOS setting uses DS DAC chip.
They simulate sample and hold.
maybe interesting to read
and maybe this as well

 

[redshift]

The RME NOS setting uses DS DAC chip.
They simulate sample and hold.
maybe interesting to read
and maybe this as well

Yep, assuming your amp can handle the mirroring in the ultrasonic bands, a nice Bessel/critically dampened LP filter followed by an all-pass to “fix” the phase would make short work of those mirrored HF frequencies in the input stages.

I’d just shove it in unfettered mode and not care about the HF. Heck, some HF content might even improve the sound of electromechanical actuators since they minimize the stiction/friction of the suspension.

And I rather have the minuscule ringing in the drivers than ridiculous ringing from sharp digital filters.

https://en.m.wikipedia.org/wiki/Ringing_artifacts

 

[solderdude]

So are you saying current DAC (chip) manufacturers and the math behind the sampling theorem is all wrong and sharp reconstruction filters are optional and wrong to use ?
I'm a bit confused about what you are trying to say.

a sample is only valid at a single point in time (the very short moment the sample was made). The rest of the time (1/44100 of a second) the actual waveform is 'ignored' and can only be reconstructed using a sharp filter NOT by holding the sample value during 1/44100 second.
The fact that R2R ladders change it into sample and hold does not make it right. It is just the only way it can be done but not the correct way. In this regard DS is much 'better', just like oversampling (using a sharp filter) is much closer to the truth than sample and hold with slow filtering.

 

[redshift]

So are you saying current DAC (chip) manufacturers and the math behind the sampling theorem is all wrong and sharp reconstruction filters are optional and wrong to use ?
I'm a bit confused about what you are trying to say.

a sample is only valid at a single point in time (the very short moment the sample was made). The rest of the time (1/44100 of a second) the actual waveform is 'ignored' and can only be reconstructed using a sharp filter NOT by holding the sample value during 1/44100 second.
The fact that R2R ladders change it into sample and hold does not make it right. It is just the only way it can be done but not the correct way. In this regard DS is much 'better', just like oversampling (using a sharp filter) is much closer to the truth than sample and hold with slow filtering.

Argumentation fallacies such as “Appeal to authority” causes my tilted obnoxious battery to go full loose cannon.

I’m saying that sharp filters is bad since they cause time based artifacts, ringing etc.

We are boldly assuming that the human auditory system works purely as a frequency analyzer. I wouldn’t be so sure of that and there is perhaps a more fundamental reason why people perceive vinyl and tape as “better sounding” despite its limitations.

Just giving some food for thought.

My proposal is to use oversampling DAC’s going full tilt at 24bit 192kHz+ and then slap analogue Bessel filters + all pass filters phase correction before the analogue signal hits the output terminals of the device.

 

[ElNino]

My proposal is to use oversampling DAC’s going full tilt at 24bit 192kHz+ and then slap analogue Bessel filters + all pass filters phase correction before the analogue signal hits the output terminals of the device.

That's a valid approach. Pretty retro, but reasonable.

 

[ShinMolina]

Why is that?

There is no inherent need for LP filtering for DAC that isn’t noise shaped. The reason for those filters is because the noise is shaped to higher frequencies and has absolutely zero with aliasing to do. Your analog gear post DAC doesn’t do any “sampling” of the signal.

For the noise shaped ones, there is even less need for that since the mirroring will be that high in frequency for it to matter anyway. Your preamp/amp and loudspeakers/cans won’t care about those frequencies anyway.

Just go unfettered and DSP/EQ those + 6dB with an upward sloping shelf. Or just do nothing if your loudspeakers and cans are relatively flat.

It's Nyquist principle, if you are playing audio recorded at 48 kHz then you have to get rid of any signals above 24 kHz because they shouldn't exist in the recording.

 

[redshift]

That's a valid approach. Pretty retro, but reasonable.

I’m old school new school. One step back, two steps forward, a giant leap for enthusiasts kinda obnoxious schmuck.

Repeat after me: It’s all good with some noises as long as the DR and SnR of digital is preserved.

 

[voodooless]

I’m saying that sharp filters is bad since they cause time based artifacts, ringing etc.

Here you go: http://archimago.blogspot.com/2018/01/audiophile-myth-260-detestable-digital.html

 

[redshift]

It's Nyquist principle, if you are playing audio recorded at 48 kHz then you have to get rid of any signals above 24 kHz because they shouldn't exist in the recording.

Can you hear 24kHz audio? I’m sure your amp can, so a bit of critically dampened LP filtering plus phase correction would sort that out in a hurry.