• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Topping D30Pro Review (Balanced DAC)

Rottmannash

Major Contributor
Forum Donor
Joined
Nov 11, 2020
Messages
2,968
Likes
2,604
Location
Nashville
Thanks. This fs/2 filter business doesn't seen to get the attention it should, considering its importance in Nyquist-Shannon <https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem>
Wow-as much as I squinted, concentrated and re-read that, I just couldn't get past not having a basic understanding of the underlying principles. I really want to develop a better understanding of these filters because both my DACs and my DAP have multiple to choose from.
 

chris719

Senior Member
Joined
Mar 22, 2019
Messages
373
Likes
423
Wow-as much as I squinted, concentrated and re-read that, I just couldn't get past not having a basic understanding of the underlying principles. I really want to develop a better understanding of these filters because both my DACs and my DAP have multiple to choose from.

Well, good/bad news is the odds are overwhelming that none of them are "technically correct". You should probably just pick whichever is the linear phase sharp roll-off, or a minimum phase option if you have something against pre-echo.

The concern is IMO way overblown on the DAC side. Yes, it's technically wrong. Yes, you can measure it. You might even be able to hear it with specific test waveforms, but it's not some new thing. The Motu M2 seems to be the only thing Amir has even tested with a "correct" filter response, and that might have been a Windows / ASIO issue according to @bennetng.

The fact that it comes up on some random Topping DAC thread just rubs me the wrong way. I'm not affiliated with Topping, nor do I own any of their products, but I have to ask why this long-standing issue has not been raised in reviews of, say, RME or Benchmark products by the people that started off the tangent by complaining in this thread.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
I thought that the ripples caused by sharp filters were all about smearing in the time domain?
Ringing in time domain is the ideal outcome of bandlimited system. Just look at ideal sinc function.
And our auditory system is also band limited. So there's no benefit having "better" time domain.
In contrast, NOS has best time domain wave form, but it's the worst of all filters.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Well, good/bad news is the odds are overwhelming that none of them are "technically correct". You should probably just pick whichever is the linear phase sharp roll-off, or a minimum phase option if you have something against pre-echo.

The concern is IMO way overblown on the DAC side. Yes, it's technically wrong. Yes, you can measure it. You might even be able to hear it with specific test waveforms, but it's not some new thing. The Motu M2 seems to be the only thing Amir has even tested with a "correct" filter response, and that might have been a Windows / ASIO issue according to @bennetng.

The fact that it comes up on some random Topping DAC thread just rubs me the wrong way. I'm not affiliated with Topping, nor do I own any of their products, but I have to ask why this long-standing issue has not been raised in reviews of, say, RME or Benchmark products by the people that started off the tangent by complaining in this thread.
There's a fundamental difference between mathematically correct and engineering. Also under difference assumption and conditions math can be very different.
People need to understand this.
Who said one has to filter out exactly fs/2 IF the aim is not to reproduce fs/2 frequencies. No one. An audio dac is to produce 20KHz as perfect as possible. When this is the context, who would give a crap about imaging over 20khz. No one. Technically better if there isn't but you lose nothing. And better you get flatter 18k-20khz for people can potentially hear those frequencies. This is the key difference between text book and reality/in practice.
Btw this is not directing to you but take this as an opportunity to express.
 

chris719

Senior Member
Joined
Mar 22, 2019
Messages
373
Likes
423
There's a fundamental difference between mathematically correct and engineering. Also under difference assumption and conditions math can be very different.
People need to understand this.
Who said one has to filter out exactly fs/2 IF the aim is not to reproduce fs/2 frequencies. No one. An audio dac is to produce 20KHz as perfect as possible. When this is the context, who would give a crap about imaging over 20khz. No one. Technically better if there isn't but you lose nothing. And better you get flatter 18k-20khz for people can potentially hear those frequencies. This is the key difference between text book and reality/in practice.
Btw this is not directing to you but take this as an opportunity to express.

I fully agree. There is always a tradeoff to be made. If Cirrus really wanted to make the filter sharper then they have to trade other parameters or make the filter longer, which increases latency.

Sigma delta converters or Class D amplifiers are compromises by themselves in that regard if someone is that worried about a little ultrasonic content.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
The fact that it comes up on some random Topping DAC thread just rubs me the wrong way. I'm not affiliated with Topping, nor do I own any of their products, but I have to ask why this long-standing issue has not been raised in reviews of, say, RME or Benchmark products by the people that started off the tangent by complaining in this thread.
To be honest, I was a bit scared when you corrected my claim that CS43198 has great SPDIF performance. I specifically avoided mentioning the brand of the DAC (rather than the chip) so that people won't call me a shill. That said, I have zero interest in buying another DAC-only product, I am interested in multi-functional products like interfaces and such even if they have somewhat inferior measurement results, because I really trust my ears and know my limits.
 

chris719

Senior Member
Joined
Mar 22, 2019
Messages
373
Likes
423
To be honest, I was a bit scared when you corrected my claim that CS43198 has great SPDIF performance. I specifically avoided mentioning the brand of the DAC (rather than the chip) so that people won't call me a shill. That said, I have zero interest in buying another DAC-only product, I am interested in multi-functional products like interfaces and such even if they have somewhat inferior measurement results, because I really trust my ears and know my limits.

I see, I understand now. I have seen people get confused occasionally since some (all?) ESS DACs can actually accept SPDIF, as I'm sure you know. To be honest, I only came here because I saw the review linked on another forum and I'm designing a PCM4222 + CS43198 interface for my own personal use.
 

AudioSceptic

Major Contributor
Joined
Jul 31, 2019
Messages
2,690
Likes
2,534
Location
Northampton, UK
Ringing in time domain is the ideal outcome of bandlimited system. Just look at ideal sinc function.
And our auditory system is also band limited. So there's no benefit having "better" time domain.
In contrast, NOS has best time domain wave form, but it's the worst of all filters.
Yes, but isn't this what these filter choices are all about, trading flat FR and image supression against alleged time domain smearing?
 

chris719

Senior Member
Joined
Mar 22, 2019
Messages
373
Likes
423
Yes, but isn't this what these filter choices are all about, trading flat FR and image supression against alleged time domain smearing?

I'm not sure I'd use the term "smearing". It doesn't really have a formal definition since it was cooked up by a bunch of salesmen.

I think you may want to clarify if you are asking about passband ripple, ripple in the filter's impulse response, or visually evident ripple on a signal like a square wave (Gibbs phenomenon). I'm going to guess the latter since that's the part that looks wrong to most people, even though it's mathematically correct.

It depends how you phrase it, but you're sort of right I suppose. I would say that there is little point to the filter choices aside from a short delay filter for latency reasons. A slow roll-off filter makes "illegal" signals look better, yes. In a perfectly band-limited system, you can't have these "transients" the salesmen like to talk about.

If you are designing a digital waveform generator that has to generate perfect looking square waves, then yeah, you either have to have a ridiculously high sample rate or you have to relax the filter. Audio is just trying to be as perfect of a band-limited system as possible. The filter closest to the ideal sinc gets you the best result.

Some DACs let you choose between linear phase or minimum phase filters if the idea of pre-ringing offends you. I don't really think it's an issue either, but a decent minimum phase filter is fine, too.

The bottom line is that a band-limited system is supposed to be band-limited, so the example usually given of a square wave is wrong - it's got discontinuities.
 
Last edited:

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Yes, but isn't this what these filter choices are all about, trading flat FR and image supression against alleged time domain smearing?
There's no smearing.
You can't do convolution with your eyes can you?
The ringings are supposed to be there. And it's there to filter the imaging clean.
 

Joaquinín

Member
Joined
Sep 10, 2020
Messages
68
Likes
45
There's no smearing.
You can't do convolution with your eyes can you?
The ringings are supposed to be there. And it's there to filter the imaging clean.
I am not technically educated, but I understand from your words that there is no real pre-echo (or post-echo) with linear phase filters?
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
You can't do convolution with your eyes can you?
The ringings are supposed to be there. And it's there to filter the imaging clean.
I am not technically educated, but I understand from your words that there is no real pre-echo (or post-echo) with linear phase filters?[/QUOTE]
Those are actually not echos. That's another BS terminology brought up by manufacturer.

The linear phase filter will have pre-ring because all frequencies need to align at the center.
Minimum phase doesn't have pre-ring and the frequencies are aligned at the front.
There's difference in phase for each frequency but that's what they do.
 

chris719

Senior Member
Joined
Mar 22, 2019
Messages
373
Likes
423
I am not technically educated, but I understand from your words that there is no real pre-echo (or post-echo) with linear phase filters?

I don't think that's what he is saying. It's hard to define because smearing isn't a real thing. What does that even mean?

Pre-echo and pre-ringing are two different phenomena and we are going very deep down the rabbit hole here. Pre-echo is an artifact of certain types of FIR filters, I would ignore this mostly. Pre-ringing is an intrinsic property of a symmetrical linear phase filter.
 

Joaquinín

Member
Joined
Sep 10, 2020
Messages
68
Likes
45
I don't think that's what he is saying. It's hard to define because smearing isn't a real thing. What does that even mean?

Pre-echo and pre-ringing are two different phenomena and we are going very deep down the rabbit hole here. Pre-echo is an artifact of certain types of FIR filters, I would ignore this mostly. Pre-ringing is an intrinsic property of a symmetrical linear phase filter.
Pre-ringing is not an artifact? I mean, doesn´t it alter the original sound in anyway that we can hear? Is a proper analogue recording-reproducing chain free from it?
 

JohnM-73

Member
Joined
Jul 17, 2020
Messages
51
Likes
57
Location
Hampshire, UK

AudioSceptic

Major Contributor
Joined
Jul 31, 2019
Messages
2,690
Likes
2,534
Location
Northampton, UK
Ringing in time domain is the ideal outcome of bandlimited system. Just look at ideal sinc function.
And our auditory system is also band limited. So there's no benefit having "better" time domain.
In contrast, NOS has best time domain wave form, but it's the worst of all filters.
I'm not sure I'd use the term "smearing". It doesn't really have a formal definition since it was cooked up by a bunch of salesmen.

I think you may want to clarify if you are asking about passband ripple, ripple in the filter's impulse response, or visually evident ripple on a signal like a square wave (Gibbs phenomenon). I'm going to guess the latter since that's the part that looks wrong to most people, even though it's mathematically correct.

It depends how you phrase it, but you're sort of right I suppose. I would say that there is little point to the filter choices aside from a short delay filter for latency reasons. A slow roll-off filter makes "illegal" signals look better, yes. In a perfectly band-limited system, you can't have these "transients" the salesmen like to talk about.

If you are designing a digital waveform generator that has to generate perfect looking square waves, then yeah, you either have to have a ridiculously high sample rate or you have to relax the filter. Audio is just trying to be as perfect of a band-limited system as possible. The filter closest to the ideal sinc gets you the best result.

Some DACs let you choose between linear phase or minimum phase filters if the idea of pre-ringing offends you. I don't really think it's an issue either, but a decent minimum phase filter is fine, too.

The bottom line is that a band-limited system is supposed to be band-limited, so the example usually given of a square wave is wrong - it's got discontinuities.
I don't think that's what he is saying. It's hard to define because smearing isn't a real thing. What does that even mean?

Pre-echo and pre-ringing are two different phenomena and we are going very deep down the rabbit hole here. Pre-echo is an artifact of certain types of FIR filters, I would ignore this mostly. Pre-ringing is an intrinsic property of a symmetrical linear phase filter.
(I've been playing Devil's Advocate here.)

Many DACs that seem to be favoured by "audiophiles" have reconstruction filters that are not ideal, or even very poor. One of the tropes is that filters giving the best FR and image suppression give a poor time domain performance. By the latter I mean the impulse response, as explained in the introduction here <https://www.soundonsound.com/techniques/mqa-time-domain-accuracy-digital-audio-quality> (you can ignore all the stuff about MQA). The claim is that the ringing tails (which I called ripples because that is what they look like) represent temporal blurring or smearing, and cause an unnatural sound. I am *not* talking about FR passband ripple.

This is why some DACs have a choice of filters, and why JA measures the impulse responses, is it not, such as here <https://www.stereophile.com/content/okto-dac8-stereo-da-processor-measurements> ?

I hope that this makes things a little clearer. ;-)
 

chris719

Senior Member
Joined
Mar 22, 2019
Messages
373
Likes
423
(I've been playing Devil's Advocate here.)

Many DACs that seem to be favoured by "audiophiles" have reconstruction filters that are not ideal, or even very poor. One of the tropes is that filters giving the best FR and image suppression give a poor time domain performance. By the latter I mean the impulse response, as explained in the introduction here <https://www.soundonsound.com/techniques/mqa-time-domain-accuracy-digital-audio-quality> (you can ignore all the stuff about MQA). The claim is that the ringing tails (which I called ripples because that is what they look like) represent temporal blurring or smearing, and cause an unnatural sound. I am *not* talking about FR passband ripple.

This is why some DACs have a choice of filters, and why JA measures the impulse responses, is it not, such as here <https://www.stereophile.com/content/okto-dac8-stereo-da-processor-measurements> ?

I hope that this makes things a little clearer. ;-)

I know these things. I don’t know what you’re trying to accomplish here.
 

AudioSceptic

Major Contributor
Joined
Jul 31, 2019
Messages
2,690
Likes
2,534
Location
Northampton, UK
It is just pure BS. I'm sick of it. Just stop it. Do convolution calculation and see for yourself. What the ringing does to the signal. Nothing. That's the point. You don't seem to have any fundamental understanding of signal processing. Basic understanding of sampling and conversions will get you very far. Learn Sinc function first. What it is what it does. What's sinc in time domain, what's sinc in frequency domain. Why everything has to be band limited. If not what happens. What's the ideal case of reproduction. Also think about image processing and then compare to audio. What's the difference what are the similarities. Why is audio this way. If you have all these understanding, you then can say oh this filter does what, that filter does that.
Seriously?
Seriously, I realised that you didn't understand that I was/am playing Devil's Advocate, so I explicitly stated that, and you still don't get it?
 
Top Bottom