OBJECTIVELY, is there any sound improvement ??

[Archimago]

On 4/28/2020 at 2:50 PM, gmgraves said:

Objectively, there are many things one can measure that will alter the sound. DACs that don’t “ring”, for instance, should sound better than those that do. DACs that have excessive jitter should sound poorer than those that have little jitter. Delta-sigma DACs should sound better than R2R DACs (Ladder DACs) because once you get past 16 bits, it becomes harder and harder to maintain resistor linearity. But many of the most expensive (and arguably considered the best) DACs are R2R or some variation there of. It seems to me that execution of a design is more important than the methodology used. This seems to be true in so many things. 
BTW, the latest Yggdrasil from Schiit measures extremely well, and sounds unbelievably good easily besting other contenders such as the Chord Hugo TT + the Chord M Scaler, or the Benchmark 3. 

 

Just a couple of comments about this:

 

1. Don't worry about the "ringing" of DACs. Impulse response ringing is a result of the "illegal" signal fed into the DAC as discussed here. You're not going to "hear" any ringing (even if Meridian all those years ago thought minimum phase upsampling filters were a good idea for their DACs :-).

 

2. Jitter is highly unlikely to be audible unless very high as demonstrated here. Things like the J-Test is great to see the level of temporal resolution the device is capable of, an indication of engineering quality way more useful that hand-waving claims of "improved jitter" by cables, filters, reclockers, and ethernet switches...

 

3. We all have different levels of auditory acuity and preferences. Some might be completely happy with 12-bit resolution, others can recognize that they want their noise floor lower at 14-bits, and others feel they're not satisfied until the DAC can achieve 18-bits for satisfaction. Depending on that preference, even relatively low-resolution DACs could sound just great; some of us might even prefer a little noise, a little distortion that NOS DACs, and R2R's add (also those who prefer vinyl playback perhaps).

 

Given the huge number of variables in play, this is why as per my blog post discussed above, we have to be mindful of what we're "looking at" so that each of us can decide "what's best"...

 

PRODUCTION issues suggest we should look for the "best" mastering for our favourite albums...

REPRODUCTION issues suggest we need a better DAC/amp/speakers - for example, if we hear a level of noise inherent in the device when playing back our quietest albums - this can be measured and compared between devices. This is where objective measurements really help.

PERCEPTION issues relate to us as biological beings with perceptual thresholds. At this time of the year, I have mild allergies and I know that my hearing perception is skewed a little to the left so I know I should not be critically listening for channel balance on bad days. So too if we know we're getting older and have high frequency loss, or know that one ear is "weaker" than the other... Or it's time to get some earwax flushed :-).

 

People always want to have a "simple" answer like "THD+N must be better than -100dB" or "jitter must be <50 picoseconds" and stuff like that. That's obviously not helpful and will not resolve the "debates" because it's only addressing a small part of the big picture... This is why I write blog posts like this because we must IMO be more sophisticated in discussions than being too focused on the partisan "subjective"/"objective" divide.

 

[Superdad]

48 minutes ago, Archimago said:

1. Don't worry about the "ringing" of DACs. Impulse response ringing is a result of the "illegal" signal fed into the DAC as discussed here. You're not going to "hear" any ringing (even if Meridian all those years ago thought minimum phase upsampling filters were a good idea for their DACs 🙂

 

So are you saying that you do not hear the differences between minimum-phase, linear-phase, and the spectrum of possible filter variation in between those extremes?

[Archimago]

2 hours ago, Superdad said:

 

So are you saying that you do not hear the differences between minimum-phase, linear-phase, and the spectrum of possible filter variation in between those extremes?

 

Only if the filtering effect is extreme like what they did with the PonoPlayer resulting in significant frequency domain effects and aliasing. Otherwise there is no difference except for some relatively minor group delay.

 

Remember, even extreme minimum vs. linear filtering was poorly heard in a blind test years ago. That's the data and it jives well with my own listening and those I know who have tried it with a more controlled procedure.

 

What's your take?

[fas42]

My take is the extra processing required by digital circuits may change the electrical noise environment enough, to "do the damage" - an ambitious CD player I heard sounded best when that part of its internals was disabled - just because there is a change, doesn't mean that what one thinks is the reason, is in fact the true cause ... 😉.

[Superdad]

1 hour ago, Archimago said:

Remember, even extreme minimum vs. linear filtering was poorly heard in a blind test years ago. That's the data and it jives well with my own listening and those I know who have tried it with a more controlled procedure.

What's your take?

 

Well if you don't hear differences between digital filter settings, I really can't help you.

I and others spent many hours fine tuning filter parameters with iZotope Advanced in the old Audirvana+, and there is not a doubt that (on a good system) even very small changes can be heard.  There was a whole long thread about it with people comparing notes and settings:

 

I don't know why you think you think digital filters are no big deal since:

a) A lot of effort goes into them by many top DAC designers;

b) HQ Player, whose raison d'être is its many advanced SRC and SDM algorithms makes it easy to compare and hear a wide range of filter and dither effects;

c) You yourself published a whole blog post on an intermediate "Goldilocks" filter you came up with in SoX

http://archimago.blogspot.com/2018/01/musings-more-fun-with-digital-filters.html

 

[Summit]

On 5/23/2020 at 2:49 AM, Superdad said:

 

Well if you don't hear differences between digital filter settings, I really can't help you.

I and others spent many hours fine tuning filter parameters with iZotope Advanced in the old Audirvana+, and there is not a doubt that (on a good system) even very small changes can be heard.  There was a whole long thread about it with people comparing notes and settings:

 

I don't know why you think you think digital filters are no big deal since:

a) A lot of effort goes into them by many top DAC designers;

b) HQ Player, whose raison d'être is its many advanced SRC and SDM algorithms makes it easy to compare and hear a wide range of filter and dither effects;

c) You yourself published a whole blog post on an intermediate "Goldilocks" filter you came up with in SoX

http://archimago.blogspot.com/2018/01/musings-more-fun-with-digital-filters.html

 

 

The question is which filter sounds/measures most accurate? Not as obvious.

[Superdad]

4 hours ago, Summit said:

The question is which filter sounds/measures most accurate? Not as obvious.

 

Well Archimago wrote:

 

"Don't worry about the "ringing" of DACs. Impulse response ringing is a result of the "illegal" signal fed into the DAC as discussed here. You're not going to "hear" any ringing (even if Meridian all those years ago thought minimum phase upsampling filters were a good idea for their DACs."

 

and

 

"Only if the filtering effect is extreme like what they did with the PonoPlayer resulting in significant frequency domain effects and aliasing. Otherwise there is no difference except for some relatively minor group delay.

Remember, even extreme minimum vs. linear filtering was poorly heard in a blind test years ago. That's the data and it jives well with my own listening and those I know who have tried it with a more controlled procedure."

 

I was questioning those assertions as I found filter parameters to have a very large affect on SQ and instrument realism/attack. Not hard to train ones ear to it after a few hours of bracketing with iZotope Advanced. Get it just right and you know it right away. 

[4est]

7 hours ago, Summit said:

 

The question is which filter sounds/measures most accurate? Not as obvious.

There is no universal best. That is the point of having the options.

[The Computer Audiophile]

9 minutes ago, 4est said:

There is no universal best. That is the point of having the options.

When it comes to filters, this is especially true. There's no way to make the perfect filter. I look at them like balloons. Push in one side and the other sides sticks out, and vice versa. It's all about what people like. 

[MetalNuts]

New MSB Dac has no option of choosing filters.  I am not against objective view of how good a DAC can be but I believe it cannot be absolute objective in evaluating a DAC so far as listening is part of the evaluation. 

[Summit]

@Superdad seemed to believe that one filter type is clearly better than the others. I don’t think so. It is easy to hear the difference between gears and filters. But it is not as easy to tell which one sounds correct, because both has there flaws. This is especially true when we discuss different filters where you literally have to choose between plague (ringing) and cholera (phase error). The same applies if we measure the filtering effects on ringing or phases instead. Which is most correct? 

[Superdad]

30 minutes ago, Summit said:

@Superdad seemed to believe that one filter type is clearly better than the others.

 

Sorry, I never said that! (Though in the end I had clear preferences based on: the DAC I was using at the time, the type of music I was listening to, and my room/speakers.) Please reread what I wrote.  I was simply pushing back on @Archimago's notion that DAC filters parameters are:

a) Not that important unless the filter is really bad;

b) Not readily discernible.

Anyone--with a decent system and a sense of what real (acoustic) instruments sound like in a real space--who has spent some time with Audirvana/iZotope/SoX parameters or HQ Player's many truly excellent SRC algorithms knows that he is wrong. 

[Summit]

21 minutes ago, Superdad said:

 

Sorry, I never said that! (Though in the end I had clear preferences based on: the DAC I was using at the time, the type of music I was listening to, and my room/speakers.) Please reread what I wrote.  I was simply pushing back on @Archimago's notion that DAC filters parameters are:

a) Not that important unless the filter is really bad;

b) Not readily discernible.

Anyone--with a decent system and a sense of what real (acoustic) instruments sound like in a real space--who has spent some time with Audirvana/iZotope/SoX parameters or HQ Player's many truly excellent SRC algorithms knows that he is wrong. 

 

That is great when you may apprehend my question and attempt to discuss filers more objectively .

 

Which filter sounds/measures most accurate and why?

[The Computer Audiophile]

3 minutes ago, Summit said:

Which filter sounds/measures most accurate and why?

That's inherently subjective, in this objective thread :~)

 

The answers will of course depend on one's processor between his/her ears and upon which type of music is being played. 

[davide256]

2 minutes ago, Summit said:

 

That is great when you may apprehend my question and attempt to discuss filers more objectively .

 

Which filter sounds/measures most accurate and why?

I'm quite sure the answer lies in " garbage field theorem' ... push down garbage in one area and it will pop up in another. Which means yet more engineering band-aids

[Summit]

20 minutes ago, davide256 said:

I'm quite sure the answer lies in " garbage field theorem' ... push down garbage in one area and it will pop up in another. Which means yet more engineering band-aids

 

If that is the case a NOS DAC would be the most accurate…

[Summit]

23 minutes ago, The Computer Audiophile said:

That's inherently subjective, in this objective thread :~)

 

The answers will of course depend on one's processor between his/her ears and upon which type of music is being played. 

 

I do not think so. No filter is totally transparent, but that dosen't mean that the accuracy can't be tested between different models and types. 

[PYP]

I understand the concept of filters, but my DAC doesn't have selectable filters and it makes me think the designer should just choose what fits the design (perhaps a simple minded view of a complicated issue).  I feel like:  I don't need no stinking filters.  

 

 

[davide256]

56 minutes ago, Summit said:

 

If that is the case a NOS DAC would be the most accurate…

there is always the difference between theoretical science and applied science.... what models well in theory may not be practical in the limits of current material science. I do find that with the two NOS DAC's I've had (Metrum Octave and multibit Gungnir) there was only a marginal difference with applying oversampling vs

with a simple Chord Mojo the difference/improvement was easily detected. no idea why.

[mfsoa]

Really? Not significant differences in the sound between different filter types? Seriously? 

 

[John Dyson]

AIEEE -- the thing about digital filters and 'ringing' is all confused and silly.  Here are the absolute facts:

 

1)  The ripples that you see with brickwall FIR filters are NOT ringing, but they are Gibbs effect that results from a residue from MISSING frequency components.

2)The Gibbs type effect 'moves around' depending on the kind of filter (minimum phase, linear phase) based upon different delays vs. frequency, so the time location where the missing frequency components would have been.  So, the timing of the missing residue moves around relative to the rest of the signal.  (Kind of backwards wrt reality though.)

3) Even though some people say that you cannot hear the difference between minimum and linear phase, it MIGHT be possible in some circumstances -- but I haven't measured it myself...  Let me explain, a linear phase filter is 'perfect' except for a fixed delay vs. frequency.  A short (few taps) 'minimum phase' filter will have variable delay vs frequency, but a short filter cannot have much delay.   A very long (many taps) 'minimum phase' filter might have audible effects because of long relative time delays.  I don't know if they are audible, but a long (lots of taps) delay minimum phase filter MIGHT have audible artifacts.  It is up to those who measure these things WRT real world hearing tests to determine the answer'  @Archimago has  done some tests and I trust his results, whatever they are!!!

 

John

 

[pkane2001]

1 hour ago, John Dyson said:

AIEEE -- the thing about digital filters and 'ringing' is all confused and silly.  Here are the absolute facts:

 

1)  The ripples that you see with brickwall FIR filters are NOT ringing, but they are Gibbs effect that results from a residue from MISSING frequency components.

2)The Gibbs type effect 'moves around' depending on the kind of filter (minimum phase, linear phase) based upon different delays vs. frequency, so the time location where the missing frequency components would have been.  So, the timing of the missing residue moves around relative to the rest of the signal.  (Kind of backwards wrt reality though.)

3) Even though some people say that you cannot hear the difference between minimum and linear phase, it MIGHT be possible in some circumstances -- but I haven't measured it myself...  Let me explain, a linear phase filter is 'perfect' except for a fixed delay vs. frequency.  A short (few taps) 'minimum phase' filter will have variable delay vs frequency, but a short filter cannot have much delay.   A very long (many taps) 'minimum phase' filter might have audible effects because of long relative time delays.  I don't know if they are audible, but a long (lots of taps) delay minimum phase filter MIGHT have audible artifacts.  It is up to those who measure these things WRT real world hearing tests to determine the answer'  @Archimago has  done some tests and I trust his results, whatever they are!!!

 

John

 

Filters can have some level of ripple in the pass-band (not ringing) That's determined by the filter design. For FIR filters it's often a tradeoff between the sharpness (slope) of the filter, amount of out-of-band attenuation, and the amount of pass-band ripple for the same number of taps. Here's a cool on-line resource for constructing and visualizing various filters: http://t-filter.engineerjs.com/

 

For example, 113 taps filter is needed to get the pass-band ripple to be under 5dB in this example:

 

And if I want to reduce the ripple to under 0.01dB, this requires 235 taps, all else being equal:

 

With modern, mega-tap filters, the ripple in the pass-band can be made fully insignificant.

[John Dyson]

59 minutes ago, pkane2001 said:

 

Filters can have some level of ripple in the pass-band (not ringing) That's determined by the filter design. For FIR filters it's often a tradeoff between the sharpness (slope) of the filter, amount of out-of-band attenuation, and the amount of pass-band ripple for the same number of taps. Here's a cool on-line resource for constructing and visualizing various filters: http://t-filter.engineerjs.com/

 

For example, 113 taps filter is needed to get the pass-band ripple to be under 5dB in this example:

 

And if I want to reduce the ripple to under 0.01dB, this requires 235 taps, all else being equal:

 

With modern, mega-tap filters, the ripple in the pass-band can be made fully insignificant.

You are right about the frequency domain ripple also.  I am sorry that I didn't specify that I was speaking about the time domain Gibbs effect pseudo-ringing.  The pure frequency domain ripple, as long as it is below a certain number of dB isn't really all that important.

 

Also, the shape of the response curve can make some difference.

 

However, my hot-button is about the time domain ripples (not 'ripple')  that are attributed to 'ringing', which it technically is NOT ringing.

 

If I wrote 'ripple' somewhere, it was intended to be descriptive -- because all of my thinking was about the audiblility aspects, and most would likely be from relative delays -- by the time you are 40dB down, it starts on the way of inaudibiliity.

 

John