Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?

[Mr. Widget]

As to what causes the amp to be 'congested', I suspect its a distortion and bandwidth problem; I say 'bandwidth' in this case only because the SET in question is zero feedback so lacks any correction of phase shift in the upper frequencies. Since amplifiers can behave according to filter theory, the HF cutoff frequency of the amp, being on a 6dB slope, can affect phase relationships down to 1/10th the cutoff frequency (this won't be true of an amplifier using adequate feedback). IOW I suspect this amp of having poor HF bandwidth. While our ears are no good detecting phase of a sine wave, they do use phase to delineate location. This is particularly true of echos; when the ear detects a sound it makes a short term copy of it and looks for other examples in the near term. If it finds them they are used for echo location. So you can see that getting the phase right can affect how the amplifier seems to present the soundstage.

Very interesting.

Thank you for the explanation!

 

[Thorsten Loesch]

If the amp is at full tilt, then the signal will not be below the residual.

Well, seeing the spectrum of music at full tilt has very little level at the top end compared to the range below (say) 1kHz.

In practice, this does not seem to be a concern.

Maybe. I have not auditioned your Class D Amp, nor Purifi.

All other Class D Amp's before had the problem (the treble starts sounding distorted with large LF signals). It was not present in Class AB designs at hand.

I also designed a amp with unusually high switching frequency (1.411 or 1.536 MHz) and a non standard output filter that provided around 120dB carrier suppression and freedom from hysteresis distortion by design, despite using ferrite cored inductors.

The high switching frequency pushes the rise of distortion out to higher frequencies staying nearly flat to past 15kHz and the new output filter eliminates the need for an AUX-0025 Filter. Actually, by measurement it's not possible to tell the Amplifier apart from a Class A design but it lacks the crossover distortion of class AB.

Thor

 

[atmasphere]

The high switching frequency pushes the rise of distortion out to higher frequencies staying nearly flat to past 15kHz and the new output filter eliminates the need for an AUX-0025 Filter. Actually, by measurement it's not possible to tell the Amplifier apart from a Class A design but it lacks the crossover distortion of class AB.

Thor

I think class D is doing serious damage to class A, by virtue of eliminating its raison d'etre, which is maximizing linearity of the output devices.

Being of the self-oscillating variety and having good Gain Bandwidth Product to support its feedback, our class D does not have rising distortion with frequency. Of course, switching at a higher frequency offers more loop gain, but there is a decreasing return since deadtime is a constant, so distortion will be higher with higher switching frequencies. With any class D which employs a choke filter at the output, crossover distortion is literally impossible.

 

[pma]

Well, seeing the spectrum of music at full tilt has very little level at the top end compared to the range below (say) 1kHz.

We know that there are music sounds with very different spectrum shape

Is high-resolution audio audible or not audible and a waste of data?

Say more. This sound capture would be better in 24-bit than the 16-bit attached. (Carefully with volume knob!)

www.audiosciencereview.com

 

[Thorsten Loesch]

I say 'bandwidth' in this case only because the SET in question is zero feedback so lacks any correction of phase shift in the upper frequencies.

Quel dommage! So does our tweeter in our speaker.

Someone quickly go and tell everyone their speakers sound congested because their tweeters are zero feedback and lack any corrections of the phaseshift in the upper frequencies.

Wait, the ear too has severe frequency limitations in it's design, though arguably it has a complex feedback and feedforward system, including positive feedback to enhance small sound levels.

But at least it's not zero feedback, so no need for FUD and to worry ears sound congested through lack of feedback.

Phew, you had me worried there for a moment.

Since amplifiers can behave according to filter theory, the HF cutoff frequency of the amp, being on a 6dB slope, can affect phase relationships down to 1/10th the cutoff frequency (this won't be true of an amplifier using adequate feedback).

Oh yes, it will. It still behaves entirely according to filter theory. And if your dominant roll-off slope goes significantly past 1st otlrder life gets interesting.

IOW I suspect this amp of having poor HF bandwidth.

Vladimir Shushurin was quite a "wideband" guy. We both design bandwidth to the 400k rule and like full output at 20Hz (implying < -0.1dB @ 20/20k.

So if it was one of Vlad's Amp's, I doubt that diagnosis was correct.

While our ears are no good detecting phase of a sine wave, they do use phase to delineate location.

Ahhhmmm, not quite.

At high frequencies, above around 700Hz, directional hearing is based pretty much entirely on amplitude difference.

At low frequencies, below around 700Hz, directional hearing is dominated by arrival time differences.

I made my "Tonmeister" Dissertation with a new "minimal" microphone setup for recording that deliberately made use of this. That must have been '87 or so. It really works as well and translated very well to 2 speaker stereo, without the common "wide mono" impression at low frequencies.

So HF phaseshift or indeed delay sufficient to account for many 360 degrees of phaseshift do not alter the perception of direction all that much.

What does seem to have an outsized import is impulse response though, for sounds of a "impulse" nature (mainly found in percussion).

Thor

 

[Thorsten Loesch]

deadtime is a constant

Really? To me it's a design variable determined by the choice of output device and driver circuit.

With any class D which employs a choke filter at the output, crossover distortion is literally impossible.

I usually say it differently, namely that for class D "crossover switching" is the operation principle, not the problem.

To me the most important Realisation was that a Class D amplifier is nothing more than a step-down (buck) DC-DC converter modulated by an audio signal.

2MHz are pretty standard for state of the art buck converters. In fact, once we stop viewing the "beast" through the lense of "Amplifier design" and instead through the DC-DC converting PSU with an overlaid signal, many things become possible that are next to impossible in an "Amplifier" due to the wrong viewpoint.

Thor

 

[Thorsten Loesch]

We know that there are music sounds with very different spectrum shape

Is high-resolution audio audible or not audible and a waste of data?

Say more. This sound capture would be better in 24-bit than the 16-bit attached. (Carefully with volume knob!)

www.audiosciencereview.com

View attachment 261563

-50dB is even lower than my -40dB supposition.

Can you give me an orchestral tutti at FFF?

With the carrier breakthrough of many class D amplifiers at -40dB at the "silence" position and the switching frequency in textbook self oscillating systems tending towards zero at 100% modulation (and yes, 0.5 at 0.5/50% modulation).

Thus the carrier breakthrough with actual audio and lower frequency signal levels at near 100% modulation will be much higher than -40dB AND the carrier frequency will be much lower and closer in frequency to the quite modest HF signal levels in music.

Thor

 

[atmasphere]

So if it was one of Vlad's Amp's, I doubt that diagnosis was correct.

We're talking about an SET, so yes, bandwidth is a problem. I suspect the Lamm likely didn't have power bandwidth much past 20KHz.

Really? To me it's a design variable determined by the choice of output device and driver circuit.

You misinterpreted my comment. What you say above is true; as you know the deadtime remains the same regardless of frequency for any given set of devices. So while you get greater loop gain, you need it to deal with the increased distortion. That is why you don't see super high switching frequencies in most modern class D amps.

Thus the carrier breakthrough with actual audio and lower frequency signal levels at near 100% modulation

I can't speak for all amps in this regard but as you might have surmised, quite a lot of work has to go into a self-oscillating design to insure that it handles clipping correctly. A lot of the patent work of the Purifi module is about this very issue. The result is the clipping character is like that of any conventional solid state amp.
Since most solid state amps and most class D amps have a fairly abrupt clipping onset, the solution of course is to simply make sure you have enough power.

 

[Thorsten Loesch]

We're talking about an SET, so yes, bandwidth is a problem. I suspect the Lamm likely didn't have power bandwidth much past 20KHz.

The HF side of SET is generally not a problem. Winding transformers with enough sections to have ~ 100kHz -3dB is not in itself difficult.

Tango and Tamura did it for decades, I designed a few like that being far from a transformer specialist (I remember absolutely hating the nearly half year in my EE course where we dealt with magnetics).

You misinterpreted my comment. What you say above is true; as you know the deadtime remains the same regardless of frequency for any given set of devices.

So, change the devices? After all we are designers here. GAN is quite interesting.

We are only limited if we don't do our homework and instead act like Chinese and buy in (or illegally copy) other peoples work.

So while you get greater loop gain, you need it to deal with the increased distortion.

No, you need to look at this on a system level.

For example, using different drivers can reduce dead time using the same devices. Just evacuate the gate charge faster.

How? How about a positive feedback loop creating negative impedance as long as there is gate charge to evacuate or charge.

Use your imagination.

That is why you don't see super high switching frequencies in most modern class D amps.

And that is why they are wrong. They are lagging behind DC-DC conversion tech by decades, for no good reason.

I can't speak for all amps in this regard but as you might have surmised, quite a lot of work has to go into a self-oscillating design to insure that it handles clipping correctly.

Clipping is another subject.

If we design "25W music" tube Amp's, clipping behavior is paramount, as we will face clipping on peaks, which will be inaudible as long as it is instant and turns off instantly.

Any stickage on the rail(s) or material operating point shift is a problem.

With a "500W" amp we have more headroom, clipping should normally not happen, so we can relax requirements.

A lot of the patent work of the Purifi module is about this very issue. The result is the clipping character is like that of any conventional solid state amp.
Since most solid state amps and most class D amps have a fairly abrupt clipping onset, the solution of course is to simply make sure you have enough power.

Or incorporate soft clipping? I remember using tape saturation as "limiter" during recording in the 80's. Sounds swell. Beats any other limiter I tried.

Oh, oops, the result does not measure "happy panther", mind you, the Panthers don't listen, as they are ceramic.

Past that, I have been following Brüno's path with interest. He reminds me that that there is always an alternative to precision directed fire that takes the head of the enemy and thus secures a win with minimal casualties for both sides. Unlike the RuSSian equivalent of the good old banzai charge.

Why solve the root cause of a problem, when you can add more feedback? Why, more feedbacks makes clipping recovery difficult, so let's introduce non-linear "edge of clipping" limiting elements and use more feedback. What, your output inductor adds distortion? Add even more feedback.

I usually look at what Brüno does to give me a shortcut to know where else to look.

I believe in solving problems by resolving the root cause.

Thor

 

[tuga]

That super musical 30% distortion lol

You mean like vinyl played on a budget record player like most people (non-audiophiles) listen?

 

[fpitas]

You mean like vinyl played on a budget record player like most people (non-audiophiles) listen?

Not quite that bad. But close!

 

[Germinal]

Testing dacs results in some measurements which can be ranked, but the link to the sound that comes out has not been established. The presumption is that two dacs measuring the same will sound identical, but this hasn’t been tested or proven.

Can this even be proven ? I thought Amir measures the analogue output.. when the output is the same the sound should also be the same.. Whether it sounds the same to YOU or ME is another question and it may be more connected to psychoacoustics and/or placebo?

Reminds me of blind testing and the so called importance of making sure that the output is equal in loudness.. I guess if a 100 dollar dac can't be distinguished from a 10.000 dollar DAC purely because it's playing a bit louder... doesn't this mean there is actually no audible difference? so what's all the fuss about blind-testing ABX etc...?

If I sit in a 911 GT3 or I sit in an average car.. let's be clear.. I know where the money is being spent !

When I listen to a 10.000 dollar DAC or a 100 dollar DAC.. besides the casing and marketing and a beefier power supply.. I'm not so sure where the money has been spent... certainly a reseller will earn a few 1000 selling the first and a few cents selling the latter...

A few days ago I watched a YouTube video by some guy testing a few headphone DAC's and he could apparently hear a difference.. but only when he zoned in/focused on a certain hi-hat sound.. meaning in all honesty he couldn't really tell them apart when listening normally

 

[fpitas]

Can this even be proven ? I thought Amir measures the analogue output.. when the output is the same the sound should also be the same.. Whether it sounds the same to YOU or ME is another question and it may be more connected to psychoacoustics and/or placebo?

Reminds me of blind testing and the so called importance of making sure that the output is equal in loudness.. I guess if a 100 dollar dac can't be distinguished from a 10.000 dollar DAC purely because it's playing a bit louder... doesn't this mean there is actually no audible difference? so what's all the fuss about blind-testing ABX etc...?

If I sit in a 911 GT3 or I sit in an average car.. let's be clear.. I know where the money is being spent !

When I listen to a 10.000 dollar DAC or a 100 dollar DAC.. besides the casing and marketing and a beefier power supply.. I'm not so sure where the money has been spent... certainly a reseller will earn a few 1000 selling the first and a few cents selling the latter...

A few days ago I watched a YouTube video by some guy testing a few headphone DAC's and he could apparently hear a difference.. but only when he zoned in/focused on a certain hi-hat sound.. meaning in all honesty he couldn't really tell them apart when listening normally

There's always some guy on the internet who can discern the tiniest things. Even when there's no difference. And proper testing always seems to prove that measurements work.

 

[ahofer]

Testing dacs results in some measurements which can be ranked, but the link to the sound that comes out has not been established. The presumption is that two dacs measuring the same will sound identical, but this hasn’t been tested or proven.

The burden of proof is the other way around. The null hypothesis is that there is no difference. Insisting on "proof" of a lack of difference is doubly poor thinking. We measure things that are known, scientifically, to be audible (and to much greater sensitivity than even the exceptional human ear). If you assert they sound different, and hope to advertise, sell, or recommend equipment based on that idea, you need to show it.

Fortunately, there have been lots of controlled trials that are consistent with this null hypothesis, and providing zero evidence of an audible difference.

----some guidance---

So you simply cannot prove general claims that are negative claims -- one cannot prove that ghosts do not exist; one cannot prove that leprechauns too do not exist. One simply cannot prove a negative and general claim.

The Burden of Proof

Shifting the burden of proof[edit]​

One way in which one would attempt to shift the burden of proof is by committing a logical fallacy known as the argument from ignorance. It occurs when either a proposition is assumed to be true because it has not yet been proven false or a proposition is assumed to be false because it has not yet been proven true.[8][9]

Burden of proof (philosophy) - Wikipedia

en.wikipedia.org

 

[fpitas]

Testing dacs results in some measurements which can be ranked, but the link to the sound that comes out has not been established. The presumption is that two dacs measuring the same will sound identical, but this hasn’t been tested or proven.

Uh huh. We'll hop right on that.

 

[HarmonicTHD]

There's always some guy on the internet who can discern the tiniest things. Even when there's no difference. And proper testing always seems to prove that measurements work.

Oh yeah. Isn’t the internet great.
Wait I get the popcorn. Looks like we are in for another round ;-)

 

[fpitas]

Oh yeah. Isn’t the internet great.
Wait I get the popcorn. Looks like we are in for another round ;-)

I wonder if anyone seriously thinks we'll expend effort proving their crackpot assertions. Probably just trolling...

 

[HarmonicTHD]

I wonder if anyone seriously thinks we'll expend effort proving their crackpot assertions. Probably just trolling...

Oh sometimes we do ;-) until the mods kick em and spoil the fun ;-)

 

[Spkrdctr]

So, change the devices? After all we are designers here. GAN is quite interesting.

Thor

Thor

Thor, do you by chance know any audio equipment makers using GAN? I'm not up on it enough to know any companies using this.

 

[fpitas]

Thor, do you by chance know any audio equipment makers using GAN? I'm not up on it enough to know any companies using this.

We use it at 18GHz. It's been hard finding tweeters, though!