[Freo-1]

There is an on-going debate about how much measurements impact actual sound quality.  Whilst there is general agreement that measurements are indeed important, the vast majority of reviews provided today only show a subset of parameters that may or may not explain the associated sound quality of the item under test.  This subject was debated on another thread here on AC regarding GanFET technology vs. MOSFET technology for class D amplifiers.   Many audiophiles who have actually listened to GanFET class D amps noted the outstanding sound quality, even when compared to other current Class D offerings that "measured better".  There is an argument to be made that the measurements provided are incomplete, and do not capture some of the sound qualities of the GanFET amp. 


With the above outlined. here are some additional items for consideration:



Ralph from Atma-Sphere has provided the following quotes/observations:


"As Daniel Von Recklinghausen once put it:"If it measures good and sounds bad, -- it is bad. If it sounds good and measures bad, -- you've measured the wrong thing."

Atma-Sphere has developed a Class D amp, which is interesting considering their legacy.  I had asked Ralph why the GanFET approach sounds as good as it does?  Here is the response:

Me: I realize it's subjective, but clearly, something very good is going on here. Thoughts?

Its the distortion signature.  Our class D amp has a distortion signature that is very much like a good tube amp although considerably lower distortion overall. In a GaNFET amp, it appears that the deadtime generates distortion that tends to be lower ordered harmonics which are innocuous, other than adding a bit of 'body' to the sound. Between that and non-linearities in the encoding scheme, you wind up with a smooth presentation quite unlike class D amps of only a few years ago and easily on par or better than the best class A amplifiers (we've been making class A2 tube power amps for over 47 years).




I then asked if there are measurements available that can capture some of the sonic attributes of the GanFET?  Here is the response:


Distortion is the key. You're looking for three things:1) when distortion is graphed vs frequency, there will be no rise in distortion at higher frequencies. You rarely see distortion shown this way but its important; that rise in distortion at higher frequencies is caused by the feedback falling off. Its one of the reasons feedback has gotten a bad rap. Its caused by insufficient Gain Bandwidth Product in the design.2) the distortion spectra should have a nice 2nd or 3rd that dominates the distortion products. In this way the higher ordered harmonics can be masked. This is how tube amplifiers can sound smoother than solid state despite having more higher ordered harmonic distortion than most solid state amps. This means that the distortion spectra has to be shown. Its less useful if that is only done at 1KHz and more useful if you can also see the 5KHz spectra, to back up the findings of distortion vs frequency.3) the distortion spectra (distortion signature) does not change at higher power levels, such as -6dB from full power. This is where many amps fall apart; they start making enough higher ordered harmonics that are not masked and so become bright and harsh at higher power levels where at lower power levels that might sound alright.If no feedback is used, the distortion vs frequency should still be consistent across the audio band. And you are correct, quite often this stuff doesn't show up in typical reviews.



This gets into the entire subjective vs. objective argument.  It appears that there are additional measurements that could be provided to explain subjective observations, but currently are not provided in reviews.  This is why ANY review from ANY site NEEDS to be taken with a grain of salt.  One should not blindly just accept any sites recommendations based on a subset of measurements.   

[FullRangeMan]

Some mesearuments are irrelevant, others not as impedance, sensitivity and electric efficiency for speakers. However SQ are not irrelevant at all, over the years I noted tube amps always please all persons from any age or gender.

[Freo-1]

Some mesearuments are irrelevant, others not as impedance, sensitivity and electric efficiency for speakers. However SQ are not irrelevant at all, over the years I noted tube amps always please all persons from any age or gender.

Yes.  Distortion and linearity is primarily why that is the case.

[FullRangeMan]

The best explanation I have found on this subject is the image below from PlanetHifi.com.br every human being feels, hears, speaks and thinks in a different way from other.

[JohnR]

Might I just first say, that if anyone starts insulting anyone in this thread, it will immediately be Timeout City. Thank you.

As for the thread Freo-1, I think the title is unfortunate as they are complementary.

I think you missed the point being made in the other thread. The frequency response of the amp varies with load. This is unrelated to whether or not you think it sounds good. What it does mean is that someone else listening to the amp will most likely not have the same frequency response from the amp.

As far as things like "when distortion is graphed vs frequency, there will be no rise in distortion at higher frequencies. You rarely see distortion shown this way..." - actually it's quite common for this type of module, can you provide the link to the datasheet for the GaN module, it's most likely in there so let's take a look.

[VinceT]

As for the thread Freo-1, I think the title is unfortunate as they are complementary.

Measurements vs. Sound Quality or Measurements and Sound Quality

Much better 

[Freo-1]

Might I just first say, that if anyone starts insulting anyone in this thread, it will immediately be Timeout City. Thank you.

As for the thread Freo-1, I think the title is unfortunate as they are complementary.

I think you missed the point being made in the other thread. The frequency response of the amp varies with load. This is unrelated to whether or not you think it sounds good. What it does mean is that someone else listening to the amp will most likely not have the same frequency response from the amp.

As far as things like "when distortion is graphed vs frequency, there will be no rise in distortion at higher frequencies. You rarely see distortion shown this way..." - actually it's quite common for this type of module, can you provide the link to the datasheet for the GaN module, it's most likely in there so let's take a look.

Roger Wilco. 


I didn't miss the frequency load issue.  My point was that no one knew for sure the boards are exactly the same.  Peachtree may have modified the board to address that. 


I had asked Ralph if some of what is heard subjectively could actually be measured?   The above was his response,  indicating that some of the properties could be measured,  but generally are not provided.   The distortion spectrum at 5Khz was provided as an example.  That was/is the intent of this thread.


Currently researching to get more data on GanFET.   

[mav52]

Sure I look at measurements, but my ears and my room drive the final decision.

[FullRangeMan]

+1
IMO measurements are very useful for industrial, hospital or military equipment, for home Hi-Fi its only a secondary indicator.

[Freo-1]

Sure I look at measurements, but my ears and my room drive the final decision.

I think most of us are on-board with this. 

[wushuliu]

You can find 50yr. old archived audio magazine articles hashing out this debate and I'm sure it's been going on long before that. A lot of these questions could be answered - or at least put on the road to progress - with a modicum of combined effort. There is no desire to do so. And whatever attempts may have been made, clearly weren't deemed worth crediting. Therefore I have stopped caring about anything other than my own satisfaction.

Take me down to Timeout City
where the grass is green and
the girls are pretty
Oh won't you please take me
hoooooomeeeeeee

[RonN5]

Here is the Atmasphere link to their class d

http://www.atma-sphere.com/en/classD.html

[dwmaggie]

Thank you, Ralph, for bringing a class D Gan to market.  And thank you for the explanation of distortion perceptions.  Perhaps this will bring more acceptance of class D.  Thank you, everyone for the positive contribution to class D Gan.

[Freo-1]

Here is the Atmasphere link to their class d

http://www.atma-sphere.com/en/classD.html

Excellent.  The specs are very good.  Notice the feedback value. 


Pretty interesting that Ralph is hinting that the use of GanFETs actually provides the best qualities of tube amps with distortion characteristics (2nd and 3rd harmonics dominant,  only lower), and still has the best attributes of SS.  This is what many others who have listened to GanFET amps have reported.

[JohnR]

Excellent.  The specs are very good.  Notice the feedback value. 


Pretty interesting that Ralph is hinting that the use of GanFETs actually provides the best qualities of tube amps with distortion characteristics (2nd and 3rd harmonics dominant,  only lower), and still has the best attributes of SS.  This is what many others who have listened to GanFET amps have reported.

Shame there are no measurements to back his statements. Are there going to be any measurements at all in this thread? I asked directly before and you haven't answered.

[Freo-1]

Shame there are no measurements to back his statements. Are there going to be any measurements at all in this thread? I asked directly before and you haven't answered.

I hear you loud  and clear. I've got inquiries regarding these issues out there.  Hope to hear back,  as I'd like to see some myself of some of the amps in question. 

For example,  the distortion spetra at 5Khz would be a nice start.      Have specifically requested some measurements regarding the new Atma-Sphere amp.   Previously requested additional information from Peachtree,  and they are putting together a response.   Believe Ralph's explanations above do describe the sonic qualities of the GanFET performance adequately.  The additional measurements would help.

[goryu]

I think you missed the point being made in the other thread. The frequency response of the amp varies with load. This is unrelated to whether or not you think it sounds good. What it does mean is that someone else listening to the amp will most likely not have the same frequency response from the amp.

I think the point was missed by a few people. It's a fact that even load independent amps will sound "different" to different people simply because we can not objectively define what something sounds "like". Tastes being different, some may like the amp, others not so much. When you are considering an amp that has a load variant response, even the same person may have different subjective opinions about the amp simply due to how the amp's load dependent output changes the sound from one load to another. Personally, with today's technology and amp design know how, I consider any amp with a load variant frequency response a flawed product.

"Therefore I have stopped caring about anything other than my own satisfaction." Exactly. In matters of taste, the only thing that matters is one's own opinion, which can never be argued. As I said, I don't need anyone to tell me how well to cook my steak, what the "best" bbq sauce is, or if brand A sounds "better" than brand B.

Debate on objective criteria is another matter. Most people would agree that an amp's job is to add nothing to the input signal other than gain. If one happens to prefer an amp without feedback with higher distortion, it isn't the lack of feedback per se that that is "liked", it is the sound of the distortion added that is preferred. Topology doesn't have a "good" or "bad" sound, it has different distortion spectra that are perceived as "good" or "bad", and often this judgment is not consistent across a population.

"that rise in distortion at higher frequencies is caused by the feedback falling off. Its one of the reasons feedback has gotten a bad rap. Its caused by insufficient Gain Bandwidth Product in the design."

There is a diehard group of feedback haters who simply turn their back on feedback as some sort of inherently evil process, despite it's ubiquitous use in nature and engineering. Too many have used it improperly as a band aid for poor design. Like most things, there is a proper way to exploit the advantages it can offer. When done "right", it is extremely useful. Too many have a bias that colors their perceptions. Read the Putzey's paper on it's proper implementation.

"the distortion spectra should have a nice 2nd or 3rd that dominates the distortion products. In this way the higher ordered harmonics can be masked. This is how tube amplifiers can sound smoother than solid state despite having more higher ordered harmonic distortion than most solid state amps.

I personally prefer amps that have distortion products that measure below audibility so that there is no masking and nothing "smoothing" anything. Easier said than done. What we have is a large number of products that aren't truly faithful to the signal, rather, perfume has been applied to hide the otherwise unpleasant odor, and it is this "perfume" that many audiophiles have become accustom to as "good sound". There is no debating people's tastes so again, no point in arguing subjective opinion.

"Liking" something that has a second class design or measurements doesn't mean there is any causality between said design/measurements and one's subjective preferences. Just because one might prefer an open loop design to one with feedback doesn't mean "they must be doing something right". "Liking" an amp that measures poorly doesn't mean something must be missing or "wrong" with the measurements.

[Freo-1]

I don't think the load variation was missed.  It was more of if it applied to the Peachtree.  It may or may not, depends on if the Peachtree board has any differences.


No real issues with your observations per se.   Ralph also pointed out the distortion measurements at 1 khz don't tell one much.  It would tell one more if the measurements showed distortion at other frequencies as outlined.  (5khz for example).   This could go a long way towards the convergence  of  measurements as it relates to sound quality.  Same with distortion at different frequencies at high power. 

[goryu]

This could go a long way towards the convergence  of  measurements as it relates to sound quality.  Same with distortion at different frequencies at high power.

On what logical grounds can you claim there is a correlation between measurements and the undefinable subjective term "sound quality"? If something is by nature undefinable, and thus unmeasurable, how can you correlate the objective with the subjective in any significant, predictive, repeatable way? It has been tried by many and usually ends with rather gross generalizations with limited predictive power.

[Freo-1]

On what logical grounds can you claim there is a correlation between measurements and the undefinable subjective term "sound quality"? If something is by nature undefinable, and thus unmeasurable, how can you correlate the objective with the subjective in any significant, predictive, repeatable way? It has been tried by many and usually ends with rather gross generalizations with limited predictive power.

Engineers have been doing this with audio for many years.  Rob Watts of Chord used this method to generate and modify the algorithms to get them to sound as good (subjective) as possible.  If one watches the YouTube video on the DAC Master Design class, he explains how adjusting the noise floor downwards throughout the spectrum improved the sound from the DAC setup.  (Objective measurements validate subjective performance).  This is just one example of correlating measurements to sound quality.