Why Some Audiophiles Fear Measurements
- Thread starter Jeff Fritz
- Start date
Orb, Thanks for the reply. There is a lot of symantic confusion on this collossal thread. My point is that blind measurements will often give you wrong answers. Measurement and testing are interactive, iterative processes that strive for an understanding. They are mini science experiments.
Because "you can't hear it on music that's near full scale when played at non-damaging volume levels" does not prove anything - we all listen to medium loudness and soft music at non-damaging volume levels.
Anyway, your private experience on dither does not agree with the experience of tens or hundreds of reaserchers and scientists who studied and published about the audibility of dither.
Whatever amps were used in Richard Clark's tests described here will do fine:
Richard Clark, all amplifiers sound the same
--Ethan
At normal volume levels you cannot hear the effect of dither. You have to turn the volume up unnaturally loud. And the music must then be very soft to not blow up your speakers. Again, please read my Dither Report and tell us which files are dithered and which are truncated. That's all you have to do to prove your case! If you get all 4 correct - which is not difficult even going on sheer chance - I'll reconsider my opinion.
Dithering to 16 bits is beneficial in theory. It's certainly useful in practice when reducing 16 bit audio to only 8 bits. It's also useful and needed in graphic applications when reducing color depth. But 24 bit audio to 16 bits? Sorry, the burden of proof is on you. If you don't like my audio examples, send me whatever you'd like and I'll make 4 sets of files from that. But don't send me a solo flute recorded at -40 or some such. It has to be representative of real music.
--Ethan
Unhappily you never give a straight answer to questions. I was interested in this thread because I have known for years your work about room treatment and could hope that your positions had a good fondations and we could have a fair talk about them . However you are not interested in measurements, but only in ambiguous statements and challenging everyone.
BTW , a simple software writen with the help of the MathlLab toolbox can immediately tell which file was dithered or not - if dither was properly done. But no need to reconsider - I am not interested in prizes and for me this thread is closed.
For many years, there was only one measure of distortion that was recorded in amplifier portfolios, THD. With all the ringing, obscuring, and noise, fixed any notion in my mind measurements are to be trusted. I have always trusted my ears, and will continue to. When an audio component works to increase transparency, it is a keeper. I have been very lucky in my choosing components that do that. I never developed a belief in one genre of components over another. Lately I compared the solid state NOS 47 Labs sound to my tubed AN DAC. It wasn't a fare comparison, because my AN DAC is not off the shelf. The 47 Labs didn't measure up. It would be wrong to assume I am a lover of tube distortion. There are no other tubes in my system. I do believe a better solid state DAC will show up some day. It has to be NOS, and take advantage of all the most advanced components. Taking my AN DAC in that direction has paid huge dividends.
The 47 Labs Flatfish transport stays. This tiny transport needs to stand on it's three spindly legs on a solid surface. The fact it's laser is direct coupled to what the transport stands on should spell disaster. Not if you have an imagination. I fixed that with a vibration absorbing substrate with no bounce. With the least number of stumbling blocks in the way of the music signal, the music flows unfettered in a marvelously live way.
Sorry that I brought in name brands. I can't really argue with the math and graphs that go into measuring sound. If measurements push other folk's buttons, far be it from me to knock them having fun.
The 47 Labs Flatfish transport stays. This tiny transport needs to stand on it's three spindly legs on a solid surface. The fact it's laser is direct coupled to what the transport stands on should spell disaster. Not if you have an imagination. I fixed that with a vibration absorbing substrate with no bounce. With the least number of stumbling blocks in the way of the music signal, the music flows unfettered in a marvelously live way.
Sorry that I brought in name brands. I can't really argue with the math and graphs that go into measuring sound. If measurements push other folk's buttons, far be it from me to knock them having fun.
OK, I must be a late bloomer but only now did I think to response to this
.First, let me say you are being unfair in the way you lump just about everything in distortion. It is like the old story of being given three wishes and for your last one, asking to have three more!
There are lots of distortion types and if you listed them all, your list would get a lot longer. And possibly be open-ended in that regard. For an example, read this paper on LIM:http://www.aes.org/e-lib/browse.cfm?elib=5611"A detailed examination of the digital logic Integrated Circuits (Ics) used within digital audio systems reveal modulation among the many signals generated within an individual IC, resulting in system clock jitter. In a compact disc player, digital control word signals (used for servo control, focus, etc.) and digital audio information generated from one IC affect the system clock generation found on the same IC. This jitter generation is prevalent in any digital playback system using standard IC designs. Clock jitter is measured and correlated with various digital signals."
Understanding how to measure each type of distortion and understanding what the data means can be very complex.
Putting that aside, let's look at another simple one: channel separation. Surely you don't want to bucket that in distortion.
The next one would be transient crosstalk. Imagine a 2x200 watt amp being driven by a 100 watt power supply. A sustained peak at 100 watts, will surely cause transient crosstalk in to the other channel by robbing it from its power linked to precisely what is being played in the other channel.
Compressed audio systems and FM receivers do even more whacky stuff by adaptively changing the stereo separation based on signal level and amount bandwidth available respectively.
I will stop here and get your reaction
.
I would say that all of your examples save one are forms of distortion from different causes which you clearly describe. So what if the category is large. If you don't like lop-sided kingdoms then don't become a biologist. Many eschew integrated circuits and use discrete electronics with point-to-point wiring to get around many of the problems to which you allude.
But, channel separation is surely important in stereo imaging and is an important spec in vinyl systems when considering cartridges. At a supposed >30 dB I haven't seen a problem with my Koetsu.
Frequency response, distortion, noise, time-based errors. Is that the right list? (Don't make me search.)
If so, I also await Ethan's reply also.
Of course you should listen to your hardware to make the final decision. No one is saying you shouldn't. I think what people are saying is take the bias out of the decision. Setup a blind test with a friend making the swap and see which one you really prefer. No matter how much you trust your ears, they are subject to bias just like the rest of us humans. If you can't train or think your way around optical illusions you can't do the same for audio.
I hope listing these two paragraphs adjacent shows the illogic in some of these posts. If you ask me something and I answer, please do not accuse me 20 posts later of not having answered. It's futile - and even a bit depressing - when basic logic is ignored. A thread that could have ended successfully many posts back with all in agreement is still going on, and I'm still answering the same questions, and explaining the same logical fallacies, repeatedly.
--Ethan
Dear Ethan: The only illogic on those posts is that you " read " only what is at your " convinience ". You never give an answer and I never speak of a 5th parameter.
The second paragraphs you quoted in your post comes from RonParty whom posted there too ( and obviously you don't choose to quote it. ) that you don't give a precise answer:
++++++++++++++++++++++++++++ " The burden of proof is on the person making a positive assertion. " +++++++++++++++++++++++++++++
in this case YOU.
That you don't have a scientific answer and that you are at the " defensive " on the subject going in circles around and around arguing with out that scientific answer only confirm what I posted: you are in this subject only: BLA; BLA; BLA; BLA and more BLA with out serious foundation.
So stop to try to degrade other people post about when you have almost nothing on hand to offer.
Microscript posted:
+++++++++++++ " . However you are not interested in measurements, but only in ambiguous statements and challenging everyone. " +++++++++++++++++++++
I can't agree more. I have respect for you as a person but in this non-answered subject IMHO you are a " shame " and believe me you win this " title " by sheer strength
with your negative non-answer attitude. Enough.
Regards and enjoy the music,
raul.
Last edited:
I am sorry but I don't follow. According to Ethan's rule of four measurements, those amps would prove to have no value at all as lower-end amps can meet all the same metrics.
One of the main reason high-end amps exist is because that extra class of distortion, namely, crosstalk which is absent from his list.
Actually, LIM is considered a different class of distortion in DACs. Yes, its effect eventually manifests in clock jitter. But it is important to consider it independently.
This actually brings up the same argument I made in my post. Assume you have an amp and you push it into clipping. At that point, if you measured its frequency response, you see that it is no longer flat, due to creation of high-frequency harmonics. Will you tell me that I should measure clipping then using frequency response measurement? If so, that would be odd and not nearly as revealing as plotting the power versus THD.
Well, it is my turn to say that is not a good argument. We may lose a lot of our high-frequency hearing but the critical 2 to 10 KHz range remains intact for people without hearing damage. Incredible amount of analysis can be done by the ear in that region.
So what is that if we measured it, we wouldn't be able to make sense out of it. What is the point of the measurement if it is not to teach us what is wrong with the equipment?
Heck, why does he have timing error as a category if having stuff lumped into one bucket is just fine? Why not through that out also and express it as distortion? You consider that progress if we did that?
LIM exists outside of a single IC if that is what you are describing.
To be fair, what he lists are some of the basics that we measure. But the science of understanding high-end gear performance is more complex if the idea is to learn and understand.
And as I said Ron either we missed it or feel it was not enough information or did not fit in the context of what was being said at that point (the links did not explain the process/mechanism involved or have any measured data-correlation-etc).
Sorry Ron but the amount of information Ethan is provided is not enough, it is too generalised without specifics such as explaining the mechanisms involved, or even breaking down specifically FR/Distortion.
I could say; any deviation that adds (can be added noise and distortion as an example)-removes-changes the temporal and any of the spectrum related information of the music explains all audio reproduction, in this sense Ethan's 4 parameters would be sub ones of mine.
This whole discussion in a good way got me going back to thinking and re-reading detailed theory for music-sound and instruments outside of audio reproduction and to consider what matters.
I am as correct as Ethan, and like Ethan I do not need to explain how Timbre/transparency/imaging/all temporal related matters/etc, however it does not really explain much even though it would cover variations in audio perception of music played back.
As Amir and also a couple of others have mentioned just saying distortion does not help, and I tried to explain this by showing how with different test scopes and even current published audio products tests that we can use for referencing do not show us enough information as they can measure audibly the same and yet sound different.
Now we are starting to see caveats relating so we have agreement current measurements do not reflect his position and yet it can also be proved measurement theory (how using much more complex tones beyond Nelson's paper will generate audible IMD) and do back up his speculation.
It is not possible to have both unless Ethan has done these further test measurements, which is fine but not what we have seen yet.
@Ethan,
I am going to drop the debate because it feels to me its reached the cherry picking parts of a discussion and information or not consider the full context of the partial post used that I have provided to support ones own ideas (and there is a risk of myself doing the same with so much information posted).
I provided Nelson's article to show to you that current test measurement do not relfect IM distortion - You have stated current measurements support your position on distortion (then went to IM distortion as Paul Miller's test show two products that definitely sound different but using the usual standard type tests measure comparably the same however it did not IMD include).
So yes and I said this; with a more real world test IM distortion can be audible as Nelson shows, but his conclusion that I also showed was that IF the harmonic (in these cases 2nd harmonic) is 0.02 or better, IMD will NOT be a problem.
So this is an opposite hypothesis to your take on IMD, and btw with the measurements I provided for the NAD and ARC they are far better than that.
However with IMD I tend to think that further testing is required as certain studies have shown instruments' partials/harmonics of a played note extend all the way up to the 20khz filter (sticking to CDs for now), and to see how linear IM distortion is when harmonic distortion is say 0.002 and also better/worse for products.
We may well see results like those for simulated real load on power amps and would go towards proving distortion and its role in audio playback perception when current standards of testing suggest they are inaudible (focus would have to be both the good and also excellent products and also different topologies that are implemented).
Without this though Ethan we have Nelson's view and measurements on distortion that do not back up your position just yet.
I hope this post is not coming off vocally strong as I have enjoyed the discussion and in fact it got me thinking about some other aspects relating to this discussion last night, but as I mentioned earlier we are definitely in a cycle of reiteration and this is going to cause confusion due to cherry picking what suits a poster due to so much already said.
Thanks for taking the time to respond Ethan and I appreciated it.
Cheers
Orb
Last edited:
Ah k I understand now and thanks for clarifying.
I did touch on that in my own way expanding on this much earlier and also by emphasising how measurements are used incorrectly by being taken out of context, going beyond the test's scope, not considering the test purpose, repeatable (once talking in more detail of a test process),etc, and once one of the link's I also provided touched on this I went back to summary mode as I would hope other members were following closely what was said or linked at the time.
But the problem is as you say this is a collossal thread and a lot of what I have said is going in a cycle so now just summarising.
In fact I am going to change focus to a different aspect that is interesting and not discussed because I feel once in reiteration cycle it is possible for confusion to arise due to summarising or parts of threads and their information is cherry picked (definitely not directed at you or necessarily anyone just the nature of online forum discussions).
I see it as this; if you have and understand/consider a correctly defined and implemented test procedure-protocol, with the correct scope (limitations and restrictions can be seen) and tools and the generated/measured data is used correctly, then we do not have the situation of blind measurements or partially blind accepted as full resolution, or incorrect use of those measurements and its associated test such as generally seen in quite a few audio forums.
Thanks
Orb
Last edited:
I liked post 629, nothing to comment about in it apart from you made a lot of valid points including that if using a parameter too generalised (such as seen with distortion that you highlighted very well in 629) it is very difficult to debate it when used to support a case/hypothesis/etc unless digging deeper and considering both test scope and measurement resolution.
Just to add about clearing audiophiles out of a room by mentioning hearing test
I cannot remember which two Floyd or Sean papers this is in, but together they show that in terms of preference those with normal hearing will usually trend the same way in their selection, while those with hearing loss (and also depends if low med high) have a high discrepancy that also is not consistent due to physiological bias effect.
What really makes this interesting is that taking those with normal hearing and that they trend very similar, their differences comes down to their threshold in terms of what is an acceptable level of error reproduced by the system or component.
In this it is an individual factor, which also is impacted by the music played as the error or type will be magnified or marginalised depending upon it, so it is "usually" never constant in real world (if taking music as being diverse and not deliberately being selective) for even the same listener.
Those who are about to disagree with my last sentence please note I emphasised usually
Cheers
Orb
Between this comment and your use of the word complex (in your final sentence, and if you used it deliberately in the scientific sense) raises a good point. When I first joined the forum I was raising questions about "complex behavior" in audio systems which often manifests itself in hard-to-understand chaotic phenomena. Perhaps you see what I was talking about.
I interpreted the "timing error" category as having to do with time and phase coherence. I am choosing this interpretation from the alternatives (like digital timing jitter which you and Don often talk about, which in my mind manifests as a distortion of the sound) because Ethan's focus tends to be on the final sound of the system at the listening position. Maybe that is the wrong interpretation. Perhaps he will clarify.
There are so many definitions of LIM on the web I can't be sure what you are talking about. I am thinking of intra-circuit em interactions which are well-recognized in high-frequency IC's, both analog and digital. Some audio circuit designers believe that these interactions occur with audible effects on closely-spaced audio circuits as well.
The behaviors of complex systems are poorly understood. A non-linear system with feedback can be subject to chaotic behavior (some extraordinarily effective random number generators are based on this). To some degree, this describes all practical audio systems.
The most general measurement I can think of is to compare the sound at the listening position with the amplitude vs time waveform of the source. The ratio of the two should be a constant. For simple sounds (like a few sine waves) the deviations from a constant will be explained in terms of THD and IMD. For complicated waveforms, like music, the deviations will appear random and, indeed, may not even be reproducible even if the exact same recording is played again. That may just be the nature of the beast. The only sure resolution to this is a completely linear system.
The reason I categorize this stuff is to explain that everything affecting audio fidelity is known, and can be assessed using current measuring techniques. There's no magic in audio. (Though there might be magic in a musical performance!) Too many people believe there are things they can hear that "science" doesn't know about. If I were to write "Everything that affects audio" it would end up being a five-volume set of books that I'm not even qualified to write. So I keep my ambitions to a reasonable size. Regardless, I do not lump everything under distortion. Only things that add new source-related content in the form of artifacts. So background hiss is not distortion, but added sum and difference IM tones is distortion.
Sure, but they're still all distortion, and so can be fairly lumped into that category. If you have additional subsets of distortion, give me the list and I'll add them to my Audiophoolery article.
Right, I list cross-talk under noise in my article above.
I guess that could be considered as either noise or distortion or both. Though distortion probably is more relevant since there are Amplitude Modulation artifacts. Other things can also be placed into multiple categories, such as ringing which is both a change in frequency response and also a timed-based error.
Sure, but that still doesn't alter the four basic categories. I honestly don't understand your (and others) objection to my categorizing audio parameters. The reaction I sometimes get amazes me. All I'm doing is trying to make audio more understandable and accessible to non-engineers. Why all the hostility? This is a serious question!
--Ethan
LOL, I loved that post and almost replied. This part sums it up perfectly:
And your point is?
The only reason high-end devices exist is to make money for the companies that sell them. In fairness, the same can be said for low-end gear. Personally, I rarely equate price with performance, especially with audio gear. Also, cross-talk is in my list as a subset of noise, as I explained above.
It may be important for design engineers to consider independently, but probably not for consumers which is who I generally write for.
--Ethan
Similar threads
