Master Thread: Are measurements Everything or Nothing?

[Steve57]

There are not good Engineering without good measures.

This is how things go round in the technological sphere.

Totally agree..

 

[Robin L]

"First off, please don't call me 'dear'". Ok, sorry!

 

[frullo]

Like what, to measure what?

Do you understand the measurements typically done?

You aren't really doing anything but repeating yourself over and over, with no sign that you are processing anything being said.

Like everybody else, Highway Pa-Troll!

 

[BDWoody]

Like everybody else, Highway Pa-Troll!

Ok, enough for you in this thread. That needed element of good faith is completely missing.

 

[voodooless]

You aren't really doing anything but repeating yourself over and over

You're a funny guy:

Do you understand the measurements typically done?

Master Thread: Are measurements Everything or Nothing?

No. ESS9028Q2M. Yeah, these are the smaller "mobile" versions. Yes, for a good reason; AK4490 was being phased out. Then the factory fire happened and most things are being phased out. RME is just staying ahead on parts. Any big brand will have a note somewhere that some parts can be changed...

www.audiosciencereview.com

Master Thread: Are measurements Everything or Nothing?

I did mix a lot of that together... Modern gear is what that is referring to...between two well measuring DACs for example. Ah, ok. Because the difference between a tube amp and a modern digital device can be very noticeable. Dac these days, yes, of course, even if there are differences on...

www.audiosciencereview.com

Master Thread: Are measurements Everything or Nothing?

Hmmh yeah, maybe Topping D90SE measures nice here, but, it depends how you look at things. I did run some measurements too. OK, that looked decent, let's see how it looks like with -60 dBFS input: OK, WTF is going on here? Let's try the same -60 dBFS signal with 352.8k rate: OK, better...

www.audiosciencereview.com

Master Thread: Are measurements Everything or Nothing?

pretty much Now that you get so much spanking here. Did you not say in a previous post that you wanted to test a tube amp? Try it... you (might ) like it .. (attached picture, yes, we are all adults in this thread). :) There you have something sensible you can try. There you MAY hear...

www.audiosciencereview.com

 

[Galliardist]

I understand this and agree. It is also a concept of listening very similar to my way of listening.

"First off, please don't call me 'dear'". Ok, sorry!

In that case, how do you avoid that high fidelity has always been achieved through engineering and technology? The first phonograph was built because the inventor had an understanding of engineering, acoustics and materials science. It's been the same ever since.

That is what makes measurements and specifications the best starting point for developing a system and the best initial test for the quality of components. Subjectivism only comes in when we know what will produce good sound - our cognition comes in at the end of that objective understanding.

 

[Sgt. Ear Ache]

Just a random thought, that I know has been expressed by others elsewhere:

Why do subjectivists, generally, eschew EQ and loudness? It’s a weird coexistence of preference uber alles, yet you can only get your preference through expensive trial-and-error switching of whole components.

I used to be that guy. I was the classic anti-EQ audiophile purist. If after assembling my system I felt the need to apply EQ to get it to sound good, then it was time to go shopping again because obviously my system was no good. I wanted a minimalist system...as few buttons and knobs as possible. Didn't even want a bass and treble knob. But that was because I didn't understand what EQ was for. Now, I get it. Put together a system that is not messing with the signal and then use EQ (and other room correction strategies) to as much as possible clear up any remaining inaccuracies. EQ isn't about shaping a pleasing tone. It's about correcting problems introduced by the elements of the system least able to be easily measured and controlled - transducers and rooms.

Combined with that is the audiophile conviction that there's more to sound reproduction than just frequency response and amplitude and noise. The belief that there's some "magic" that only happens almost by accident and can only be heard, but not measured. You have to find that by endless searching. The process is never finished...and it's contained within the component. You can't get it by moving sliders around.

 

[solderdude]

If the ultimate goal is to build a good reproductive system, the ear is still a good global control system, in the absence of a measurement made in multiple, exhaustive contexts. Perhaps a holographic recording and measurement system is what we should strive for in the future.

Hearing could be a tool as well.
It just isn't a repeatable and reliable tool.
There are ways to create better circumstances for using hearing as a tool.

There is no holographic recording at this point in time. Multichannel comes closest.
The problem is the vast majority of recordings is 2 channel. This we cannot change.
This means the 'recording' is the source. It is fixed and what happened before its creation is moot (except for tonal balance issues which can be corrected)

Now 2 choices remain.
1: Reproduce that was recorded as accurate as possible.
2: Make it sound pleasant to the listener.
combining 1 and 2 may lead to slippery slopes.

In choice 1 you can use measurements to see what comes closest and make improvements and quantify.
In choice 2 you can fool around, change things to your liking and do it all by ear if you want. Pleasant perhaps to individuals, accurate it may not be.

I don't care which one anyone chooses. Stating that option 2 is 'better' because someone likes it better is not the same thing as it being objectively better (as in faithful to what was recorded).
The last bit is what most folks here object to... stating something is better because their ears tell them it is. Preference is not equal to objectively better.

 

[Miska]

Confucius say "Man who not understand windowing functions, measurement methodologies, and system level variations" should not make definitive statements about their measurements on science based forums (not to mention not maintaining consistent bandwidths in the graphs).

Ahh yeah, I work on audio DSP like passive sonar systems and measurement devices since 1994 and still don't understand anything. Sure.

Maybe you are failing to read what is actually in the plots.

 

[solderdude]

My understanding is that the more one really knows about something the more one realizes they don't really know all there is to know about the subject.
One learns a bit every day.

Fortunately scientific findings should be repeatable and can be verified by others which should be done before discrediting someone else.
To disprove Miska's or Amir's findings one has to repeat the test and then can find a solution.

I remember the SMSL M500 having an issue that was missed by regular measurements and only once found out was fixed.

 

[xaviescacs]

@Miska, welcome back May I ask what are in your opinion, among all differences and artifacts that your plots show, meaning those that deviate from an ideal, the ones that can be audible and can make this two DACs sound different? Thanks!

 

[ahofer]

I used to be that guy.

Me too. But, especially with the advent of DSP, I realized it was just received dogma.

 

[Miska]

Fortunately scientific findings should be repeatable and can be verified by others which should be done before discrediting someone else.
To disprove Miska's or Amir's findings one has to repeat the test and then can find a solution.

One reason I got D90SE were the results Archimago got. Because I know ES9038PRO chip doesn't have noise floor issues. So I wanted to check whether those are related to the ADC he used (because it potentially has such issues due to insufficient anti-alias filter attenuation). But I stumbled number of related but different issues, so I didn't get conclusive result on that front. But at least I partially confirmed his findings, and that those don't exist for example on the M500 mk II I have.

But from other DACs with the same chip, and ES9028Q2M, ES9038Q2M and ES9028PRO devices, I know the noise floor of the chip is not lively. If it is there it is implementation related.

But look at the widened lobe at 44.1k and the distortion harmonics. And dirty spikes on 352.8k noise floor. Those are especially curious things. Just like tripping at 705.6k IMD test.

 

[Ken1951]

Guys, you're clearly being trolled, and it's not even vaguely creative trolling. I'd strongly suggest ignoring the fortune cookie generator unless by some miracle, it says something halfway intelligent.

That's why I put that dude on ignore ages ago.

 

[audio2design]

Ahh yeah, I work on audio DSP like passive sonar systems and measurement devices since 1994 and still don't understand anything. Sure.

Maybe you are failing to read what is actually in the plots.

The 44.1KHz, -60db plot shows evidence of either a different FFT length or window function. The frequency resolution appear to be considerably less. Alternately, there is a driver issue and where you think you are bit perfect, you are not. Either way, the difference is drastic enough to point more towards a measurement / setup / system error than an equipment bug. It would be good if each plot had FFT length, averaging if any, etc.

 

[audio2design]

We only have that input signal to work with. It’s what the artist and engineers signed off on. We aren’t going to improve it by applying our own filters and distortions after the fact. And even if one does think they can improve it by doing so, does it not make more sense to get a clean, undistorted signal to the speakers as much as possible and then apply distortions in a controlled and adjustable way via EQ rather than having some blanket inaccuracies applied to every single recording you listen to?

They didn't sign off on the signal. They signed off on what they heard, on their system, with their speakers, in their room. We won't improve it by adding distortions, but we may improve it for us personally, and quite possibly make what we hear closer to what the artist/engineer heard, by adding filters. Of course, if our speakers have less distortion than theirs, then adding distortion could even make it more "accurate" to what they intended.

I never implied that adding uncontrolled distortions and filters is the correct way to go. I am only pointing out that the "signal" is not really the reference that it so often assigned to it.

 

[anotherhobby]

That's why I put that dude on ignore ages ago.

 

[Sgt. Ear Ache]

They didn't sign off on the signal. They signed off on what they heard, on their system, with their speakers, in their room. We won't improve it by adding distortions, but we may improve it for us personally, and quite possibly make what we hear closer to what the artist/engineer heard, by adding filters. Of course, if our speakers have less distortion than theirs, then adding distortion could even make it more "accurate" to what they intended.

I never implied that adding uncontrolled distortions and filters is the correct way to go. I am only pointing out that the "signal" is not really the reference that it so often assigned to it.

We could certainly improve a specific recording for us personally. But what does that do for other recordings that don't need that improvement. There's no difference between signing off on "the signal" and signing off on "what they heard." The signal is all there is (although I would hope they at least listened to it on a few different systems). Whatever their systems are doing, the signal is all we have. It's a fool's errand to say "well we don't know what the producer was doing so we might as well just go hog wild." (I know that's not really what you are saying, but just making a point) There are great recordings, and there are not-so-great recordings. If I build a system that makes my half-dozen or so favorite test tracks sound awesome in the audio shop I buy from, how do I know that same system isn't doing harm to a bunch of other great recordings I didn't use for the audition? On top of that, I haven't really ever found that it's truly possible to greatly improve bad recordings anyway. Every adjustment you make involves a compromise. I find now with a system that is more or less built to simply pass the recording along transparently that most recordings - even stuff that I would describe as being pretty poorly recorded - actually sound fine. In other words, a transparent system allows me to get the best out of pretty much any recording.

 

[Frgirard]

Hear what the ME in a different room with different speakers is a joke.
The domestic room called living room is an acoustic battle field under the reign of the high decay and the reflections.
With the modal nightmare..
Try?

Galaxy Studios - About Us

Film Post Production and Music Recording studio in the heart of Belgium. Galaxy Studios is one of the leading film studios in Europe and home of Auro 3D

www.galaxystudios.com

Studio B

STUDIO B Ingénieur du son depuis plus de 20 ans, Benjamin Savignoni à travaillé pour de nombreux labels et artistes de styles variés et s’est forgé une excellente réputation. Il rejoint Trans…

www.translab-mastering.com

 

[Miska]

The 44.1KHz, -60db plot shows evidence of either a different FFT length or window function. The frequency resolution appear to be considerably less. Alternately, there is a driver issue and where you think you are bit perfect, you are not. Either way, the difference is drastic enough to point more towards a measurement / setup / system error than an equipment bug. It would be good if each plot had FFT length, averaging if any, etc.

Even if FFT length would be different it doesn't explain the distortions since it doesn't affect discrete tones. And differences between different input formats...

Driver is the same Linux kernel USB Audio Class driver as with all the other DACs too. I don't know if they use stock XMOS UAC2 firmware implementation, which is known to be horribly buggy.

And for example if you look at 1k and 19+20k plots for 705.6k input, I didn't touch analyzer at all between the plots, just source signal changed.

OTOH, M500 mk II is not problem free at 705.6k either, it just trips up on Jtest24 data.