What objective measurements qualify as "High Fidelity"?

[terryforsythe]

(Continuing from a discussion in another thread that went off topic.)

My perspective is that what qualifies as "high fidelity" is subjective. Others disagree.

Assuming that there are objective measurement results that qualify a component/system as being high fidelity, what are the threshold levels of those results? Below are a list of measurement types (EDIT: and other system aspects), but feel free to add additional types.

Electronics (e.g., amplifiers and DACs)

1. Signal-to-noise ratio (SNR) - ?
2. Total harmonic distortion (THD) - ?
3. SINAD - ?
4. Intermodulation distortion (IMD) - ?
5. Stereo separation - ?
6. Frequency response - ?

Speakers and speaker/subwoofer combinations

1. On-axis frequency response - ? (e.g., XX Hz - XX kHz +/- X dB)
2. Off-axis frequency response (horizontal contour and vertical contour) - ? (e.g., -X dB +/- X dB at XX degrees off-axis XX Hz - XX kHz)
3. THD - ?
4. Dynamic range (compression) - ? (e.g., X dB max. at XX SPL)
5. Closed baffle or open baffle? (Not a measurement type, but may be relevant)

Tuning curves

Does implementing a tuning curve that is non-linear disqualify a system from being considered "high fidelity" since it deviates from faithfulness to the original recording? By way of example, the Harman curve has significant bass boost - does that mean it should not be used for a high fidelity audio system?

 

[Multicore]

As I see it, fidelity is a useless concept for our purposes and has been for decades. It was a useful marketing term back in the day but as an engineering goal I can't see how to use it.

All recorded music is artifice. So fidelity to what?

 

[lashto]

..
My perspective is that what qualifies as "high fidelity" is subjective. Others disagree.
..

fully agree .. mostly
Here's an interesting presentation on the subject "what does "accurate" even mean?"

 

[Chr1]

I guess we are taking fidelity to mean accuracy or resolution. And this accuracy is to the original recording. So therefore anything that deviates from this is not strictly "highest fidelity". This would include altered frequency response, I would assume. My 2c.

 

[charlielaub]

(Continuing from a discussion in another thread that went off topic.)

My perspective is that what qualifies as "high fidelity" is subjective. Others disagree.

Assuming that there are objective measurement results that qualify a component/system as being high fidelity, what are the threshold levels of those results? Below are a list of measurement types, but feel free to add additional types.

Electronics (e.g., amplifiers and DACs)

1. Signal-to-noise ratio (SNR) - ?
2. Total harmonic distortion (THD) - ?
3. SINAD - ?
4. Intermodulation distortion (IMD) - ?
5. Stereo separation - ?
6. Frequency response - ?

Speakers and speaker/subwoofer combinations

1. On-axis frequency response - ? (e.g., XX Hz - XX kHz +/- X dB)
2. Off-axis frequency response (horizontal contour and vertical contour) - ? (e.g., -X dB +/- X dB at XX degrees off-axis XX Hz - XX kHz)
3. THD - ?
4. Dynamic range (compression) - ? (e.g., X dB max. at XX SPL)
5. Closed baffle or open baffle?

Tuning curves

Does implementing a tuning curve that is non-linear disqualify a system from being considered "high fidelity" since it deviates from faithfulness to the original recording? By way of example, the Harman curve has significant bass boost - does that mean it should not be used for a high fidelity audio system?

It's interesting to note that for the "electronics" metrics you listed, you mostly chose steady-state, frequency domain measurements. What about burst measurements or time domain measurements, and related test signals? These are mostly lacking from the gamut of FFT-based frequency domain stuff that you can find.

For the speaker related: "Closed or open baffle' is not a measurement or metric. Those are design paradigms and there are good and bad examples of each. But you might add energy storage panel and other resonances, and power response to those.

 

[terryforsythe]

So therefore anything that deviates from this is not strictly "highest fidelity".

How much deviation is acceptable to still be considered "high fidelity"?

 

[Multicore]

I guess we are taking fidelity to mean accuracy or resolution. And this accuracy is to the original recording. So therefore anything that deviates from this is not strictly "highest fidelity". This would include altered frequency response, I would assume. My 2c.

What, in this conception, is the unit under test?

 

[Chr1]

How much deviation is acceptable to still be considered "high fidelity"?

Not defined, so no idea.
Just as vague as "sounds good" really.
I guess this is the point of this thread... To clarify some widely used, but vague audio terminology!

 

[Nopiano]

Somewhere we can probably find the old European DIN standards for audio/hifi, which I dimly recall was an attempt to codify such things decades ago. They are probably still just about sufficient, as those who heard decent systems in the 1970s can probably attest. In fact, slightly scarily, some now classic designs such as the Quad Electrostatic, can still give a remarkably persuasive sound, despite some fairly glaring weaknesses under measurement. The ears’s and brain’s ability to ’fill in the gaps’ is probably what makes even table radios and car audio enjoyable, even if they are far from ‘Hifi’.

A bit of googling suggests DIN 45500 isn’t easily accessible. However, some excerpts include this:-

”The document outlines the technical requirements for HiFi standard DIN 45500 for low frequency amplifiers from 1977, including: 1) The transmission range must be between 40-16,000 Hz with permissible deviations of ±1.5 dB for linear inputs and ±2 dB for equalizing inputs. 2) Transmission measure differences between stereo channels must be less than 3 dB without balance adjustment and less than 6 dB with a balance adjustment over 8 dB. 3) Distortion must be less than 1% from 40-12,500 Hz for integrated amplifiers.”

Modest indeed! But it does approximate to the specs in your opening post.

 

[Davide]

For me the concept of fidelity can only be a reference to the real event that produced that sound or noise.
But already the fact of measuring sound (or recording) implies a step that does not allow to establish fidelity, because in absolute terms you never know how exactly you are measuring. It is a concept of metrology.
Then... if the music has synthesized components and/or each sound is mixed together by means of instruments, analog or digital, then there is no longer the reference...
Therefore it becomes difficult to establish whether / which a measure is representative of fidelity.
The discourse of subjectivity can in a certain sense refer to fidelity, but it remains irreconcilable with the measurements, for the aforementioned point.

 

[Chr1]

For me the concept of fidelity can only be a reference to the real event that produced that sound or noise.
But already the fact of measuring sound (or recording) implies a step that does not allow to establish fidelity, because in absolute terms you never know how exactly you are measuring. It is a concept of metrology.
Then... if the music has synthesized components and/or each sound is mixed together by means of instruments, analog or digital, then there is no longer the reference...
Therefore it becomes difficult to establish whether / which a measure is representative of fidelity.
The discourse of subjectivity can in a certain sense refer to fidelity, but it remains irreconcilable with the measurements, for the aforementioned point.

Unfortunately that makes the term kinda pointless then. I think that it is a useful term, but only really if defined. I think fidelity to the recording is a useful reference point. Otherwise it is far too complex and would need to be broken down. I guess we are back to whether we are talking about audio reproduction or just audio in general. I tend to assume that most discussions here, unless stated otherwise, are referring to audio playback/reproduction.

 

[Vacceo]

The closest we have as reference is the source itself; being it a recorded orchestra or recorded cars moving on a street.

Beyond the source, we only have guesswork. Don't get me wrong, considering the intent can be fun and productive, but considering that intent involves emocional reactions that may or may not be archived, the signal is the only objetive reference we have.

 

[Multicore]

Unfortunately that makes the term kinda pointless then. I think that it is a useful term, but only really if defined. I think fidelity to the recording is a useful reference point. Otherwise it is far too complex and would need to be broken down. I guess we are back to whether we are talking about audio reproduction or just audio in general. I tend to assume that most discussions here, unless stated otherwise, are referring to audio playback/reproduction.

I don't think fidelity works as a useful reference point in any engineering perspective on or decomposition of the system. Setting the recording as the input of a decomposition is one possibility but I can't think of a reason why I should consider the recording correct. Indeed I often compensate for what I consider defects in the recording. If that makes my playback system low fidelity then why would I care about high fidelity?

What does help is to work towards rough consensus on what measurable properties we like to see in the standard subsystems of the decomposed system, e.g. amplifiers, speakers, etc. Relating measurements to fidelity in a formal way doesn't work.

 

[Multicore]

The closest we have as reference is the source itself; being it a recorded orchestra or recorded cars moving on a street.

Even that doesn't work, in my opinion. Have you ever spent time choosing a recording space, positioning the musicians in it, choosing mics and positioning them? It's very instructive to do so. Tiny differences have significant effects and big differences change everything. So what is the sound source and how does it qualify as a reference? And given that none of the resulting recordings you could make are what any listener actually heard, it's clear the recording itself is an artifice involving all manner of choices, most of which were made for reasons that we can only guess at.

Beyond the source, we only have guesswork. Don't get me wrong, considering the intent can be fun and productive, but considering that intent involves emocional reactions that may or may not be archived, the signal is the only objetive reference we have.

The signal is not the source you described in the previous paragraph, it is the output of an artificial recording process.

And it's not the only reference we have. We also have test signals. Using only those we are able to build systems that will please most people.

 

[DVDdoug]

Of course the issue is, "How high is high?"

Electronics are usually "transparent" with inaudible noise and distortion, and frequency response better than human hearing, so better than human hearing overall. (That's assuming you're not over-driving an amplifier into clipping/distortion, etc.)

Studies have been done related to the thresholds of audibility (which I don't know much about). I'm sure there are AES papers and books written on the subject. You can't give a simple-single number because for example a noise at 1kHz is more audible than broadband noise or noise at lower frequencies at the dB same level. There are lots of different distortion characteristics and different frequency responses that could fall-into the same "spec". And of course not all listeners are equal but there are still limits to what trained listeners with good hearing can hear.

Things get tricker with speakers & rooms. I suppose you could have a "transparent" speaker... until you put it in a real room. (You wouldn't want to listen in an anechoic chamber).

It's impossible to get the realistic sound of a concert hall with a pair of speakers in your living room. But you CAN fake-out listeners with recordings and speakers in a concert hall.... It's been done more than once. You might be able to get "close" at home with surround sound and acoustic treatment (and the right multi-channel recording). Or you CAN treat a living room or home studio to simulate the sound of a mixing/mastering studio but not everybody wants that.

 

[goryu]

We have to differentiate between fidelity to the actual performance and fidelity to the recorded signal. There is no objective way to quantify fidelity to the actual performance for a great many reasons I won't go into. The only objective goal is fidelity to the recorded signal. What is "high fidelity"? There is no absolute, quantifiable metric- all we can do is compare to the state of the art at any given point in time and say "A is truer to the signal than B". Fidelity to the signal is not fidelity to the sound of the live performance- any actual sound recorded or produced in a room is dependent on the room to a very large degree and thus we can at best only make objective observations about a speaker's fidelity via anechoic measurements. I don't think we can say absolutely that dsp is not high fidelity as it would depend on how it is used and the result.

 

[Justdafactsmaam]

This covers it pretty well.

 

[Justdafactsmaam]

Things get tricker with speakers & rooms. I suppose you could have a "transparent" speaker... until you put it in a real room. (You wouldn't want to listen in an anechoic chamber).

This urban myth needs to be put to bed. Unless a speaker system is designed to rely on room gain like the MBLs or early side wall reflections like the modern JBLs and Revels, the more accurate the speaker the more that accuracy will come through with less listening room sound. Dead rooms are ideal for speaker systems that are designed to be accurate direct radiators.

It's impossible to get the realistic sound of a concert hall with a pair of speakers in your living room.

That is no longer true now that we have the BACCH DSP but you do need to kill first order reflections

 

[Killingbeans]

Yeah... what's on the recording and what was heard in the studio are very likely two rather different things. And what's heard through a so called "high fidelity" playback system is probably a completely third thing, not matter what definition of the term was used during the design phase. Circle of confusion and all that.

In other words; The concept of "high fidelity" doesn't seem to have done much to break the circle.

What's really needed is an actual "reference performance". Not in the sense of higher performance than the competition, but simply a standard that both the studios and the home listeners can follow to make sure that they get roughly the same experience from the recording.

It's easy (IMO) to set a standard for audio electronics that's close enough, in any practical sense, to "wire-with-gain", and I don't see any reason to do otherwise.

If a person wanted to deviate from the standard in order to satisfy a personal preference, they would easilly be able to do so by using EQ or adding effects, but the standard itself would not suffer.

Speakers is a lot more of a minefield.

Flat anechoic response, and in-room response with a slight slope towards the treble, + a smooth off-axis listening window covering roughly +/-60 degrees, seems to be a good compromise, that gets close enough to the ideal speaker while also keeping unfavorable room interactions at a minimum.

I wouldn't personally mind seeing that set as a standard, but good luck making people agree on that

 

[Justdafactsmaam]

Yeah... what's on the recording and what was heard in the studio are very likely two rather different things. And what's heard through a so called "high fidelity" playback system is probably a completely third thing, not matter what definition of the term was used during the design phase. Circle of confusion and all that.

In other words; The concept of "high fidelity" doesn't seem to have done much to break the circle.

What's really needed is an actual "reference performance". Not in the sense of higher performance than the competition, but simply a standard that both the studios and the home listeners can follow to make sure that they get roughly the same experience from the recording.

It's easy (IMO) to set a standard for audio electronics that's close enough, in any practical sense, to "wire-with-gain", and I don't see any reason to do otherwise.

If a person wanted to deviate from the standard in order to satisfy a personal preference, they would easilly be able to do so by using EQ or adding effects, but the standard itself would not suffer.

Speakers is a lot more of a minefield.

Flat anechoic response, and in-room response with a slight slope towards the treble, + a smooth off-axis listening window covering roughly +/-60 degrees, seems to be a good compromise, that gets close enough to the ideal speaker while also keeping unfavorable room interactions at a minimum.

I wouldn't personally mind seeing that set as a standard, but good luck making people agree on that

With a standard do we not also set a ceiling for performance?