Flat Frequency Response/Different Sound?

[JJFUSA]

To add to the discussion a little: If the FR is flat and the THD isn't insane, any audible differences in the DAC (or amp) usually exist in the listener's mind only, as this thread and others have pointed out.

However, you will probably come across speakers that have "the same FR" that sound different, too. In this case, they actually do sound different. This is because of "directivity" / "dispersion" which refers to how the FR varies depending on what angle the speaker is facing relative to you and the room.

This is why we look at "spinoramas" which measure the on-axis (facing straight at the speaker) frequency response, as well as off-axis (speaker is angled away from the listener) frequency response. They can be very different and they can vary in complicated ways.

What you hear when you listen to a speaker is a combination of the on- and off-axis measurements. All of the sound from the speaker bounces around the room and some of it reaches your ears. So you really need to see both to get a good idea.

Many times you only get to see the on-axis measurement. This is probably the single most important measurement, but you can't get a good sense of how the speaker REALLY sounds unless you also see several off-axis measurements.

Speaker measurements are both useful, more complicated, and more interesting than DAC or amp measurements. On- and off-axis FR are really only the beginning, but they are the biggest part of the story when it comes to speaker performance.

Very interesting, thank you, in fact just last night I was reading a great article by Ethan Winer exactly about the same room acoustic differentiations:

Why We Believe

ethanwiner.com

To be honest, I was extremely surprised to see that moving around even by 4 inches could make a huge deal and affect the FR.

Well, that doesn't apply to headphones and IEMs obviously, yet it shows how easily sound can be altered by the surroundings and effectively ruin any hi end setup.

 

[restorer-john]

@JJFUSA Your initial question is a good one.

To add to the discussion a little: If the FR is flat and the THD isn't insane, any audible differences in the DAC (or amp) usually exist in the listener's mind only, as this thread and others have pointed out.

However, you will probably come across speakers that have "the same FR" that sound different, too. In this case, they actually do sound different. This is because of "directivity" / "dispersion" which refers to how the FR varies depending on what angle the speaker is facing relative to you and the room.

I take some issue with those two paragraphs above.

Frequency response when measuring amplifiers is done at an arbitary level, usually 1W@8R (Amir uses 5W@4R). If an amplifier is not ruler flat across the audible bandwidth at 1W, it has no hope. Power bandwidth tests look for the -3dB points (1/2 power loss) and note those points. Running a frequency response plot at 1W, half power and full power is not the same as spot frequency power vs thd plots that Amir uses. One shows you how the response changes across the entire spectrum under loads, the other shows a few spot frequencies and when they clip/distort. Amir's is better in some ways, worse in others. Bottom line is frequency response plot is useless unless it is run at elevated powers as well as low powers in amplifiers. DACs, it doesn't matter.

And speakers will also change their response significantly as the power input changes. Some compress more than others, but all speaker exhibit a big change to their frequency response as they heat up.

 

[MRC01]

One factor that has been mentioned deserves stronger recognition: differences in level/loudness are not always perceived that way. Large and obvious changes in volume are perceived as volume. But smaller more subtle differences in volume are not perceived as volume, but as "richer, thinner, fuller, leaner, brighter" etc. all the audiophile terms. So an honest reviewer with good intentions and honed listening skills may faithfully describe what he heard in these terms, not realizing that the differences he was actually hearing were nothing more than subtle changes in level.

This especially affects reviewers who are smart enough to know that level matching is important, yet they do it in an informal ad-hoc manner that gets it close enough that perceptually, they think it's matched, yet it's actually mis-matched just enough to cause the above perceptual differences.

Also, frequency response may be most of what we hear, responsible for the most easily audible differences, but it's not the whole enchilada.

 

[restorer-john]

This especially affects reviewers who are smart enough to know that level matching is important, yet they do it in an informal ad-hoc manner that gets it close enough that perceptually, they think it's matched, yet it's actually mis-matched just enough to cause the above perceptual differences.

This is true. Audible/electrical Level matching is done most often at the spot frequency of 1kHz. And yet we are most sensitive to around 3kHz- right where a lot of two way speaker crossovers sit...

 

[kemmler3D]

@JJFUSA Your initial question is a good one.



I take some issue with those two paragraphs above.

Frequency response when measuring amplifiers is done at an arbitary level, usually 1W@8R (Amir uses 5W@4R). If an amplifier is not ruler flat across the audible bandwidth at 1W, it has no hope. Power bandwidth tests look for the -3dB points (1/2 power loss) and note those points. Running a frequency response plot at 1W, half power and full power is not the same as spot frequency power vs thd plots that Amir uses. One shows you how the response changes across the entire spectrum under loads, the other shows a few spot frequencies and when they clip/distort. Amir's is better in some ways, worse in others. Bottom line is frequency response plot is useless unless it is run at elevated powers as well as low powers in amplifiers. DACs, it doesn't matter.

And speakers will also change their response significantly as the power input changes. Some compress more than others, but all speaker exhibit a big change to their frequency response as they heat up.

Thanks @restorer-john , of course this is all correct, but for beginners we can use some oversimplified models before introducing the reality. The Bohr model still shows up in beginner physics classes! I also didn't feel up to typing all that out.

Well, that doesn't apply to headphones and IEMs obviously

Unfortunately, it does, sort of. Moving a headphone around relative to your ear can cause noticeable differences in FR. This is VERY noticeable in measurement rigs because of the artificial nature of a rubber ear with a mic in it, and is one of the reasons measuring headphones is so hard. It's less severe with real ears (i think) but the relative position of the headphone driver and your ear does matter, although a 4" shift is usually out of the question.

Ultimately sound is all waves bouncing around, and they don't stop doing that, ESPECIALLY when we want them to.

Anyway, you are on the right track it seems, Ethan Winer is a good one and ASR is a good place to find the deeper explanations of a lot of this stuff.

 

[restorer-john]

Moving a headphone around relative to your ear can cause noticeable differences in FR. This is VERY noticeable in measurement rigs because of the artificial nature of a rubber ear with a mic in it, and is one of the reasons measuring headphones is so hard. It's less severe with real ears (i think)

So true. I've come to the conclusion that headphones are so variable in that regard, that reccommendations cannot really be given as everyone places/shifts the 'phones to suit their own shape of head/ears/glasses etc. The response changes can be huge for just a few mm of movment. A tiny bit more pressure on the cups from a larger head etc.

 

[kemmler3D]

So true. I've come to the conclusion that headphones are so variable in that regard, that reccommendations cannot really be given as everyone places/shifts the 'phones to suit their own shape of head/ears/glasses etc. The response changes can be huge for just a few mm of movment. A tiny bit more pressure on the cups from a larger head etc.

Totally agree. And when you factor in that people's HRTFs are different, not to mention their actual equal loudness contours, you realize that the harman curve is really only a starting point (a good one, especially if EQ is available).

If you want the best headphone, you need to audition them, because performance of headphones is ultimately personal, if not subjective per se.

So, yeah, I think headphones are probably like shoes. There are objectively better and worse shoes for any given task, but no shoe fits everyone and nobody should spend too much on a shoe without trying it on.

Now, if personalized audio ever really gets dialed in, the analogy will shift, and headphones will be more like heat-moldable ski boots. You buy the best one you can afford and mold it to your own anatomy.

 

[restorer-john]

Now, if personalized audio ever really gets dialed in, the analogy will shift, and headphones will be more like heat-moldable ski boots. You buy the best one you can afford and mold it to your own anatomy.

Aren't there some IEMs being done that way already? I seem to recall someone on ASR talking about laser scanned ears and custom moulding? I could be wrong. Maybe I dreamt it.

 

[kemmler3D]

Aren't there some IEMs being done that way already? I seem to recall someone on ASR talking about laser scanned ears and custom moulding? I could be wrong. Maybe I dreamt it.

This sort of thing does exist already, the Nuraphone (released several years ago) uses mic-based measurement to tune the FR for your ear. It has a very unusual construction, (basically a closed back circumaural IEM) but people seem to like it.

I think the Apple Airpods Pro were also supposed to have some kind of customization but I don't know what came of it.

Physical custom molding for IEMs has been a thing for ages, but I was thinking more along the lines of tuning the FR to account for your personal HRTF etc.

 

[okok]

If the FR is flat and the THD isn't insane, any audible differences in the DAC (or amp) usually exist in the listener's mind only

you don't need dac or amp, there's some output from the mobo, save tons of money

 

[JJFUSA]

Thanks @restorer-john , of course this is all correct, but for beginners we can use some oversimplified models before introducing the reality. The Bohr model still shows up in beginner physics classes! I also didn't feel up to typing all that out.

Unfortunately, it does, sort of. Moving a headphone around relative to your ear can cause noticeable differences in FR. This is VERY noticeable in measurement rigs because of the artificial nature of a rubber ear with a mic in it, and is one of the reasons measuring headphones is so hard. It's less severe with real ears (i think) but the relative position of the headphone driver and your ear does matter, although a 4" shift is usually out of the question.

Ultimately sound is all waves bouncing around, and they don't stop doing that, ESPECIALLY when we want them to.

Anyway, you are on the right track it seems, Ethan Winer is a good one and ASR is a good place to find the deeper explanations of a lot of this stuff.

Thank you very much for the kind words, I've always been in favor of the pure unaltered original source representation, fidelity and faithfulness to the initial recording is kinda my philosophy of the music perception, that's why I was quite concern about confusing reviews online, so I made the post here to seek professional unbiased response, which I received indeed.

Considering moving a headphone and FR you're absolutely correct, even when I adjust my little IEMs in my ears I can hear different sounding, I was rather talking about movement around the room like Ethan Winer did during his testing - that would not affect headphones, well seems like everything affects everything and there are no perfection for sure. lol

 

[JJFUSA]

One factor that has been mentioned deserves stronger recognition: differences in level/loudness are not always perceived that way. Large and obvious changes in volume are perceived as volume. But smaller more subtle differences in volume are not perceived as volume, but as "richer, thinner, fuller, leaner, brighter" etc. all the audiophile terms. So an honest reviewer with good intentions and honed listening skills may faithfully describe what he heard in these terms, not realizing that the differences he was actually hearing were nothing more than subtle changes in level.

This especially affects reviewers who are smart enough to know that level matching is important, yet they do it in an informal ad-hoc manner that gets it close enough that perceptually, they think it's matched, yet it's actually mis-matched just enough to cause the above perceptual differences.

Also, frequency response may be most of what we hear, responsible for the most easily audible differences, but it's not the whole enchilada.

Perception of the differences in level/loudness you mentioned could be connected to the Phon I presume.

Phon - Wikipedia

en.wikipedia.org

That's why it's so important to match sound levels of the units during any A/B testing, like you said even tiny deviation in volume would affect what we hear and lead to the wrong conclusions since FR of our ears changes with the loudness as equal-loudness contours suggests.

 

[kemmler3D]

fidelity and faithfulness to the initial recording is kinda my philosophy of the music perception,

Well, you've come to the right place. Opinions vary here but the predominant one (different from most sites on the web) is that this is the most appropriate objective for audio gear.

seems like everything affects everything and there are no perfection for sure. lol

This is pretty much true which is why audio is so deceptively tricky! Not just the gear, but what's going on in your own brain tends to change the outcome quite a bit. This is why a science-y approach is necessary, because without strict controls you have no idea what's causing changes in what you're hearing!

If you want a serious rabbit hole, start looking up DIY speaker design stuff... it's really just one tradeoff after another.

 

[kemmler3D]

you don't need dac or amp, there's some output from the mobo, save tons of money

I mean, that is still a DAC and amp... and IME mobo-based audio is true garbage, like not even "good enough for casual use". However I know that some are pretty good. You raise a good point which is if you have technically decent but un-glamorous gear, use it anyway and ignore the haters.

 

[JJFUSA]

I mean, that is still a DAC and amp... and IME mobo-based audio is true garbage, like not even "good enough for casual use". However I know that some are pretty good. You raise a good point which is if you have technically decent but un-glamorous gear, use it anyway and ignore the haters.

I confirm that the most of mobo-based audio is true garbage especially in laptops, even basic sound card / USB DAC sounds better that onboard sound, no needs for any measurements the difference is drastic and hearable pretty well.

 

[IAtaman]

I do find people who talk about the tonality of an amp as if it is a new breed of apple utterly irritating and have been on the if it measures the same, it sounds the same camp since forever, but these days I do find myself to be more and more skeptical whether that is in fact always true.

Not so much for DACs, but more so Amps I have to say, and that's mainly because I am of the perception that a conclusive test to check the linearity of DACs can be devised but I am not so sure if same can be said for Amps, and whether the measurements with which we have been presented are conclusive enough to prove different amps sound the same under various operating conditions. I really hope this is "just a phase", and as I learn more about them it will all go away

My main issue is that I don't know whether we are measuring all parameters that might have an impact on the sound reproduction; and can, with reasonable accuracy, conclude that these devices are linear and low distortion across all operating conditions. I mean there are THD vs Freq and RMS vs Freq graphs in reviews but is that it, is that all we need to know? For example are those graphs true for all temperatures and all varying impedances? Do they also account for transient response capabilities of amps, do we measure to see if all amps have enough headroom in their output stages and power supplies to handle all kinds of wanky input signals and headphones people will throw at them, or does this even matter? I don't know.

This is not to give any credit to bullshit connoisseurs mind you- a lot of low lives have been and still do live in the shadows of science and tested knowledge, telling all kinds of stories to people for things we don't conclusively know. But I think it is also very much in the spirit of the same scientific method to be skeptical, and do not assume we know and can measure all that is relevant to the topic in hand, and understand the limitations of our current knowledge and measurement capabilities.

 

[antcollinet]

......no needs for any measurements the difference is drastic and hearable pretty well.

Always be wary of statements like that - whether they come from others, or are your own conclusion. The cognitive biases talked about upthread can - and often do - result in perception of differences that seem like "night and day" or "hearable pretty well" even though those differences don't actually exist in the sound reaching the listeners ears. Everyone is subject to this. It doesn't mean you are less skilled, or have faulty hearing - it is just the way people are built.

The simple fact is we can't trust sighted listening as giving valid results - even when it is us listening.

If it is just a case of "do I like what I hear" then it is fine. We can just enjoy it if we like it. If, however we are going to make statements we expect others to use in their own decision making "onboard audio is always bad" then we should really have something other than our own sighted listening to back that up.

 

[Cars-N-Cans]

I do find people who talk about the tonality of an amp as if it is a new breed of apple utterly irritating and have been on the if it measures the same, it sounds the same camp since forever, but these days I do find myself to be more and more skeptical whether that is in fact always true.

Not so much for DACs, but more so Amps I have to say, and that's mainly because I am of the perception that a conclusive test to check the linearity of DACs can be devised but I am not so sure if same can be said for Amps, and whether the measurements with which we have been presented are conclusive enough to prove different amps sound the same under various operating conditions. I really hope this is "just a phase", and as I learn more about them it will all go away

One thing with amplifiers, and electronics in general, is that unlike speakers and headphones its very well defined, and we are able to measure many orders of magnitude beyond what can be audible. Since we have an electrical signal at both ends measurements and performance are much more straight-forward since everything is well defined and pretty every aspect with respect to what we actually hear can be measured. With amplifiers deviations in linearity will show up as distortion, and the nature of those deviations can be viewed by looking at the distortion residuals. This is because deviations in linearity will cause the amplification to vary as well, and a 1 kHz sine in will no longer be a 1 kHz sine out. That's the beauty of using a sine wave is that the equation for a linear system like an amplifier will be output = gain * input, and in order for that to be true, the gain must be constant with respect to frequency and amplitude. Any deviations from that can be easily measured. (Edit: This is a bit of a TL;DR in that actual measurements will be more complex and comprehensive, but you get the general idea. I would say we can have a high degree of confidence that we can measure anything we can hear in electronics.) The one parameter that can vary is the output impedance. Usually with solid state its down in the milliohm territory so to the speakers it will basically look like a perfect voltage source with zero impedance. With tube amps, however, they can have substantial impedance, and this will alter the sound because the speakers and the resistance of the amplifier will form a voltage divider with the speakers impedance variations being imposed on the overall frequency response of the system. From the third edition of Dr. Toole's book Sound Reproduction:

This is why its good to have low output impedance amplifiers. This is especially true of headphone amplifiers where the output impedance can sometimes be many ohms or even tens of ohms. That will fairly dramatically alter the response of some dynamic headphones.

I can understand where the sentiment comes from, that electronics do sound different. I even still like to think that way myself But, objective comparative listening shows otherwise, in that two devices that I think sound different, will ultimately sound the same in A/B listening tests.

 

[IAtaman]

I can understand where the sentiment comes from, that electronics do sound different. I even still like to think that way myself But, objective comparative listening shows otherwise, in that two devices that I think sound different, will ultimately sound the same in A/B listening tests.

No not for me, I have very little trust of a human brain's ability to make objective judgement, and I do not find myself more skeptical these days because I heard "differences" in electronics

That's the beauty of using a sine wave is that the equation for a linear system like an amplifier will be output = gain * input, and in order for that to be true, the gain must be constant with respect to frequency and amplitude.

... and load impedance, and input impedance, and temperature, and supply voltage, and output voltage and output current...

Let me try to give an example to be able to explain what I mean:

This is from the datasheet of TIs OPA 161X series of audio opams. Just as an example, and not beucase this is the most important metric to evaluate for linearity, just look at IMD. Is it -136db or -142db? These are all at Ta=25C, RL=2k ohm, and Vout = 3Vrms by the way. What about Ta=45C, RL=1k and Vout = V supply, what is IMD there? Just becasue I measured IMD to be is -142db for 60hz and 7khz at 25C, 2k RL does this mean IMD is never a problem all across the real life operating range this product is going to be exposed to?

 

[Cars-N-Cans]

No not for me, I have very little trust of a human brain's ability to make objective judgement, and I do not find myself more skeptical these days because I heard "differences" in electronics

... and load impedance, and input impedance, and temperature, and supply voltage, and output voltage and output current...

Let me try to give an example to be able to explain what I mean:
View attachment 264037

This is from the datasheet of TIs OPA 161X series of audio opams. Just as an example, and not beucase this is the most important metric to evaluate for linearity, just look at IMD. Is it -136db or -142db? These are all at Ta=25C, RL=2k ohm, and Vout = 3Vrms by the way. What about Ta=45C, RL=1k and Vout = V supply, what is IMD there? Just becasue I measured IMD to be is -142db for 60hz and 7khz at 25C, 2k RL does this mean IMD is never a problem all across the real life operating range this product is going to be exposed to?

I would say as a rule of thumb, typically when distortion products are more than 50 dB less than the main audio, they can be considered to be completely inaudible. This is about 0.5% THD, and seems to hold fairly well from my experience. Once something is better than that, it’s transparent. While the IMD will vary, it will always be several orders of magnitude below what can be heard at all times in the example given. Electronic designers also have a number of tools to combat these effects so that performance does not vary significantly in terms of audio. It may show up in measurements, but we will never hear it as it’s so much smaller than what we can distinguish. To get back to your original point on linearity, yes it will absolutely vary with amplitude (edit: Oops! I made a mistake in that the rising distortion below is due to noise (the ‘+N’) being higher relative to signal at lower output levels):

However, these figures are well below anything we can hear. Other than noise, audio is actually very undemanding as far as performance goes, to the point of it being embarrassing. While human hearing is very acute, in terms of electronics and measurements we are, for all intents and purposes, deaf as posts compared to something like Amir's analyzers. The high performance is needed for other applications like small signal amplification. Those you will care about parameters varying. However, for audio its actually quite easy to make equipment that is actually transparent to us. I would say in electronics things have to be VERY broken for them to become audible. Usually your biggest problems will be things like noise getting into the audio, and clipping if something is being overdriven. But if you can check those two off the list, then the only things that will dictate the sound quality will be the source material, and the acoustic transducers being used, i.e. speakers or headphones. Unlike electronics, its harder to make linear drivers, and most of the time headphones and speakers will actually have audible distortion in some parts of the audio spectrum at certain SPL levels. They are the dominant sources of nonlinearity in the system. If you are using solid state amplifiers and digital audio as your music source, the electronics are going to be as transparent as the wires that connect them in terms of tonality and detail, and I would not be too concerned as long as you are using competent equipment. I would say with 100% confidence that in electronics there is nothing we can hear that can't be measured. The same is not true for things like headphones, where we cannot measure exactly how someone will perceive them and what the soundstage/imaging will be, only the overall tonality via the Harman target.