[A Jedi]

In reference to the standard 1KHz 2Vrms measurement for amps, DACs, etc., a smart person on that other forum is doing actual science. Some here may be interested. I fully expect this post to be flagged because ... but maybe a few of you will get the benefit before that happens. Here's a link to the outcome:

Spoiler: Spoiler

https://www.superbestaudiofriends.o...ace-old-measurement-new-approach.13754/page-3

 

[bigshot]

That red dot is the approximate location of the most sensitive spot in human hearing I guess. It's an arbitratry point, but the standard 1kHz measurement is so you compare apples to apples. The question is, what do you want the chart to represent? The largest area of distortion, or the most audible area of distortion? Perhaps the best way is to average it all out and come up with a total amount of distortion. I don't know what's best. You just have to learn to read the charts and understand what they mean in real world sound you can hear with human ears.

I'm not quite sure how he arrived at the black box though. That is 10dB below peak across the whole audible spectrum. The loudest part of the track isn't going to be evenly spread from 20 to 20, and whether or not that is what sits at 70 to 90dB depends on how the track is compressed and normalized. And who listens to headphones at 90dB with a +10 peak on top of that? That black box is confusing. Maybe I just don't understand what that represents. The red at the bottom left corner... is that -65dB plus -30dB (-90dB?) And at the red at the upper left corner, is that -55dB (+10 from -65?)

Looking at those charts, even the worst distortion is between -55 and -65dB. That isn't high enough to be audible in normal home audio use. Both those amps would most likely sound the same when it comes to distortion.

But if you make something red and yellow like fire, it's going to look bad no matter how audible it is!

 

[VNandor]

I don't think you'll find much pushback against a more elaborate measurement method than measuring SINAD at one particular frequency at one particular load. Separating the distortion from the noise and then breaking down THD to the individual harmonics and measuring it at multiple frequencies and output levels is going to be more informative than just one SINAD number. An other dimension could be added for separate load impedances. The amp is going to behave differently with a 300ohm load impedance than with a 30ohm load impedance even if either the output level or the output power were fixed.

Personally, this would be the kind of effort I expect from "technical reviews". Single point measurements and sweeps could be done by just about anyone with a decent soundcard but despite that, there aren't many people doing this which is why I think these simpler tests can still garner a good amount of attention.

I'm not quite sure how he arrived at the black box though. That is 10dB below peak across the whole audible spectrum. The loudest part of the track isn't going to be evenly spread from 20 to 20, and whether or not that is what sits at 70 to 90dB depends on how the track is compressed and normalized. And who listens to headphones at 90dB with a +10 peak on top of that? That black box is confusing. Maybe I just don't understand what that represents.

The black box represents how much Vrms is actually being put out by the amp when someone is using a "typical headphone" at "typical listening levels". The rectangle should definitely be tilted though because there is way more voltage being put out at low frequencies compared to higher ones. I think the poster does imply that 90dB SPL on average is about the highest average level one would want to listen it while 75dB SPL is the lowest one. The poster also assumes that a round ~102dB/V sensitivity is what's typical.

The red at the bottom left corner... is that -65dB plus -30dB (-90dB?) And at the red at the upper left corner, is that -55dB (+10 from -65?)

The picture you are looking at is essentially surface graph. It has the X axis, the Y axis and the color. The X axis shows the output level in dBu/volts. Y axis shows the frequency. The color shows the SINAD associated with the dBu and frequency pairs. So the red at the bottom left corner means that:
1) X axis: At -30dBu (0.0245V) rms
2) Y axis: at 50Hz
3) Color: the amp has a SINAD around -66dB.

 

[bigshot]

Maybe 90dB is different with headphones than speakers. With my speaker system, 90dB would be way too loud. 75dB would be loud. I can’t get over 80 without flinching.

 

[71 dB]

Maybe 90dB is different with headphones than speakers. With my speaker system, 90dB would be way too loud. 75dB would be loud. I can’t get over 80 without flinching.

Average level or peaks? 75 dB average level is somewhat loud, but 75 dB peaks isn't.

 

[bigshot]

He's talking about average level isn't he? Because he's allowed 10dB for peaks. Maybe 90dB in a room is different than 90dB in your ears. I don't know. But it seems to me, if you wanted to put a box around the loudest part of most music, it would be the midrange in the top 20dB, not full spectrum at -10dB.

I don't understand how he arrived at the black box. I can read the chart well enough. It doesn't look like either would have audible levels of distortion unless there's something off the scale that isn't showing. Distortion really shouldn't be an issue with DAC/amps.

 

[VNandor]

He's talking about average level isn't he? Because he's allowed 10dB for peaks. Maybe 90dB in a room is different than 90dB in your ears. I don't know. But it seems to me, if you wanted to put a box around the loudest part of most music, it would be the midrange in the top 20dB, not full spectrum at -10dB.

What do you mean by top 20dB and -10dB? In reference to what? The poster on SBAF says that he is interested in the -25dB to -15dB (referenced to 0.775Vrms!) area because that's how much voltage it takes to him to reach his preferred listening levels with his preferred headphones. He mentions this because apparently the 1kHZ sinad measurements are typically taken at +8dBu which he finds not as telling as if the measurement was taken at a lower level, around the output voltage that he (and most likely many others) actually uses.

 

[bigshot]

How is it telling if it isn't audible? A miss by an inch is as much of a miss as a miss by a mile. What I can't hear won't hurt me.

I guess I'm thinking in terms of audible dB, not dBu. When I judge things, I want to try to translate it out of theoretical realm and into real world sound. Distortion up around -40dB begins to get my attention. I don't know what that translates to in dBu, but I would guess that would very from headphone to headphone. I don't know how to translate that into real world sound. So I guess that chart would be mostly in theory unless you defined the headphones you're talking about... and even then, I don't see anything remotely audible. If you understand how to translate dBu to dB do you see anything there that gets into the audible range?

I could be completely misunderstanding though. Whenever I'm presented with a chart with almost no defining of its parameters or analysis in descriptive text, I get lost. It's like, "Here! Look at all that red! It's bad, right? I put a black box around it. That's really bad!" That doesn't really help me to understand what is being measured. Maybe other folks are used to seeing info presented like this. I understand the chart, I just don't see what that black box is intended to represent, or why all of the frequency range within the box is of the same importance.

I'm looking at this again and struggling to try to find audible distortion. Maybe something in the lower left corner of the black box in the "bad one" goes beyond the range of the measurements into the edge of audibility, but that would be sub bass, which isn't the strong point in headphones anyway. I don't know if there would be any model of cans that would be clean enough in the sub bass for the marginal distortion of the amp to ever reveal itself. It seems to me that the distortion of the amp would be swamped by the distortion of the headphones themselves.

 

[VNandor]

I genuinely think that the chart is fairly easy to read at least to people involved in the hobby but I guess it's not as universal as I thought. I would find it super weird if someone published amp specs in terms of "real world sound" instead of electricity.

 

[bigshot]

Well then let me ask an obvious question... Is any of the distortion represented in that chart at an audible level using typical headphones (or even less typical ones)? Would someone be able to hear a difference between the two amps based on the distortion levels alone?

 

[gregorio]

Well then let me ask an obvious question... Is any of the distortion represented in that chart at an audible level using typical headphones (or even less typical ones)?

That’s not an obvious question, in fact it’s not even a relevant question! The relevant question is: Does this measurement scheme give a more accurate representation of the distortion produced at typical operating levels than the existing standard (1kHz @ 2V). The answer IMHO is clearly “Yes”!

The next obvious question would be: “Will manufacturers adopt this measurement?” - And the answer is almost certainly “No”. That’s a shame because now they just have to make their amps perform particularly well at one voltage and one frequency in order to get a competitive measurement, rather than having to compete throughout the freq spectrum at a range of more realistic voltages. If review sites such as ASR adopted it though, that could be interesting!

G

 

[VNandor]

Actually, for less typical headphones, probably yes. The most sensitive IEM i can think of right now is the Shure SE215 which sits at 126dB/V. This IEM would already be quite loud at -30dBu but the chart does not go lower than that. The first chart shows -66dB of THD+N over all frequencies. Note that if the chart went lower, it would most likely show an even worse THD+N. The -66dB of noise might not be completely masked during a quiet part of a well recorded song and especially if the song was made without recording (but still with sampling) and the inherent noise that comes with recording. It could actually be annoying if there was no music playing over the noise. Some people always just slap on their headphones/IEMs when sitting down in front of their computers even if they don't plan to listen to anything right away. The IEM with its noise isolation would make this problem even worse.

The second chart's distortion at -30dBu at 500-1000Hz shows around -53dB THD+N and the 500-1000Hz band area is fairly important. It shows even and even worse -51dB THD+N at lower frequencies. If this was coming from noise that would be not as noticeable as if it were coming from THD. -51dB of noise at such low frequencies might not be audible but a -51dB THD would certainly be under the right conditions. The right condition would mean headphones with low distortion around these frequencies and of course sine test tones. Music with a sparse arrangement and sine bass or even just heavily filtered bass could also reveal such a high amount of distortion.

 

[A Jedi]

For those wondering how much distortion they can hear, take the Klippel test:

https://www.klippel.de/listeningtest/?v=3

 

[danadam]

That’s a shame because now they just have to make their amps perform particularly well at one voltage and one frequency in order to get a competitive measurement, rather than having to compete throughout the freq spectrum at a range of more realistic voltages. If review sites such as ASR adopted it though, that could be interesting!

It's not exactly like in the OP but ASR already does SNR @ 50mV and THD vs power @ different loads, for example ifi Zen CAN Review (Headphone Amp) (Jul 2021):

 

[bigshot]

Just so everyone understands where I'm coming from... I always start out asking "Is this something worth worrying about?" Specifically, is this something that is audible when someone sits down to listen to commercially recorded music at normal listening levels on home audio equipment. I understand that there is a strong faction of audiophiles who keep splitting the difference on better and better specs, even when it results in absolutely no audible difference. They believe that if the numbers are better, it has to sound a little better. I don't believe that. I believe in thresholds of audibility. I don't want audible noise or distortion, but once I reach audible transparency, I don't care how far below that threshold the specs lie.

It may be that there are better ways to measure distortion levels in amps, and that might be interesting to engineers and equipment designers, but if distortion levels never reach audible thresholds, I don't feel the need to measure any further. Better specs beyond transparency makes no difference to me when I sit on the sofa and listen to the Eroica symphony.

So my first question will always be, "Is this audible?" If the answer is no, then I'm done. If the answer is "maybe under the right circumstances" my next question is "How likely would it be that a normal music listener would stumble into those particular circumstances."

It's surprising how difficult it is to get straight answers to those seemingly straightforward questions. People feel the need to concoct exaggerated circumstances or "best, best, best case scenarios" to think up a reason why stuff that really doesn't matter MIGHT THEORETICALLY matter. I don't care about those situations. They just send me back to my first two questions again. If someone is honest and just says, "No, it's highly unlikely that this could be a problem for most if not all home music listeners.", then I'm done and can move on.

Actually, for less typical headphones, probably yes. The most sensitive IEM i can think of right now is the Shure SE215 which sits at 126dB/V.

Is that one of the IEMs with non-standard impedance/sensitivity that needs a specific kind of amp to work properly? I'm not familiar with the two amps being measured here, or the Shure. Does Shure recommend the SE215 be used with the impedance/sensitivity of this kind of amp?

Sometimes IEMs seem to be deliberately designed to not work with a wide range of amps. If you own that kind of IEM, you should buy the amp recommended for it. It isn't the fault of the amp if you use the wrong kind.

This IEM would already be quite loud at -30dBu but the chart does not go lower than that. The first chart shows -66dB of THD+N over all frequencies. Note that if the chart went lower, it would most likely show an even worse THD+N.

Yeah, I was curious about that. It would be good if the chart went into clearly audible territory and made THAT red, instead of labelling "less good but still inaudible in most cases" as red.

The -66dB of noise might not be completely masked during a quiet part of a well recorded song

I'd like to hear a sample where that is the case. In my sig file, Ethan Winer takes a horrible buzzing sound- the worst kind of noise you can possibly imagine- and runs it at descending levels under music. In his test, beyond -40dB the buzz is completely inaudible under the music. Link to his downloadable sample files: http://ethanwiner.com/aes/ -66dB is significantly further down than -40dB. I can't imagine a situation where any kind of noise at that level could be audible with commercially recorded music. The noise floor of the recording studio where the music was made is likely higher than -66dB.

-51dB of noise at such low frequencies might not be audible but a -51dB THD would certainly be under the right conditions. The right condition would mean headphones with low distortion around these frequencies and of course sine test tones. Music with a sparse arrangement and sine bass or even just heavily filtered bass could also reveal such a high amount of distortion.

I'd like to hear an example of that too... with music. I don't generally listen to test tones with my home audio system.

I think the disagreement we have isn't with the chart, it's what the numbers represent in sound. You're basing your threshold of audibility on theory. I find that doing that is a slippery slope into the "a little bit better is always better" syndrome.

If you happen to know how you could set up a test track with the right kind of music, the right kind of noise and the right kind of distortion to take it from the realm of theory into actual practice, I'd appreciate it if you'd share your tracks with me. I haven't been able to set up a circumstance where I can hear anything anywhere near -66dB or even -50dB under music, especially in frequencies at the outside edges of the range. I can hear noise and distortion clearer in the sweet spot of 1 to 4kHz.