Nope, that has not been established so generally(!) and definitively in the research. That kind of conclusion seems to be entirely based on lazily generalizing a simple comparison of single-number THD metrics - without looking at the comparative spectral distributions - between transducer and source device. We have already agreed that IMD is far worse yet you’re going around claiming transducers are the weakest link in the chain, all while IMD in transducers is so damn low nobody even bothers to measure it.
Look, I’m not going point by point again, I will just share this recent-seeming article on RTings (with dr. Olive consulting, apparently) because it supports the same essential idea I’ve been trying to push in all of this: we don’t have good enough measures of the audibility(!) of distortion, nor in some cases even the true level of distortions (due to measurement gear limitations as mentioned for example in this SBAF post) to be going around hammering people over the head for claiming to hear the levels of distortion they’re hearing from the gear they say they can hear it from, placebo effects notwithstanding.
So the article from Dr. Olive is a very good one, and points to precisely the things I’m talking about re the masking window. But I don’t see how this relates to your claim that the notion “transducers are a worse source of harmonic distortion than typical headphone amplifiers” is false - if that is indeed your claim here.
Like… this point about higher harmonic distortion being more typical of transducers has nothing to do with the problems of summing THD metrics, as you’re suggesting is the problem here, and I fully agree that it is a problem.
Yes, the increase in sensitivity means that some amps that were previously not good enough now are.
A cheap amp option that should do better now is the Schiit Midgard.
For a single DAC/amp solution, the Schiit Jotunheim is still an extremely solid pick (esp for the 400$ cost, a bit more with the Mesh DAC add on) which is powerful enough even for the Tungsten V1 as the amp puts out 1.2W into 300ohms.
Also to be clear… THEY might not have good enough equipment to measure this stuff at this SPL, and the post even goes so far as to note this. But for headphones, a GRAS capsule is actually extremely precise for this kind of thing. It’s also notably better than even the more advanced B&K 5128 system in this regard.
And, for source equipment, you don’t even need the APX555 or something like it to get extremely precise measurements, there are less expensive, albeit less ‘official’ ways to do this. So as far as metrics are concerned, we’re not limited by what we can actually measure. We’re limited by the analysis of what’s being captured and how it’s being impacted by other variables.
“It can be done (today)” is not the same thing as “it has been done so routinely that that and only that is the quality of evidence all these online/forum authorities are basing their pronouncements on”. People could be working on a combination of good and bad measurements made in the past, and teaching newbies… not exactly the most up to date conclusions.
Aaanyway, maybe this mile-long off-topic detour should be split off as its own thread (or merged with the existing thread “Distortion”?), as it will probably remain a subject of interest for the foreseeable future.
Yeah but is anyone here actually doing that? Like… did you read the rest of my post there about higher order products, the masking window and IMD?
Maybe I’m wrong about this but I feel like these days the only THD summing you get is from the weird SINAD absolutists - or places that aren’t actually taking any of this stuff seriously in the first place.
At the bottom of the stack of turtles holding up the Earth lies human perception. Psychoacoustics and philosophy of mind tackle perception, and as such they form the primary analysis.
Every electrical measurement derives from and serves human perception. There is no ultimate God’s Eye perspective on the outside world, but we do have some really good models.
Basically yeah. Measurement tools aren’t perfect, but as far as the ‘what’ of it, we can measure all the things that would be perceptually relevant. It’s the interpretation and understanding that’s lacking.
Now since you mentioned this, and since I happen to be afflicted by a tragic inability to touch GRAS , can you estimate whether the GRAS model you’re thinking of would be able to measure the harmonic distortion from some representative S-tier headphones played at, say, 60 dB RMS?
Sure, but again… even if you’re listening at 60dB on average, if you’re interested in understanding that device’s distortion performance, you’ll want to test it at much higher SPL. If you’re interested in what’s audible at 60dB specifically, that’s a different question, but also one that’s particularly unrealistic. Like the purpose of understanding harmonic distortion should not be to answer the question “how much distortion does this thing have?”, it’s to answer the question “at what SPL do I need to care?”
Sure you could estimate, or sure it would measure the distortion (without hitting noise floor limitations)?
How is that unrealistic when that’s the level I listen at in my best-case quiet environment when I’m in that critical-listening ‘zone’? That is the reference scenario where all those extreme SINADs have a chance to come into play, that’s when the DAC and/or amp’s noise and distortion could surpass that of the headphones, or not, and that’s when I’m best positioned to hear it. That answers the question “what source gear SNR/THD/IMD delta is worth spending extra money on”.
LE:
Oh you know what, that’s just a guesstimate I keep repeating based on those dB charts comparing each level to everyday noise sources, but I never took into account the “missing 6 dB” of headphone listening. So let’s call it 66 dB.
Well, distortion is related to level, so while you can get a reading that’s still above the noise floor, it kind of doesn’t matter because 60dB is extremely easy for transducers to achieve and most people are not listening at 60dB.
Have you actually tested your listening level with calibrated industry standard equipment? Because 60dB is rather quiet. But beyond that, there’s still a misconception here about the average listening level and what SPL threshold would be relevant for distortion products. If you have a 1khz tone at 60dB, that’s not 60dB peak, or with music.
But for the theoretical example for where I think you’re going here… Say your playback is sufficiently low to where the transducer is not generating any audible distortion products. Then sure, it may matter to look more at products generated by the rest of the equipment. Like I think even in less extreme examples, as I said earlier, things like IMD matter for these devices, and it should be tested.
I just have a hard time believing that people are actually going to be listening this low. And as far as review practices are concerned, it makes perfect sense to see where those distortion thresholds actually occur, even if on an individual basis you’re not listening that loud. If most people are listening around 75-80dB average, it’s not unreasonable to test the transducer at 90-100dB.
I should also note, it also kind of depends where in frequency the distortion products occur - as Dr. Olive’s article above also indicated. Higher frequency products are going to be more audible than say 2nd order stuff in the bass.
You are discounting the environmental background noise of any room (e.g., ~30 dB). The 60 dB signal in an ultra-quiet room still must overcome general air movement.
This is not reference listening, rather, this is using background noise as a de facto way to mask nuances.
Of course not, same situation as with my access to a GRAS. Best I could do is I could try to compare it to playing the same music through speakers and measuring that with an SPL meter at my position. Should be whatever that comes out to +6 as my headphone loudness.
No, but if that’s the primary reason for picking that level it’s still weird. You’d expect them to measure at 75-80. No, I think they picked 93-94, 104 and so on just because they knew they had bad microphones that couldn’t reliably go below that - Solderdude pretty much admits to it in his article about reading distortion graphs, then you saw that other guy on SBAF diagnosing his own lackluster setup… Amir might be that one special case where he’s testing at his real listening loudness (considering his headphone review comments where he’s pushing everything so loud it’s making his ears flap etc.).
Also true. The signal has no problem overcoming the background noise at 60 vs 30 (or 20 as I’ve measured it at night), the question is will the source device N+D add anything detectable to that. I think any additive component has to be a good -15 dB vs. the primary one to not make an audible difference, so -40 + -15 = -55 dBr should be good enough for me to not hear anything bad if I’m right about my listening loudness in that scenario.
Thing is, plain noise I don’t think masks certain tone-based distortions very well, they could be audible even lower than -15, I don’t know. That’s why that continued research RTings were talking about is still needed - we have to find out how to apply perceptual weighting to all of these things.
It’s my reference listening, if turning things up past that level makes the experience annoying and not worth continuing. But it’s true avid IEM users a couple of decades younger might be even better positioned than me to protest that TOTL source gear could make a difference for them as advertised since their earplugs also protect them from most of that environmental noise. Especially the ones who have discovered foam eartips.
It would be rare of me to come to the defense of Amir on any of this because I have an extremely low opinion of him as a human and his value system for product evaluation generally… it’s typically a narrative-led skewing of the science for the purposes of playing authority games. But when it comes to testing for peak - and it’s likely I’ve criticized him for this in the past - it’s not unreasonable to go quite a bit louder than your average listening level. The thing is… what he’s missing is that the ability to hear those distortion products decreases as the fundamental SPL goes up. So in those instances where he’s hearing distortion products on a test fixture… he ain’t hearing it with music if he were to actually listen to it at that level.
As to the idea that they’re testing this loud because they have bad mics, I think maybe other folks doing it have this problem, but Amir doesn’t. He’s testing at 114dB or whatever crazy volume threshold because that’s where those transducers to incur their excursion limits or reveal themselves to have massive distortion problems. It could be a matter of edge case edging… or maybe to him there’s a practical benefit to be able to say “this product is bad” if he doesn’t like it, when there’s nothing otherwise wrong with it, or maybe he genuinely believes it matters at that SPL.
But either way, it’s another category along which a product can be evaluated - to see where a transducer’s limits are is still a useful thing to know even if you’re not hitting them with your typical listening level. I like to be able to know if something has an excursion limit, because that’ll impact whether or not I can EQ the bass to the high heavens or whatever.
I think the core ideas put forward here though are to do with whether you can dismiss harmonic distortion metrics in headphones because you’re not hearing them anyway at lower listening volumes. While that’s generally true… I’m not sure this then makes a case that we need to care more about source equipment distortion metrics. Though maybe I could be persuaded otherwise should more research emerge on thresholds of audibility or preference for low order products.
No no, I just want to know the performance of the headphones at real listening levels so I can properly compare it to that of different source devices and say more confidently “this amp is worth buying, that one is not”. I don’t know which class of device to give the priority until I have the right human-centric numbers.
Got it. Well I think this is an exercise in futuility to be honest, though I suppose I get why you’d want that. So much of this is dependent on the crest factor of the music you’re listening to anyway. But in short… I don’t think you’ll hear distortion products either way with most competent devices. The exception being to do with particularly nasty IMD stuff inside the audible band, so worth watching out for that.
Your question pertains to personalized testing. There is no ground truth for everyone, and each person falls somewhere in the performance envelope (potential range) of a human. Look into Signal Detection Theory and receiver operator characteristic (ROC) curves. It was a big deal for the military with radar data interpretation and whatnot in WW2, and sparked a bunch of follow up work.
The human senses are probabilistic. You can’t establish your personal ROC curve from anyone other than yourself.
, can you estimate whether the GRAS model you’re thinking of would be able to measure the harmonic distortion from some representative S-tier headphones played at, say, 60 dB RMS?
Best I could do is I could try to compare it to playing the same music through speakers and measuring that with an SPL meter at my position. Should be whatever that comes out to +6 as my headphone loudness.