-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Master Thread: Are measurements Everything or Nothing?
- Thread starter amirm
- Start date
- Joined
- Feb 23, 2016
- Messages
- 20,802
- Likes
- 37,715
Miska has some credibility with me. Because I've seen him more than once uncover very strange DAC behaviour and it was so if you had the DAC to try for yourself. He has now offered his test signals anyone can download and give it a whirl if they have the D90 SE.
Where I disagree with Miska is the general opaqueness about details like here even when he eventually divulges more of them, and what level of things he says is audible. It is certainly possible he hears better than me or has more experience at hearing faults than me, but he doesn't provide enough for me to say one way or the other in most cases. That is his choice whether I like it or not. However, I don't discount everything he says, and he is often met with that kind of response.
vineyard_aus33
New Member
- Joined
- Jan 24, 2022
- Messages
- 1
- Likes
- 0
I'm of the view it's a combination of measurements, one's physiology, gear and environment. And that measurements vs perceivability is an entire subject that could be separated until it is fully understood
1. I'm fairly certain I can hear a difference between lossy and lossless, ripped from the same source of course (but only some songs)
- I'd think that since most if not all lossy codecs employ some psychoacoustic modelling and considerations, it could depend on the recording, ie. If a track has more content that has been deemed to 'stay' or 'removed' by the modelling
2. I'm thoroughly annoyed when it's summer here (35C+) and i just feel all music sounds trash in that kind of environment
- Yes i know temperature vs air density etc. I'm not talking about the measurement, i just, don't enjoy any music in those kind of heats, no matter how good the system or recording is, even if the entire system is re-tuned to create a replica of what it would do at 21C... just nah.
3. Ear structure?
- Pushing the top of the Helix forwards from behind yeilds a stronger midrange
- Pushing the Pinna forwards from behind yeilds a stronger midbass
- Pushing the Earlobe forwards from behind yields a snappier, fun sound
- I'd think that since we all have different ears, like our fingerprints, the end perception of sound could be skewed greatly
- FaceID my ears and inner canals with your infrared voodoo if you need to Apple, let me know the difference between what Amir and myself are hearing based on our physiology at least! So that i know when Amir says bright, i might know this as snappy, and now i know that on the graphs he's referring to that 3dB bump from 6-9khz which to me becomes 5-7.75khz gddammit
4. Past conditioning?
- I've been brought up in a pretty silent environment, all audio sources were rather warm sounding (aha low quality), all my family members have low, smexy voices.lol. So i can't stand anything in the vicinity of 'shrill, sweet, tingy' etc. Basically, no triangles, no bagpipes and some chihuahuas
5. Stuff sounds different to me after right after waking up, after an exercise in the evening, or after some alcohol and the other good stuffs. Aha. So to me, I know the graphs and numbers say one thing, but when i be vibing, someone gotta be graphing my brain on an electrogram or something, it's just another level. Glad i can cue my personal build-in atmos/HRTF processor by being me or ingesting some things and it changes the electrical impulses which changes the percept...... oh no. no freaking way...
6. I think measurements are EVERYTHING if we could make full sense of what the measurement is intending to tell us. All those quantization tests, reviews between between lossy formats, what do they mean? Do they mean that MP3 discards 5% information between -30 and -22dbfs at 256hz, for a set time per each 441 samples? And then does it at 10dbfs lower at a bigger delta at the next harmonic? I mean, that's probably not true. Yes it's not bitperfect... but how... exactly where and exactly what changes?
- After all, if you close your eyes and get someone to snap a finger at dead-center of your nose, pretty sure you could identify it's center. Move the finger 2cm to the right and snap again, you could possibly tell the difference. A difference of 0.09ms in arrival time (and the other bunch of room reflections) to your other ear has allowed you to spot this difference
- 44100hz = 441 samples per ms right? No. I want to know whats going on each 90000nanosec at least. Which is how many samples? lol
- I'd like to think we all have golden ears, all good up to a point, thereafter we just can't tie it to the data we are seeing and this is the missing link
- Or perhaps the data we collect is not nearly enough or the one we need, or we havent discovered it yet?
- And instructions are unclear. When they say we can't detect a 0.3db difference in volume change, does it mean 0.3db at a single frequency? Or does it mean.. go ahead, chop 0.3db between 250hz to 8khz on an EDM track, no you wont hear a change at all.
- If we can have a high speed camera filming the speed of light on at 1 trillion FPS for research proposes (ala Youtube) or how a vinyl breaks apart at 10000fps, I'd think we'd kinda need analyzers at the a sample rate of 50000khz to fully understand what we are studying no? You read that right. Show me what a fifty million sample rate RTA is seeing for a 2 second excerpt of a track stretched out to fill 10 minutes of analysis
7. Measurements are NOTHING as well because of enviroment, factors that are not taken into account (or thought as no perceptible difference), physiology etc. I'm just a casual audiophile. I use the measurements to tell if something is 'good enough' or 'within the ballpark of what i'm looking for'. Then i'd use my ears to tell 'if this goes well with that'
1. I'm fairly certain I can hear a difference between lossy and lossless, ripped from the same source of course (but only some songs)
- I'd think that since most if not all lossy codecs employ some psychoacoustic modelling and considerations, it could depend on the recording, ie. If a track has more content that has been deemed to 'stay' or 'removed' by the modelling
2. I'm thoroughly annoyed when it's summer here (35C+) and i just feel all music sounds trash in that kind of environment
- Yes i know temperature vs air density etc. I'm not talking about the measurement, i just, don't enjoy any music in those kind of heats, no matter how good the system or recording is, even if the entire system is re-tuned to create a replica of what it would do at 21C... just nah.
3. Ear structure?
- Pushing the top of the Helix forwards from behind yeilds a stronger midrange
- Pushing the Pinna forwards from behind yeilds a stronger midbass
- Pushing the Earlobe forwards from behind yields a snappier, fun sound
- I'd think that since we all have different ears, like our fingerprints, the end perception of sound could be skewed greatly
- FaceID my ears and inner canals with your infrared voodoo if you need to Apple, let me know the difference between what Amir and myself are hearing based on our physiology at least! So that i know when Amir says bright, i might know this as snappy, and now i know that on the graphs he's referring to that 3dB bump from 6-9khz which to me becomes 5-7.75khz gddammit
4. Past conditioning?
- I've been brought up in a pretty silent environment, all audio sources were rather warm sounding (aha low quality), all my family members have low, smexy voices.lol. So i can't stand anything in the vicinity of 'shrill, sweet, tingy' etc. Basically, no triangles, no bagpipes and some chihuahuas
5. Stuff sounds different to me after right after waking up, after an exercise in the evening, or after some alcohol and the other good stuffs. Aha. So to me, I know the graphs and numbers say one thing, but when i be vibing, someone gotta be graphing my brain on an electrogram or something, it's just another level. Glad i can cue my personal build-in atmos/HRTF processor by being me or ingesting some things and it changes the electrical impulses which changes the percept...... oh no. no freaking way...
6. I think measurements are EVERYTHING if we could make full sense of what the measurement is intending to tell us. All those quantization tests, reviews between between lossy formats, what do they mean? Do they mean that MP3 discards 5% information between -30 and -22dbfs at 256hz, for a set time per each 441 samples? And then does it at 10dbfs lower at a bigger delta at the next harmonic? I mean, that's probably not true. Yes it's not bitperfect... but how... exactly where and exactly what changes?
- After all, if you close your eyes and get someone to snap a finger at dead-center of your nose, pretty sure you could identify it's center. Move the finger 2cm to the right and snap again, you could possibly tell the difference. A difference of 0.09ms in arrival time (and the other bunch of room reflections) to your other ear has allowed you to spot this difference
- 44100hz = 441 samples per ms right? No. I want to know whats going on each 90000nanosec at least. Which is how many samples? lol
- I'd like to think we all have golden ears, all good up to a point, thereafter we just can't tie it to the data we are seeing and this is the missing link
- Or perhaps the data we collect is not nearly enough or the one we need, or we havent discovered it yet?
- And instructions are unclear. When they say we can't detect a 0.3db difference in volume change, does it mean 0.3db at a single frequency? Or does it mean.. go ahead, chop 0.3db between 250hz to 8khz on an EDM track, no you wont hear a change at all.
- If we can have a high speed camera filming the speed of light on at 1 trillion FPS for research proposes (ala Youtube) or how a vinyl breaks apart at 10000fps, I'd think we'd kinda need analyzers at the a sample rate of 50000khz to fully understand what we are studying no? You read that right. Show me what a fifty million sample rate RTA is seeing for a 2 second excerpt of a track stretched out to fill 10 minutes of analysis
7. Measurements are NOTHING as well because of enviroment, factors that are not taken into account (or thought as no perceptible difference), physiology etc. I'm just a casual audiophile. I use the measurements to tell if something is 'good enough' or 'within the ballpark of what i'm looking for'. Then i'd use my ears to tell 'if this goes well with that'
Last edited:
You’ll note I did not say he was wrong. I said he had no reason to expect people to take his word for audibility levels at below -120 dB without being willing to share his evidence, and no reason to be offended when people challenge his unevidenced conclusions.
It’s about instincts. Does he want to get to the bottom of the D90SE behavior? Does he want to figure out why the effect he’s claiming is audible despite being far below the audibility levels observed in other research? Or is he just using it as a stick to beat ASR writ large over the head? I see much more of the last, and it casts doubt on his motives.
Rick “understanding the difference between motive and intent” Denney
- Joined
- Feb 23, 2016
- Messages
- 20,802
- Likes
- 37,715
As an odd phenomena I don't even know is true, I mentioned a Sony developer earlier in the thread.
In the early days of sigma-delta ADCs and DACs he found noise floor modulation over a certain time scale audible while the level of such things was 90-100 db down from 0 dbFS. He setup and demonstrated he could hear this blind. He showed it to some of their recording professionals and once shown it they too could hear it blind with music. None of the fine details of that are public.
What was happening was an interaction with early forms of dither and certain modulation chains. They hypothesized that when you walk into a room your ear in a few seconds creates a filter letting you ignore much of the effects of a room. The problem was early dither was only pseudo-random (which actually I suppose all such dither is). So with these early sigma delta parts it was slowly altering the low level noise floor and very slightly modulating it with music in a repeating pattern. So that over a couple minutes it was like being in a room where the dimensions were slowly morphing back and forth in a pattern that repeated. If you listened to minimalist recordings you could hear the apparent size of the space waver around a bit over a couple or three minutes.
The fix was to use dither much closer to being fully random with a pattern that wouldn't repeat for much longer periods of time. They never investigated if their hypothesis of why you heard this was true, but once they figured out the right dither made it go away no one bothered otherwise. Modern dither and modern chips aren't going to have this effect so no way for us to see if it was truly audible for most people or not.
Are there any other odd things like that out there to hear? Certainly anyone who claims it should provide something for us to try it out for ourself if they expect to be believed.
In the early days of sigma-delta ADCs and DACs he found noise floor modulation over a certain time scale audible while the level of such things was 90-100 db down from 0 dbFS. He setup and demonstrated he could hear this blind. He showed it to some of their recording professionals and once shown it they too could hear it blind with music. None of the fine details of that are public.
What was happening was an interaction with early forms of dither and certain modulation chains. They hypothesized that when you walk into a room your ear in a few seconds creates a filter letting you ignore much of the effects of a room. The problem was early dither was only pseudo-random (which actually I suppose all such dither is). So with these early sigma delta parts it was slowly altering the low level noise floor and very slightly modulating it with music in a repeating pattern. So that over a couple minutes it was like being in a room where the dimensions were slowly morphing back and forth in a pattern that repeated. If you listened to minimalist recordings you could hear the apparent size of the space waver around a bit over a couple or three minutes.
The fix was to use dither much closer to being fully random with a pattern that wouldn't repeat for much longer periods of time. They never investigated if their hypothesis of why you heard this was true, but once they figured out the right dither made it go away no one bothered otherwise. Modern dither and modern chips aren't going to have this effect so no way for us to see if it was truly audible for most people or not.
Are there any other odd things like that out there to hear? Certainly anyone who claims it should provide something for us to try it out for ourself if they expect to be believed.
- Joined
- Feb 23, 2016
- Messages
- 20,802
- Likes
- 37,715
You don't understand the sampling theorems. It lead you to make several erroneous conclusions or ideas in your post. Like how many samples are needed to hear that finger snap has moved. Or the idea you think we need 500,000 khz sampling (which a few devices now exceed btw). Or the need for a fifty million sample RTA (which you could create with REW btw).
Go watch that Monty Montgomery video about digital show&tell part 2.
All of those physiological and psychological influences can be controlled, and by controlled mean accounted for in testing. A simple ABX test is usually enough, because those effects don’t change from one sample to the next rapidly enough.
I thought I could tell the difference between high-rate lossy and lossless, too, until I took my own blind ABX test. I couldn’t tell the difference even at lower bit rates, even listening for quiet artifacts, without extreme gain-riding. Those tests are available online. I’m sure there are those who can, but evidence suggests it requires extensive training to do so, and unrealistic listening techniques.
Yet routinely we hear that these are night and day differences.
Rick “lots of interest ABX tests online” Denney
@Miska: can you try to explain this: if you look ook at the IMD graph of the D90SE:
Why is the SINAD not worse at the -60 generator level than any of the other SOTA DAC's? If the issue you found was present here, should it not show here as well? Or does it disappear when playing two tones? I'm not 100% sure, but I think these measurements are done at 44.1/24 (@amirm can you confirm?)
Why is the SINAD not worse at the -60 generator level than any of the other SOTA DAC's? If the issue you found was present here, should it not show here as well? Or does it disappear when playing two tones? I'm not 100% sure, but I think these measurements are done at 44.1/24 (@amirm can you confirm?)
Last edited:
solderdude
Grand Contributor
A lot of this has already been researched. Perception is a bitch. It changes depending on many factors.
Depends on the used codec, bitrate and quality settings and is very dependent on training and recordings themselves.
One can't generalize lossy codecs and circumstances.
The mood changes so perception changes.
Yes, pinnae and ear canals differ. The rest of the auditory system also is likely to differ in certain aspects.
Here's the thing though.
2 people, one with outwards protruding big ears and another person with very small pinnae as good as flat against the head are both listening to a girl with guitar in the middle of the street.
We measure the response at the ear drum. Not surprisingly the response will be different.
The question is what both listeners actually hear.
The answer is simple they hear a girl with guitar and see and hear the 'truth'.
This is what the brain (hearing process) 'calibrates' to. Hearing everyday sounds are regarded as truth.
Put some filtering like dense cloth (or push your pinna and change filtering that way) and you will immediately hear differences.
Slowly change your hearing abilities (aging) won't register as such. At 60 we still hear the world the same as when we were young. Yet we don't.
Hearing is complex.
If it would really work this would not matter as long as both individuals use their own 'correction'.
The bright vs snappy bit is a vocabulary thing and mutual agreement on how to describe sounds.
Preferences. Fortunately we have tone control for that.
Every one is experiencing this. Some are more aware then others. Some think they can 'hear past all this'.
It is one of the hurdles in perception and especially long term perception and 'remembering' sound.
yes, the snag is in fully understanding all relevant measurements. Only very few people, often those that actually do measuring and live measuring, fully understand. It is the main reason why many folks say measurements say nothing. Because they do not understand the implications or do not combine all relevant measurement results. Then there is the snag of visualizing measurement results and understanding test conditions.
Nulling will show the differences but the audibility of those measured differences is totally a different matter.
What proper codecs should do is remove that what is supposed to be masked by other signals so that signal does not have to be encoded as well. The lower the available datarate the more needs to be removed. At some point, depending on the recorded material, this becomes audible.
record the fingersnap using binaural recordings. listen to it with a good headphone and record this on 44.1/16.
In both cases (reality and recording) close the eyes.
You will find 44.1/16 is perfectly capable of recording the timing differences.
As Blumlein already mentioned... it seems like you are not getting the sampling theorem. Few people do.
Conversions from acoustic to electric and electric to acoustic are problematic for tons of reasons. In this case the measurements still have meaning. They don't mean nothing, they just aren't equally conclusive as electrical measurements.
Sgt. Ear Ache
Major Contributor
Lol, one can easily imagine the thought process going on there. Fremer taking the test is thinking "oh my these sound exactly the same...this is impossible" but he makes his choices. Then, on discovering he'd hit on all 5 - "YES! Holy crap I'm never doing that again!"
I mean it's not particularly incredible to flip a coin 5 times and get 5 heads. I just went over to flipsimu.com and flipped a coin 30 times. After the first 20 I was a perfect 50:50 - 10 heads and 10 tails. The next ten flips - 9 heads and 1 tail.
Keep in mind- self reported.
Sgt. Ear Ache
Major Contributor
Sgt. Ear Ache
Major Contributor
This being the case, it seems to me that relying on the subjective impressions of a bunch of individual audiophiles on youtube and elsewhere for gear assessments makes little sense. I mean not only are their rooms and setups completely different from your own, but their very EARS are completely different!
Perhaps a solution to this conundrum might be to find some objective way to assess what the gear is doing so that we can at least know with confidence (and without relying on our endlessly variable ears) what is happening prior to the signal reaching our speakers. Then, maybe somebody - some third party that has the resources necessary - could conduct a bunch of experiments where people listened to a bunch of different speakers (blind so that they aren't tempted to pick the prettiest one) and selected the ones they felt sounded the best. Then, you could examine (measure?) all those speakers and try to determine if there were commonalities between the ones that seemed to be most preferred. Maybe after having done a whole slew of that sort of research it might be possible to arrive at some sort of broad consensus regarding what makes for a good-sounding speaker.
From my understanding, this was part of a mass demo at an AES meeting rather than a formal, well-controlled test. I contacted the organizer of the demo many years ago when this was being strutted around as something significant. "No, we did not get his score sheet in advance of making the key public." John Atkinson claims to have witnessed this, for what that's worth.
Putting aside the questionable evidence of this claim, the fact that in a group of 100 or so people answering randomly one would expect a couple 5/5 scores was waved away. "Isn't it curious that one of those 5/5 scores was achieved by a Golden Ear?"
This is the kind of self-promotion I can't take seriously.
@Miska: can you try to explain this: if you look ook at the IMD graph of the D90SE:
Why is the SINAD not worse at the -60 generator level than any of the other SOTA DAC's? If the issue you found was present here, should it not show here as well? Or does it disappear when playing two tones? I'm not 100% sure, but I think these measurements are done at 44.1/24 (@amirm can you confirm?)
Does that include all wide-banwdith noise or just IMD distortion harmonics? At least based on the graph it is about the IMD harmonics. So I rather look at pure spectrum analysis.
And as mentioned earlier. The DAC supports 28 source formats. So those measurements should be repeated for each. It may be different with 44.1/32 input vs 44.1/24 input too. As an example, I just found out about bug in iFi's XMOS firmware causing 24-bit input formats to be broken, but 32-bit input formats work fine (so 24-bit data needs to be zero-padded into 32-bit words for those). And I think many XMOS chips run out of steam at higher rates when they have the MQA code which burns extra cycles for the MQA detection, etc.
DDF
Addicted to Fun and Learning
- Joined
- Dec 31, 2018
- Messages
- 617
- Likes
- 1,360
This is a double edged sword. It's true that those who claim audibility should demonstrate proof. But there can be a reciprocal fallacy in assuming gear is faultless if it passes a few limited use case audio tests. ASR has uncovered a few oddities in gear over the last couple of years, but hearing itself also plays tricks that can be stimulated by signal or playback chain use cases a few simple tests can't explore.
One of my favourites is the work by Schroeder and Mehgart "Auditory Masking in the perception of speech". They showed that they could make the threshold for perceiving distortion 30 dB more sensitive by altering the phase relationship (playback chain's group delay) between a distortion component and a speech like masking signal. Changing the phase changes the signal envelope and allows the distortion component to "peak out" when the signal envelope is low.
There are all sorts of other context and signal dependent auditory effects that are odd. Diana Deutsch (some of her research here) wrote an article in the JAES, Vol. 31, Num. 9, Sept. 1983, "Auditory Illusions, Handedness, and the Spatial Environment" where she presented two repeating octave spaced tones via headphones so that the left ear receives the high tone and the right ear receives the low tone: right handers hear the high tone in their right ear and the low tone in their left ear, left handers hear the high tone in the the left ear and the low tone in the right ear. Wild.
I guess that's why I'm so sensitive to ASR posters constantly jumping all over people that make audibility claims. Sure, many are imagined, but once you study the cognition science literature long enough, in some cases, reserving judgement feels like a healthy alternative.
Fremer also supposedly quickly blind AB’d a high end cable (at an audio show, where else). So he’s taken these two results and retired from blind testing with a perfect record.
I thought he also claimed to have aced the amplifier comparison at an audio show that was written up in Stereophile (in a statistically questionable way, originally).
(more here:
Blind Listening
John Atkinson sets the stage Nothing seems to polarize people as much as the vexed question concerning the importance of audible differences between amplifiers. If you think there are subjective differences, you're an audiophile; if you don't, you're not.
www.stereophile.com
Post in thread 'If "Tube Sound" Is a Myth, Why Tubes?'
https://audiosciencereview.com/foru...be-sound-is-a-myth-why-tubes.8656/post-594714)
It’s where his bread his buttered, we probably shouldn’t expect better.
Finally, it turns out Stereophile called him David Clark and I somehow parroted it.
The Blind leading the Deaf Letter: Why Amplifiers Sound the Same
Letter: Why Amplifiers Sound the Same
www.stereophile.com
Which contains this gem from Harvey Rosenberg. ROFL.
Last edited:
Sgt. Ear Ache
Major Contributor
hahaha...oh boy. That's good stuff...
I contacted Archimago and he added some D90SE tests.@Miska: can you try to explain this: if you look ook at the IMD graph of the D90SE:
Why is the SINAD not worse at the -60 generator level than any of the other SOTA DAC's? If the issue you found was present here, should it not show here as well? Or does it disappear when playing two tones? I'm not 100% sure, but I think these measurements are done at 44.1/24 (@amirm can you confirm?)
REVIEW & MEASUREMENTS: Cayin RU6 R-2R USB Dongle DAC/Headphone Amp. Considering "Intentional Imperfections". (Topping D90SE -60dBFS input graphs.)
A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.
archimago.blogspot.com
Similar threads
