Our auditory perception pays particular attention to the attack segment of a sound (we need to categorize the sound as early as possible, for obvious reasons). My premise with noise modulation is that the noise floor dynamically modulates (maybe in correlation with the music signal or maybe loosely correlated to it, not sure?) and if the noise floor has been increased due to processing of a current demanding signal & this hasn't fallen back to base level when the system is now processing a soft sound, the timing of the start of the attack of the gentle sound is less clearly perceived - perceptually, this leads to the perception of a less defined sound stage - the sound objects are just not as solidly placed in the space (out ITDs are not as sharply defined in time).
This is pretty much what Mallinson (when ESS chief engineer) said in his video of the state variables not returning quickly enough after a loud sound & it's perceived but not by him (he didn't specify how it was perceived by those who do).
This is pretty much what Mallinson (when ESS chief engineer) said in his video of the state variables not returning quickly enough after a loud sound & it's perceived but not by him (he didn't specify how it was perceived by those who do).
Yes, it would seem so! Was there any follow up in how the preferred one measured compared to others?
Last edited:
Sorry I was referring to the BBC reporting on Mr. Smith's demise (twice).
I talked to Martin years ago, and I don't think they did any tests that would make anyone happy.
That all seems more like digital filter mysticism. By definition nothing can happen before the event except as a consequence of filtering and associated delays. Energy decaying in electronics seems unlikely. Its possible that an amp could have higher noise at a DC point away from zero. I'm not sure anyone has looked but Scott may set us straight. Digital filters with many taps could do something. Energy decaying in transducers is very likely and probably at a much higher level than in any electronics at least in all of my measurements of transducers.
That seems like a pretty easy thing to test for. Fundamentally it's like testing for settling time. Analog tools to test to .01% (-80 dB) have been around for a long time. Verifying its audibility would be more difficult but creating a simulation of it is not that challenging. Take a recording and mix modulated noise per the supposition at different levels and see at which level it becomes audibly obvious. The dreaded Foobar ABX test would be an easy way to explore it.
Normal curve is always Gaussian. You can not say all people can do football as good as Lionel Messi or Chistiano Ronaldo.
And that includes all of the grandpas hit by tinnitus. But they are very experienced listeners and remember exactly how it used to sound 40 years ago for comparison.
Yes, dynamic noise floor is a differentiator of systems, IOW systems that measure similarly low static THD/IMD but 'sound' different.
Lowered noise modulation does result in change in precision of 'solidity to the soundstage & solidity to individual instruments/voices in that soundstage' and 'a more believable sound with realistic perceived dynamics'
You mention 'remove' noise modulation......how to achieve this ?.
Yes, a suitable test signal standard would be handy (maybe), in the meantime comparing loopback recordings of deliberately changed system noise floors is informative.
Yup, the system and the final output acoustical transducer excess noise dynamic noise floor has Attack, Sustain, Decay characteristic and according to the device/system and the program and the program embedded noise.
The fun part is when you can A/B all of this OTF, spectrally different Noise Floor ADSR equals subjective changes that can be good and/or bad.
Passive and active component choices and cables and connectors and power quality all affect/effect system noise floor and consequently the final in room (kitchen/headphone) sound.
Tweaking the above items as is the 'accepted norm' and can help to dial in an enjoyable system sound but it is by definition a crap shoot.
There are other ways to achieve preset definable system sounds or 'no sound' but that is getting firmly into Bybee territory.
Dan.
Any natural music signal has "attacks" in orders slower than a step from the generator (1000x and more) that is used for amplifier testing. Music signal is the easiest signal for the amplifier to work with and testing by sines, twin tones, multi tones, noises, modulated narrow band noises tells much more than testing with any natural music signal.
So, 1/. Propose a way to measure this, and do it. I'm sure others will then repeat your test.
And 2/. What about mono, if we are not interested in the stereo illusion?
Surely none of this is an issue for an amp that can operate into the GHz range?
Now, speakers, that's a different story of course
Interestingly when a test was setup to try and see if there was a prefernce between the Revel Salon 2 and the JBL M2 that was done initially in mono with one speaker. People preferred the Salon, which actually didn't suprise me.
It wasn´t meant to criticize you, more to show that even when taking the maximum number possible, it still could be true that all belong to the mentioned "5%".
I must admit not having read GedLee´s statement myself, so can´t swear which way he meant it, but usually such a comment is related to the population overall. At least i am not aware of any sensory group test taken on the members of diyaudio.
We can´t break down the percentage from the population estimate to a subgroup of members as we don´t know if this subgroup (not even all the members) are sufficiently representative for the population.
Instead it is not unreasonable to assume that humans more interested in audio quality are engaging in such a special interest group and that indeed a higher percentage have better listening abilities.
As said earlier, the group "others" was i did not expect before when doing listenng tests with others. We had a few people with excellent listening abilities although untrained but having zero interest in audio reproduction quality.
Again, as stated before, i understand a lot of these concerns (even share some often), but from an objective point of view, these more extreme examples were provided but as we all know, it isn´t different in case of less extreme claims.
And any attempt to provide some experimental evidence is ridiculed, negated, belittled, and if all that does not help it simply will be ignored as if it never has happened.
Again i don´t see any sense in the demand of "blind tests" - in fact it seems now really to be a demand of "ABX Foorbar style" - if only negative results were accepted.
By not explaining to potentially testers what the pitfalls are and giving advice how to avoid these, it is more like a "cargo style" testing than interest in good experiments delivering (most of the times) correct results.
Last edited:
Is there not a noise floor on every analogue output? If this fluctuates for whatever reason, is it not a modulating noise floor? If this modulation is patterned in some way we are more likely to perceive it. If it is patterned in such a way that it interferes with the perception of the starting point of low level sounds, we are likely to perceive such sounds as less accurately defined in time & possibly in resolution also - the subjective description often given for this is that the sounds emerge from blackness.
Most modern DACs use noise shaping - if this is not 100% stable we have modulations in the noise floor. I would also refer you to the Mallinson video referenced a while back where he talks about state variables settling time. I also remember some talk here years ago from Peufeu about thermal memory in circuits. There are more ways for noise floor modulation than the few I have outlined.
If you look at the link I gave to a guy's ABX results you will see that the files they created are meant to be a simulation of jitter - they are created as follows (Post link)
This seems to be somewhat close to what we are talking about & there are a series of files with different levels of disturbance mixed into them. As you will see from reading the thread the guy reports positive ABX results for all files & the files are analysed for any other issues which might provide a 'tell' in ABX testing because, based on the accepted "thresholds of hearing" some of these files have no distortion in them that would be considered audible by 'accepted wisdom'
So maybe this is 'proof' that some forms of modulating noise being perceptible & we need to look into it?
Of course what we 'hear' arises from processing, but for the word 'hear' to have any useful meaning the processing must be largely restricted to handling input via the ears. Otherwise it is not hearing but something else, however much we may perceive it to be hearing.mmerrill99 said:
This allegation is repeatedly made by people who firmly believe that the test finds the 'wrong' result. Repetition is not a valid method of proof. It is to be expected that distinguishing between small differences will be difficult; I don't see why this is regarded as a criticism of the test. If it was easy then we could conclude that they were not small differences. I think the problem arises because small difference may be falsely thought to be larger differences when ears are supported by sight or other knowledge; it is well known that zero differences may be thought to be obvious differences when sight or other knowledge leads in this direction.
That is because I remain unconvinced. On the contrary, your attempts to convince me have had the opposite effect.
Yes, that I can agree with. Audiophiles believe that they can hear cables, that bad cables sound better than good cables, that poorly-engineered 'high end' audio has better sound reproduction than well-engineered mid-range, that coupling caps have a sound, that sighted tests are a valid way of testing hearing etc. etc.
I would not use the word 'believe'. It is simply a test protocol, it has strengths and weaknesses. For many people the main weakness is that it finds the 'wrong' result.
I would not call an alleged difference 'heard' in a sighted test a false positive; I would call it a mistake. The mistake is to confuse hearing with other sensory inputs. Sighted tests can cause people to 'hear' 'differences' which are simply not present (I don't mean too small to hear, but non-existent e.g. A and B are the same system). That is why sighted tests have no validity in testing audio systems, or testing human hearing.
I'm interested in the underlying mechanism for all of this & I believe it is that the signal reference voltage is being disturbed, usually some small disturbance on ground. I've seen lots of efforts which seems to work in this way & seem to result in the same perceived improvements - from hugely overspecced PS, battery PS, to over the top inter-chassis earth strapping. I keep an open mind about some of the more esoteric solutions but don't have a 'belief' in them just an inquisitive mind.
See the link I gave to the ABX results - this may well be a test of the audibility of noise modulation?
Although it is trivial it should be noted/remembered that in case of a chain with several different links each one must have quite higher bandwidth to meet the criterion for the whole system.
Some microphones do, others do not.
PMA, really, you demand proof for opinion?
I used "i think" and that should be sufficient to let readers know about the opinion nature of the statement. (*)
@billshurv,
"couldn´t tell a strad from a cheapy" means i think being unable to distinguish, but that wasn´t the outcome of these experiments done by Fritz et al.
Afair "telling them apart" was no problem but choosing one of the participating "strads" as favourites was a problem.
Btw, "cheap" was none of the participating instruments ...
Otoh, with all due respect, if you are concerned of unsupported claims, you should not post so often about differences that must be easy to detect in a "Foobar ABX" if a perceptable difference exists.
(*) Nevertheless there is some evidence given by the experimental results published by Oohashi et al. and the authors of several follow up studies.
Soloudre published some interesting results on the topic of intrachannel phase differences at higher frequencies in the audio band, that might of interest.
If that all is of much relevance for normal recording/listening is another question, but bringing in another low pass is always easy while bringing back something that was not recorded might be more difficult.....
Last edited:
I don't dispute that. It is a very powerful argument for avoiding sighted listening tests. It therefore puzzles me that you seem to regard it as a reason for embracing sighted tests.mmerrill99 said:
I did not specifically mention Foobar ABX. I was merely expressing surprise at Jakob's claim that a mountain of evidence would indicate test weakness rather than truth.
No, I am simply asking 'under these conditions can they distinguish A from B?'. That is all.Jakob2 said:
In most cases people do listening tests with equipment they regard as good, under conditions they regard as typical (or better) than domestic listening. Your criticism might be valid if people were doing blind tests with 2" computer speakers next to a building site and an airport radar, but I think such conditions are rare. The issue is a simple one: under something like normal conditions can people reliably distinguish between A and B, by hearing A and B? That is all. We know that allowing them to see A and B renders the test fatally flawed, because then they may reliably hear differences between A and A if they see B and hear A. 'Test stress' is only an issue for people whose reputation is on the line. If A and B are similar but different (which will often be the case) then it is to be expected that some people will hear the difference, some people will not, and some will hear the difference part of the time. The statistics will show this.
True. What is the distraction in this case?
OK, I take your point. I find statistics to be rather like quantum mechanics: lots of people can do the sums, but far fewer can actually explain what they mean - and there are different schools of thought who each claim to have the whole truth. I remember in first-year QM we were taught the Copenhagen School understanding, but we were not told that this was merely one school and there were others; I was a bit annoyed when I later discovered this.
I will sometimes talk nonsense on statistics, so I ask for continued patience. However, I sometimes feel that statistics are being used to obscure things rather than reveal them.
- Status
- Not open for further replies.