See the following.
After reading your posts quoted above, I want to ask you couple of things.
1. Which advocate of level matched audio DBT supported just doing something "blind (or double blind)"?
2. Which actual example of level matched audio DBT failed to reveal the audible difference that was really there?
You take it wrong.
I am somewhat skeptical of the description, and the claimed procedure, if it actually happened, won't even get you close to that figure.
0.1 dB ~= 1% as a V ratio which would require -40dB noise below your test/listening SPL in a room
so 70-80 dB SPL listening level requires below 30-40 dB background noise - not always met in home listening rooms but if you have that much noise why can't the "objectivisits" laugh at your "poorly resolving" listening environment
I don't see the logic in arguing room noise making matching difficult AND claiming you can rely on "just listen" for hearing meaningful differences under the same conditions
so 70-80 dB SPL listening level requires below 30-40 dB background noise - not always met in home listening rooms but if you have that much noise why can't the "objectivisits" laugh at your "poorly resolving" listening environment
I don't see the logic in arguing room noise making matching difficult AND claiming you can rely on "just listen" for hearing meaningful differences under the same conditions
Last edited:
Which brings me back to the whole notion of carrying out (and insisting upon) a home, level-matched A/B listening test, and its worth.
Such tests are valuable, if conducted in a controlled lab environment minimizing as many random variables as possible, using as many participants as possible, repeated as many times as possible, on various age groups , then shoved through a statistical number crunching software, to get some meaningful result.
Joe public has no chance to do such a test, a simple A/B comparative listening test (if) done by a handful outside of a lab becomes an almost a meaningless exercise.
There are many who have argued that one needs a much longer session of probably weeks with each unit to decide on its worth.
I belong to this group.
Last edited:
Nonsense. If you want to determine, "Can I hear the difference between A and B?" it is quite simple to do. If you're content with never testing or challenging your own perceptions, and have some sort of emotional stake in the outcome, that's fine; not everyone has intellectual curiosity.
it is not how much background noise you have to begin with, but how much it changes, at what frequency bands when you go from A to B, and what control you can have over it.
in otherwords, what you thought you heard (or did hear) was it because of changing background noise of a few dB's in 500-1000Hz range, or was it because A sounded sweeter than B?
If enough ears say it does, then it may exist. The problem is that in many cases it is not unaided ears which 'hear' but ears aided by sight.KenTajalli said:
If they are blind tasters then it is likely that it is different. If they are sighted tasters then their perceptions carry much less weight.
Some of them are. Some of them may even be fooling themselves.
Lots of audio stuff is sold to gullible purchasers with stories which are obvious nonsense. That must raise the suspicion that some other audio stuff is sold with stories which are less obvious nonsense, but nonsense nonetheless. Having bought into the story (often, sadly, literally with hard cash) the gullible then demand proof from critics that the story is false; in many cases this is impossible to provide because the buyer can't actually understand the issues involved so the valid argument goes over his head. He then declares that in the absence of proof to the contrary the story must be true. This is a logical fallacy (you can't prove something true simply by failing to prove that the converse is false) but logic is rare after money has been spent.
Find a way of demonstrating that it exists. If you fail, try again. Eventually, give up and accept that it doesn't exist. As has been said before (by SY?) 'data' is not the plural of 'anecdote'.
You may be confusing "Do I want to buy this?" with "Does this reproduce sound accurately?".KenTajalli said:
What if I don't?
I don't care if product A is different to product B, I'm only interested that when I invest my money and time on a DAC or any hifi gear, it gives me more repeated satisfaction for years to come, and I can not determine that by a simple A/B comparison, unless (again) the differences are indeed night'n'day.
I do not employ lab environment testing procedures, when I am deciding which DAC is better (to me, to my ears with all my flaws).
If enough people choose the same DAC , that DAC is better.
If the changes to sound quality become harder and harder for me to detect, then I have to decide, if I want to pay $3000 extra to get them (diminishing returns).
In my case the ~$2000 mark is about as much you can get from a good quality DAC that's worth it to me.
For you it may be $100.
for one it is 'reproduce music more faithfully' I don't care what it can do with steady state test tones.
it seems my definition of a 'better DAC' is different to some (yours).
Hifi DACs are made to reproduce Music at home, indeed all hifi gear are, they are not meant to be anything else.
If buyer/listener can not achieve his goal: 'reproduce Music more faithfully' then that equipment has failed.
The Asus sound card that I do have (lab results below) although excellent, does not have the same sound quality of my other DAC the Teac UD-501 , tried in three different systems. both measure very well (and both sub $1k).
Asus Results
Teac Results
Am I really?
this is a thread about commercially available hifi DACs made for domestic use, or did I get that wrong too?
I always thought we were discussing things in that context.
it does look that way - mag tape, vinyl from way more than 2 decades ago does seem to be many people's gold standard
Last edited:
It was just an abbreviation, as a lot of people seem to think that adding the "blind property" to a test would in a miraculous way ensure getting correct results.
Controlled listening test means more than adding "blindness"; "blind" only means that _one_ bias effect is avoided, all the other bias mechanism (distractors) are still at work (rosenthal effect, hawthourne effect, internal criterion problem, presentation order, time errors, expectation bias and so on)
A listening test is just a sensory test and in general it is a hypothesis test and is therefore prone to both error possibilities, which means alpha error (i.e. to reject the null hypothesis although it is true) and beta errors (i.e. to _not- reject the null hypothesis although it is false).
Accurate level matching is most important in quite short comparisons with rapid switching between DUTs, as a level difference will be a systematic error.
If the results of such tests are of practical relevance is indeed questionable and if one is aiming for something more comparable to everyday use, then rapid switching between DUTs is of minor relevance, as one is looking for a difference he will be able to appreciate afterwards.
In this case level matching is of minor importance (provided that any level differences will be randomly assigned to each DUT) as it then will work as a random error.
Listening time will be extended to several days and the DUT will be used during this time at different level settings.
If the results of such tests are of practical relevance is indeed questionable and if one is aiming for something more comparable to everyday use, then rapid switching between DUTs is of minor relevance, as one is looking for a difference he will be able to appreciate afterwards.
In this case level matching is of minor importance (provided that any level differences will be randomly assigned to each DUT) as it then will work as a random error.
Listening time will be extended to several days and the DUT will be used during this time at different level settings.
ok, I will get a 0.1db mic and redo another test once I receive my mosaic dac (may take a while).
what I can say is that when I compared the sabre hifimediy dac to the hrt ms2, those two cheap dacs sounded very close to each other. but then going from those to a buffalo 2 and then the sd-1 player, the difference were substantial and evident in many different ways.
what I can say is that when I compared the sabre hifimediy dac to the hrt ms2, those two cheap dacs sounded very close to each other. but then going from those to a buffalo 2 and then the sd-1 player, the difference were substantial and evident in many different ways.
SY,
could you cite some papers with real test results where level differences of 0.2dB were detected with music?
Literature normally give JNDs of 0.5 - 1 dB, the lowest number i remember came from Clarke´s JAES article in which he mentioned >= 0.3 dB with music but without any detailed description of the test (of course it was an ABX).
could you cite some papers with real test results where level differences of 0.2dB were detected with music?
Literature normally give JNDs of 0.5 - 1 dB, the lowest number i remember came from Clarke´s JAES article in which he mentioned >= 0.3 dB with music but without any detailed description of the test (of course it was an ABX).
Why would you need to do it with a mic?
Measuring the voltage on the outputs of each DUT is sufficient as long as the level is within the linear region of each subsequent processing unit.
Although your normal mic might not be really precise overall, the reproducibility at a specific frequency is usually nevertheless around 0.1dB or better.
Last edited: