In the process of reading through some IRE journals in an attempt to gain some understanding regarding AB testing I have come across the mention of this phenomena several times. Does anyone know if it is genuine and if so what impact this effect may have on testing modalities?
I know nothing of this law but a quick goggle search came up with the following. There were many other sites found. You might want to try a search on goggle. It's spelled "Weber-Fechner Law"
Thanks Mr V thats an excellent place to start. This is one of the quotes from the audio field circa 1954; " The increase of a stimulus neccessary to produce a just discernible increase in the resulting sensation bears a constant ratio to the total stimulus. It is sometimes stated in the form that the magnitude of the sensation produced is proportional to the logarithum of the stimulus. This law applied to the hearing sensation means a fractional increase in intenstity, which is just perceptable as a change in intensity; should be a constant independent of the intensity." I have not seen this made public yet by the AB posts; yet it seems pretty critical to any test involving the ABX format. What gives?
Wayne Parham Messages: 18758 Registered: January 2001
Illuminati (33rd Degree)
This is why the volume levels are to be held constant in any A/B test. It's also why I don't consider power increases less than 2x to be particularly significant. If you A/B test two things with 10% difference, you'll surely hear it and the louder one has a slight advantage. But if you're not A/B testing, you really have to double power to notice much of a change. A ten percent increase seems like almost nothing at all, barely perceptible unless compared side-by-side.
I see your point regarding level matching as it seems that would also affect the amps and distortion and overload conditions and every other aspect of reproduction. Is that the sum total of what they mean here though? The couple of papers I have encountered this mentioned it was used to describe quality issues as the variable while the operating conditions were held stable; that is the part I found curious. Looking at Mr Vinyls link they say for instance that humans hear Pitch Distance in a logarithmic fashion, ie the distance between 100hz and 150hz sounds the same as the difference between 1000hz and 1500hz. One article mentions transient overload in the output stage of a PP amp and conditions that skewed the AB results. Just something to mull over. Looks interesting though.
Wayne Parham Messages: 18758 Registered: January 2001
Illuminati (33rd Degree)
In the musical scales most of us are familar with, there are twelve half-notes per octave. Each half-note is separated by a factor of 21/12, approximately a 1.06-to-1 ratio. The math behind musical scales is very interesting. Check out these links:
Absolutely; that is one of the points they make on the site, they include the formula for determining all of the relationships between tones. What I don't follow is how that phenomena affects reaction to percieved quality of sound. That seems to be the sticking point as this Weber-Fechner Law modulates all results. I only bring it up as it appeared in a couple articles on testing and I had never heard reference made to the effects it might have in any of the current literature. Its new to me also.
Wayne Parham Messages: 18758 Registered: January 2001
Illuminati (33rd Degree)
It's definitely interesting stuff. I took piano lessons for seven years when I was a kid, and I played synthesizers when they were analog. So I studied musical scales and note and chord progressions and also waveforms, harmonics, attack and decay and that kind of thing. It's holistic and reductionistic at the same time; There is art and mathematics intertwined in the most interesting way.
Wayne my friend; I must apologise for my inability to properly communicate on this issue. I really appreciate that you are willing to pursue this and regret that we seem to have become sidetracked here. While it is a good discussion and enjoyable we have not addressed the issue of the post at all. See this statement of the Weber-Fechner Law is not addressed specifically to sound or audio. It is a consequence of pshycological response to external stimuli as far as I can tell. One way I found it too be expressed is that in comparison testing the two options at test and the control are governed by the perception of difference in as much as people will always choose the option that is most close in quality to the last choice. If there are any differences, and that is what the test seeks to prove, then the proximity of the choice will determine the outcome. This was considered the fundamental flaw in all comparison testing and as a result caused many of those types of tests to be suspect or abandoned. Would you agree with this understanding; or am I not getting the meaning of this theory? Please understand this is my interpretation of the process; it is not an opinion I found somewhere. Thanks for offering your valuable insights into this.
Nice Link, Mr. Vinyl, and thanks for bringing this topic up John. What it tells me is that improvements in audio reproduction beyond a certain point suffer from dimininshing marginal returns, so an AB testing of high quality equipment will usually lead to a significant null result. -akhilesh