Ben Bradley wrote:
> On Fri, 22 Aug 2008 17:11:24 -0500, Richard Owlett
> <
[email protected]> wrote:
>
>
>>SteveSmith wrote:
>>
>>>Here's what I found when doing the research for my book-- 15 years ago.
>>>Sorry,
>>>I don't have the reference anymore. This also gives you another topic to
>>>look under, Companding.
>>>Steve
>>>
>>>http://www.dspguide.com/ch22/5.htm
>>
>>Hmmmm, Are you and Mr. Avins in partnership to whack dumkoffs(sp?) up
>>beside head until they actually begin to *THINK* ROFL
>>
>>The first paragraph of your reference was an appropriate weight 2x4 ;)
>>It demonstrated that solving my immediate problem had little to do with
>>my specific question.
>>
>>HOWEVER, I'm still fascinated by the answer to my question:
>> 1. how loud must signal B be to mask signal A?
>> 2. what is minimum amount of distortion that is detectable.
>
>
> These are indeed psychoacoustic questions, but I've seen them
> discussed in relation to amplifier distortion, and especially the
> "tubes vs. transistors" debate that's been going on ever since
> transistor amplifiers were commercially available.
> I recall that the 4th edition Radiotron Designer's Handbook (the
> venerable 1000+ page vacuum-tube electronics design reference from the
> 1950's may have someething on this.
Now that may be motivation to dig thru a stack of boxes I haven't really
looked at for 20 years.
> One more resource might be the
> Journal of the Audio Engineering Society (http://aes.org), or some of
> their other publications.
> There's also Google's Usenet archives, where I've seen this sort of
> thing discussed on rec.audio.pro and perhaps also
> sci.electronics.design as well as comp.dsp.
> Here's a general answer that you might already know (I now look at
> what I wrote below, and some may say I'm wrong about something, but
> whatever, it may spur some good discussion).
> The first thing to keep in mind is that "percent harmonic
> distortion" figure used for amplifiers is a near-meaningless number
> for audibility of distortion. It's a conglomerate figure for all
> distortion, and some types of distortion are more audible than others.
> Lower harmonics (say, second and third) are more easily masked by the
> fundamental and have to at a higher volume to be audible than do
> higher (just to pick numbers, maybe fifth and up) harmonics. Also,
> when they ARE audible, lower harmonics are more "pleasing" and
> "euphonic" than higher harmonics, and might be mistaken for part of
> the original signal rather than percieved as distortion (so it could
> be important to compare with an "undistorted" original signal, with
> something like the PCABX I mention below).
> The transfer characteristic of vacuum tube amplifiers is generally
> in the shape of an 'S' curve, as tubes gradually go into cutoff and
> saturation. Even in their "linear region" it's a slight curve, so it
> can generate distortion well below the extremes. This generates mostly
> lower-order harmonics. Transistor circuits tend to have transfer
> curves that look like a stretched out "Z" due to their hard turnon and
> turnoff characteristics, generating higher harmonics.
> Also, tubes are operated at "full gain" with little or no negative
> feedback (perhaps one reason is they are, or certainly were,
> relatively expensive, and it takes more tubes to make up for the
> lowered gain when using high negative feedbback), whereas transistor
> circuits are designed for high gain and high negative feedback.
> While negative feedback DOES lower distortion, it also changes the
> characteristics of the distortion that remains. One interesting point
> is crossover distortion in push-pull transistor amplifiers (circuits
> are usually biased on enough to make this insignificant, but let's
> pretend not for the moment). This is generally regarded as a very
> audible and very bad sounding distortion (it's also "counterintuitive"
> in that it DECREASES a percentage of the signal as the signal level
> increases, the opposite of most types of distortion). Negative
> feedback will reduce the percentage distortion in this case, but will
> force the remaining distortion products into higher harmonics, perhaps
> even making the problem MORE audible than with less negative feedback.
> rec.audio.pro poster Arny Kruger used to have a "PCABX" site with
> various .wav files with different amounts and types of distortion you
> could download, along with software that would pick randomly between
> two .wav files - you could play A, B, or the one the program picked,
> X, and the idea was to determine by ear whether X was A or B. You
> would push a button to indicate your decision of X=A or X=B, and the
> program would then tell you if you're right. By doing several trials,
> the progra determines statistically whether there's an audible
> difference (to that listener!) between the two sound files.
> The site pcabx.com is no longer available for reasons I don't know
> (it went away while I wasn't keeping up with Usenet), but I recall he
> did have a large number of .wav files, and the hosting costs may have
> been significant.
> But I suppose you just want to find results and not do your own
> original research (especially when it surely has already been done in
> this area). What's your application for this? How much time and money
> are you willing to spend to answer these questions?
>
This outgrowth of an offshoot of a general interest in problems related
to speech recognition. The only budget I have for this is my time and
access to the Web. My current project is representing
time/frequency/intensity of sound in 3D - the spectrograms that are
typically used just don't "work" for me.
The purpose of this round of questions was to get some idea of how to
scale the plot to be both "pleasing" and useful. My current idea is to
experiment with plotting the data on a linear scale with contours
displayed at logarithmic intervals.