There's nothing wrong with the premise of Bill Waslo's test tracks. Proving someone can't hear the marching band in his sample tracks proves just that: it's hard to hear the marching band masked by a louder choir. But those tests are not representative of the phenomena that can occur with cable effects.
There seems to be a misunderstanding in this thread about the interdependence of the source / cable / load in manifesting audible effects. Perhaps that is why the subject of cable sound is so contentious. On a forum almost no two people are comparing the same setup so experiences are inherently going to differ.
With the blind comparator test setup I've used for the last couple of decades, a demonstration that excludes random variables easily demonstrates differences between two speaker cables A and B to anyone in minutes.
The setup is such that although the listener can see both cables neither I nor the listener can tell which one is being used. the process starts by randomising whether cable A is X or Y, and B is therefore Y or X. The listener can switch between the two as often as they wish and whenever they want, until they can hear a difference, or they can't. Then it can be revealed whether A is X or Y, and b was Y or X. The next round randomises again so the known A/B becomes an unknown X/Y again.
After say six different rounds of picking whether X = A or X = B correctly, the chance of guessing is 1/128. But once the listener's auditory system has tuned in to the difference, picking which is which often just takes a few seconds. I've done this hundreds of times with hundreds of people. There is no need for someone to have golden ears for them to be able to identify that two different speaker cables, X and Y, sound different with the blind comparator.
Deciding which cable is better is harder, because you need to know what the test recording should sound like before you can determine which cable is altering the sound less. but there are ways to work that out too. The best cable is no cable, a shorter cable reduces the amount of cable effect, longer increases it. This is how to determine what the cable is doing.
Last edited:
As you are aware a cable is an LCR network, ie filter, so although you can say it doesn't technically distort it will change the frequency response, this may or may not be audible. Although it's not exactly the same thing, level matching is very important in DBLT, so it seems reasonable to assume the frequency response change could be audible.
But of course they are representative! If you can't hear a full band at 60dB below the main music, how on earth can you hear, say, -80dB distortion of a cable below the main music? This is basic reasoning.
Jan
It might seem basic reasoning, but it is faulty reasoning. The masking effect of unwanted audio signals is very much dependent on the signal structure of both the wanted and unwanted signals.
For example bit rate reduction for compressed audio files is possible because of perceptual masking which determines which low level signals are masked and thus can be 'thrown out' and which low level signals must be retained in order to preserve fidelity.
Piano notes are a great revealer of N+D because the transient energy of the hammer strike excites resonances, whereas the sustained note offers very little masking of N+D, whether linear or nonlinear, correlated or uncorrelated.
Tape noise at -60dB is quite audible because the noise is not correlated to the wanted signal and thus not masked.
Jitter is audible in digital playback at even lower levels because aliasing creates distortion products that are unrelated in any way to the harmonic content of the wanted signal.
It took decades since the introduction of CDs to sort that problem, at least partly because CD playback was technically 'perfect' and didn't need fixing according to the enlightened non-audiofools. Anyone who didn't like the edgy sound of early CD players and discs was a nutter in the mid-1980s; how times have changed with regard to digital!
The Audio DiffMaker AES paper makes interesting reading. John Dunlavy and I used dual channel FFT to subtract the difference signals of the 'in' and 'out' ends of loudspeaker cables (and other audio components) in the mid 1980s, about 30 years before Bill Waslo's paper. Using dual channel FFT in real time with simultaneous 'before' and 'after' in the analogue domain means almost all of the post processing required for DiffMaker to work was not necessary, i.e. no need to compensate for clock variations, sampling errors, different environmental noise or EMI/RFI at the two different recording times etc, etc.
Last edited:
"John Dunlavy and I used dual channel FFT to subtract the difference signals of the 'in' and 'out' ends of loudspeaker cables (and other audio components) in the mid 1980s, about 30 years before Bill Waslo's paper. Using dual channel FFT in real time with simultaneous 'before' and 'after' in the analogue domain means almost all of the post processing required for DiffMaker to work was not necessary,"
That sounds quite interesting, has that been documented somewhere?
Jan
That sounds quite interesting, has that been documented somewhere?
Jan
Your test plan is not good enough. Set aside my math tells the chance to get 6 out of 6 is 1/64 (which is a rather high probability, make it 10 or more and we can talk) you need to specify the pass and fail limits. Check out the binomial distribution and the formula that gives the probability to guess exactly k times out of n trials when the probability to guess in each trial is p. It will be clear why for p=0.5 n=6 is way too low.
No, the stuff we did together at Duntech wasn't documented. Maybe John Dunlavy had his own files. I was at Duntech as an invited guest to use his very well funded research facilities because at the time I was the Sound Master at the Adelaide Festival Centre. At the Centre we had our own research and development program for sound equipment including speakers; we were kind of like a mini BBC technical department of the era.
We didn't exist to publish research, just make better stuff than we could buy.
I did do about ½ dozen or so AES presentations, including double blind speaker evaluations. I was an AES committee member for a decade or two, and was often asked to write and present papers for AES conventions, but I was too busy just doing it and content with reading AES research papers and doing stuff properly.
It's pretty clear that either I have not explained the method well enough, or you have not understood my system. With my system there is no incentive, motivation or reason for a listener to guess one way or the other. There is no reward at all to be right or wrong, because there is no right or wrong. If someone correctly makes six out of six identifications, they will more than likely make 20 out of 20 or even 100 out of 100. If they cannot hear a difference, they do not make any choice because there is no incentive or reason to state a choice that might embarrass themselves.
I developed this A/B to random X/Y comparator system when I was employed as an electronic engineer and designer by a loudspeaker manufacturing company, and was frustrated by the owner's golden ears approach to design. I had previously designed and presented a formal double blind professional loudspeaker evaluation for the AES in accordance with AES recommended practice, so I do have at least a tiny grasp of some of the pitfalls of subjective sound quality assessment.
The reasons for developing my simple blind comparator was that I wanted to be able to demonstrate what was audible and what was bunkum to others, including the messiah owner of the "high end" loudspeaker manufacturing business who employed me. Eventually I was sacked for disagreeing with the company messiah's ethos, so I took the blind comparator apparatus with me knowing the they would never use it. It has since been used hundreds of times by me for development of product, and for other people to be objective about how they spent their hifi dollars.
What levels? Lots of words from you with no numbers. Published research shows that Jitter sensitivity is not the boogyman many say it is. Of course didn't stop 'jitter reduction' boxes getting rave reviews even if they made matters worse.
The perfect part was philips marketing dept. Was Jitter even a problem before Miller and Dunn worked out how to test it around 1992 and suddenly it became a new FUD that manufactures could harp on about?
And what has this got to do with cable distortion other than a bit of slight of hand to deflect?
Your words, not mine. At one extreme it is possible to add an uncorrelated disturbance signal to another signal at the same power level of the wanted signal and for the effect to be inaudible. At the other extreme, an uncorrelated signal added to the wanted signal is obvious at diminishingly minuscule amounts. What is and isn't audible is all about the the nature of the two signals.
Last edited:
I'll leave that for you to work out. If you care, you might start by exploring the context in which the posts were made.
Last edited:
I did now, after reading this post. You have no clue about designing and executing a relevant test.
Jan, you know better. How many times do you have to be reminded that some people can hear if a CD has not been dithered? The quantizing distortion is calculated to be at -93dBFS.
We could talk about why it might be that brass bands and quantizing noise apparently have different thresholds of audibility, if anyone wants to. In the meantime can we please remember the facts that tend to confirm, and the facts that tend to disconfirm, one's preferred beliefs should not be misused in an attempt just to win an argument. Not if we want to be scientific, anyway.
Yes, the nature, correlated or not. Or what sort of 'nature of signals' did you mean? Or was this just a remark with no specific meaning?
Jan
Is it to do with patterns, ie, they are both musical, similar to the cocktail party effect, and tape noise would be like a steam train went by outside, you'd be aware of the sound because it wasn't correlated in the same way?
Yes whenever jitter is wrongly raised to try and prove black is white then soon after dither is raised as well. Again nothing to do with cables.
- Status
- Not open for further replies.