John,
An example from recent experience where measurement is trumped by the ear in a big way:
I fumble around with speakers quite a bit, and presently I am developing something with a small woofer that goes low by virtue of a large excursion, and it does so very well. With one exception, which is that it has long leads between the driver terminal and the attachment to the voice coil, which run through the cone. These flop around at larger movements, and these movements are acoustically coupled to the cone.
This is clearly audible. However, does it show up in an FFT as distortion? Not really. The leads flop at the same frequency as the fundamental, so whatever comes out of the measurement mike falls either in the bin of the fundamental, or will show up as an increase in the noise floor. The latter one is very low definition with acoustic measurements. So, audible artifacts, but no way for me to measure it with my halfway decent setup.
The same white spot might show up in measurements of gear such as amplifiers, where the audible artifacts are less pronounced, but could still be discernable to the true GEB nonetheless.
Last edited:
wrt perception, absolutely I dont have the answers and i'm not sure we, individually, or as a group, ever will. A single person can have 2 different experiences of the same electronic/electro-mechanical signal repeated twice, but that does not mean it was actually different. a change cannot be made in the equipment to correct for something that did not originate in the equipment. that is a completely separate issue and its where these 2 are confused that causes the majority of argument.
transducers, the room and the perception (as well as individual preference for something other than transparency) are IMO the main areas of concern. the recording and playback of audio signals in the box, to well beyond our ability to discern the difference; not so much.
transducers, the room and the perception (as well as individual preference for something other than transparency) are IMO the main areas of concern. the recording and playback of audio signals in the box, to well beyond our ability to discern the difference; not so much.
Last edited:
qusp,
I am in agreement with you on where we need to improve before we get any more carried away with the electronics side of things. As long as a speaker can have 10^2 or even 10^3 power greater distortion why are we always chasing such small improvements in the electronics side while the acoustic transducers can be so ridiculously bad? IM and harmonic distortion in the transducers is the major flaw in our systems, not the fine detail distortion we often chase.
I haven't got the test gear to look at the distortion mechanism in the electronics side that makes one amplifier sound better than another but I do suspect it has much to do with the 2nd harmonic distortion numbers that make one amp sound stale or analytic and one that has a smoother almost tube like sound that though not technically correct is something many would chose as the nicer sound.
I am in agreement with you on where we need to improve before we get any more carried away with the electronics side of things. As long as a speaker can have 10^2 or even 10^3 power greater distortion why are we always chasing such small improvements in the electronics side while the acoustic transducers can be so ridiculously bad? IM and harmonic distortion in the transducers is the major flaw in our systems, not the fine detail distortion we often chase.
I haven't got the test gear to look at the distortion mechanism in the electronics side that makes one amplifier sound better than another but I do suspect it has much to do with the 2nd harmonic distortion numbers that make one amp sound stale or analytic and one that has a smoother almost tube like sound that though not technically correct is something many would chose as the nicer sound.
There is a whole class of new measurement solutions for that and similar problems. Its called rub and buzz testers. Some work better better than others but they do work well in manufacturing environments.
This is typical if a problem can be identified clearly. It becomes a clear task and a test can be devised that works. For "uninvolving sound" the description doesn't lead to something that can be tested for.
There is an issue of how much distortion and what kind matters. If you don't have clear guidelines or rules you overbuild. The safe way to go is to reduce any identifiable distortion. Since its possible to test for distortion products at -130 dB you must come to terms with how far you can go and at what point you have really eliminated distortion as an issue. I defy anyone to hear something even at -100dB.
The second/third harmonic levels may be an odd issue. It could be inferred from stuff earlier in this thread that significant second harmonic can enhance the ability to determine pitch which leads to a possible "better than real" experience. If so it could explain the curious tenacity single ended triode amps have in the market for example.
Demian
I was going to mention that a waterfall display is better for showing the rub kind of problems.
Buzz is much easier to detect.
The humor is that for my current stadium project my group will be measuring and adjusting issues that although mentioned elsewhere in this thread are almost universally ignored.
Test equipment when properly designed and used can read artifacts below the hearing threshold. The issue is what parameters do you want to view.
The problem I have with any testing including SY's blind tests is that you really can't be sure you are controlling all the parameters.
The weirdest example I know of this is the effect of RF fields on HF atmospheric attenuation. More common uncontrolled variables include humidity and atmospheric pressure. Then we could talk about reproduced bandwidth vs live.
So there always is room for the GEB to say something doesn't sound right. Of course I have been amused see some rate the highest distortion components as the best, when others rate the lowest the best. So there clearly is personal preference.
ES
I was going to mention that a waterfall display is better for showing the rub kind of problems.
Buzz is much easier to detect.
The humor is that for my current stadium project my group will be measuring and adjusting issues that although mentioned elsewhere in this thread are almost universally ignored.
Test equipment when properly designed and used can read artifacts below the hearing threshold. The issue is what parameters do you want to view.
The problem I have with any testing including SY's blind tests is that you really can't be sure you are controlling all the parameters.
The weirdest example I know of this is the effect of RF fields on HF atmospheric attenuation. More common uncontrolled variables include humidity and atmospheric pressure. Then we could talk about reproduced bandwidth vs live.
So there always is room for the GEB to say something doesn't sound right. Of course I have been amused see some rate the highest distortion components as the best, when others rate the lowest the best. So there clearly is personal preference.
ES
I think this is the key point that means that this discussion is destined to go on for all eternity . . .
Some of us believe the human auditory perception can hear elements in audio reproduction that it is very hard to measure. This is my view
Other's appear to believe that modern day measurement methods have long "exceeded our ability to ear".
Ultimately listening to Music / Audio has to be an individual subjective experience so this stuff may never be able to be "prooved" either way.
I think to save energy it's best we just accept that we can never know what others can or cannot "hear". Let's not give each other a hard time about the fact that there are different viewpoints here.
As someone said earlier, it doesn't matter what other people hear ( unless we're feeling emotionally insecure about it ) The only significant thing for each of us is what we can hear.
It would seem there are pros & cons to both viewpoints which we're just gonna have to live with coz I don't see this dichotomy going away any time soon.
whatever you say... if you want to let people think they can hear clear differences in bitperfect data, and things that are not happening, can be clearly heard by someone skilled in the art of listening; go right ahead. i'll keep my sanity thanks.
being 'politically correct' and taking the ground that anything can happen in someone elses world is just pissweak IMO. there is perception and there is objective and verifiable reality; one is verifiable by others and by recording, the other only takes place in a very flimsy/temporary way inside an individual mind. most people cannot even communicate that experience effectively to others.
just because we cannot prove that something subjective didnt happen in someones mind, does not mean that it actually did happen in reality
perception and listening is incredibly important as part of development, but (by iteself) it is proof of nothing
being 'politically correct' and taking the ground that anything can happen in someone elses world is just pissweak IMO. there is perception and there is objective and verifiable reality; one is verifiable by others and by recording, the other only takes place in a very flimsy/temporary way inside an individual mind. most people cannot even communicate that experience effectively to others.
just because we cannot prove that something subjective didnt happen in someones mind, does not mean that it actually did happen in reality
perception and listening is incredibly important as part of development, but (by iteself) it is proof of nothing
Last edited:
Youre right though, the conversation is both perennial and impossible to resolve. Why people cannot just be happy they have had a wonderful subjective experience and instead attempt to ascribe it to external factors that make no sense, is beyond me.
mikelm,
At some point we have to have some reference points in listening and testing to use for comparison reasons. Even though there are circuits that may have a S/N ratio of up to let's say 120db that does not mean that we can actually hear with that acuity. Think about how far apart the loudest to quietest material must be to have that wide a ratio. That is like saying that someone with the best hearing can hear a fly on the wall while listening to rock concert level program material, it doesn't make practical sense that humans have that wide a range of discrimination. And here we are talking about S/N levels that are worlds greater than most dynamic loudspeakers could reasonably accomplish. The average speaker specification will have at least a 1% distortion rate at a given SPL and many are listed at 10% at rated output. How can we say we can distinguish a 120db S/N at 0.001% distortion when the background loudspeaker distortion is multiple factors greater than the electronics?
Demian and Simon,
I would say that in loudspeaker design and development the waterfall plots have been one of the most informative changes from a simple FR curve. Without this we can never see a resonant problem or see the decay properties of a material or the assembly in a loudspeaker. A perfectly nice FR may be the result of a simple static response curve but the waterfall is going to show the time domain response that is probably more important than a perfectly flat response. I understand the rub and buzz testing protocol but I am not sure how the lead-out wire slapping the cone would look in the waterfall response? I would think that would be very level dependent and may not show up at all in a common 1 watt @ 1 meter test protocol used in standard QA testing. In a long excursion driver you may have to approach max excursion before this phenomena would happen. I see that JBL and others have patents on the lead wire becoming integral with the spider but I don't know if this introduces any new problems?
At some point we have to have some reference points in listening and testing to use for comparison reasons. Even though there are circuits that may have a S/N ratio of up to let's say 120db that does not mean that we can actually hear with that acuity. Think about how far apart the loudest to quietest material must be to have that wide a ratio. That is like saying that someone with the best hearing can hear a fly on the wall while listening to rock concert level program material, it doesn't make practical sense that humans have that wide a range of discrimination. And here we are talking about S/N levels that are worlds greater than most dynamic loudspeakers could reasonably accomplish. The average speaker specification will have at least a 1% distortion rate at a given SPL and many are listed at 10% at rated output. How can we say we can distinguish a 120db S/N at 0.001% distortion when the background loudspeaker distortion is multiple factors greater than the electronics?
Demian and Simon,
I would say that in loudspeaker design and development the waterfall plots have been one of the most informative changes from a simple FR curve. Without this we can never see a resonant problem or see the decay properties of a material or the assembly in a loudspeaker. A perfectly nice FR may be the result of a simple static response curve but the waterfall is going to show the time domain response that is probably more important than a perfectly flat response. I understand the rub and buzz testing protocol but I am not sure how the lead-out wire slapping the cone would look in the waterfall response? I would think that would be very level dependent and may not show up at all in a common 1 watt @ 1 meter test protocol used in standard QA testing. In a long excursion driver you may have to approach max excursion before this phenomena would happen. I see that JBL and others have patents on the lead wire becoming integral with the spider but I don't know if this introduces any new problems?
Last edited:
Actually on this topic I have a question for John
I'm wondering, when you have designed & built a pre-amp that many who have heard it regard as the very best ever produced, why are you dissatisfied with the methodologies that you used to develop it.
What is wrong with doing final voicing of audio equipment subjectively ?
You seem to be managing very well with this method thus far.
Are you expecting significantly better results if you can develop and objective testing regime ?
I'm wondering, when you have designed & built a pre-amp that many who have heard it regard as the very best ever produced, why are you dissatisfied with the methodologies that you used to develop it.
What is wrong with doing final voicing of audio equipment subjectively ?
You seem to be managing very well with this method thus far.
Are you expecting significantly better results if you can develop and objective testing regime ?
Last edited:
I have constantly evolved my test techniques over the decades.
Once, long ago, I thought that .005% SMPTE IM was about all that was necessary, especially if the harmonic structure was low.
Then, I graduated to .001% SMPTE IM, and 10-50000 Hz harmonic distortion with both wave analyzers and THD meters.
Then, TIM, with a sine-square waveform that was amazing of separating lousy from good circuits.
Then, I tested my circuits with a Hirata distortion system.
Ultimately, I used modified THD meters with FFT based secondary measurement, with signal averaging and 100K bandwidth, to reach below -120dB or .00001% distortion, especially for higher order 5,7,9th.
Here is where I have had to stop as I cannot afford even more elaborate equipment.
If I could, I would get a Stanford Research SR-1 and I probably will, sooner or later.
However, I have found that NO MATTER WHAT I MEASURE, the true test is with the ears, usually with STAX electrostatic headphones.
For us, 'voicing' is just a general term for picking the best passive components that we can discriminate differences by listening. This includes: wire, connectors, caps, solder, circuit board material, power supply improvements, etc.
For me it is NOT changing the actual sound measurements, as is what is usually possible with typical audio test equipment, and we NEVER deliberately add distortion.
That's about it. Of course, I ALWAYS MEASURE to the best of my ability, but I have been fooled before, and if I don't get independent evaluation by others of the sound quality of the product, I have been very embarrassed on occasion.
Once, long ago, I thought that .005% SMPTE IM was about all that was necessary, especially if the harmonic structure was low.
Then, I graduated to .001% SMPTE IM, and 10-50000 Hz harmonic distortion with both wave analyzers and THD meters.
Then, TIM, with a sine-square waveform that was amazing of separating lousy from good circuits.
Then, I tested my circuits with a Hirata distortion system.
Ultimately, I used modified THD meters with FFT based secondary measurement, with signal averaging and 100K bandwidth, to reach below -120dB or .00001% distortion, especially for higher order 5,7,9th.
Here is where I have had to stop as I cannot afford even more elaborate equipment.
If I could, I would get a Stanford Research SR-1 and I probably will, sooner or later.
However, I have found that NO MATTER WHAT I MEASURE, the true test is with the ears, usually with STAX electrostatic headphones.
For us, 'voicing' is just a general term for picking the best passive components that we can discriminate differences by listening. This includes: wire, connectors, caps, solder, circuit board material, power supply improvements, etc.
For me it is NOT changing the actual sound measurements, as is what is usually possible with typical audio test equipment, and we NEVER deliberately add distortion.
That's about it. Of course, I ALWAYS MEASURE to the best of my ability, but I have been fooled before, and if I don't get independent evaluation by others of the sound quality of the product, I have been very embarrassed on occasion.
I can fully agree to this for anyone who builds his system (either by writing a check or by soldering it together) and enjoys the fruits of his labors. When I design a power amp and think it sounds great, I have nothing to prove to anyone else.
Except. If I go on line telling everybody that brand x resistor will bring out a whole new universe in music from your amp, that's a different case. If I tell people that unless they use the genetic engineered speaker cables that provide quantum resolution (and by the way, I will sell you such a cable), the fact that I hear it (real or imagined) is not sufficient.
There are many here that cannot keep the two situations apart.
jan
Ah, but I do have to worry about what others hear for my projects. So having the ability to make real meaningful measurements is sometimes important.
And of course there really are better resistors! 🙂
And of course there really are better resistors! 🙂
In a power amp this "static" S/N ratio with 1khz applied, although not unimportant, I believe is far less important than the "dynamic" noise which comes when a complex multi frequency signal is applied with the amp driving a real life complex load. In the same way that a complex signal excites resonaces in loudspeakers & their cabinets, the after effects of which can be seen in a waterfall plot. I believe in many amps this dynamic signal / noise ratio is well within the range of human hearing and I imagine it could be measureable.
Did anyone try doing a waterfall plot at the amp's output instead of the speaker's output ? Would it be technically possible ? I would be interested to see that if it could be done.
If John is hoping to see measured differences in amps where all such internal resonaces have been eliminated - for instance the difference from changing resistor or capacitor types - I can't imagine that could be easily measured - even though I believe those differences can be heard subjectively.
Edit: Ahh - posted this before I saw John's reply above
Last edited:
Well but when you look at the ear as some kind of microphone this
is a highly non linear microphone. But these non linearity is used by the
brain in advance to extract additional information.
Not all "faults" of the reproduction system are created equal so to speak,
just reducing distortion (or whatever is the figure of merit of the day) is
just the wrong approach.
You cannot extract additional info if it isn't already present in whatever hits your timpany.
What the brain does is 'interprete' those vibrations. And that is out of your control, and unconsious. Thus, unreliable and not consistent.
Microphones, while not perfect, are quite a lot better at it.
jan
- Status
- Not open for further replies.