Is this really appeal to authority? I had to read the definition right now, but my understanding says that it would be if JDPA had done the study and Spy Handler said "they're usually right, so they're right here". But here, Spy Handler is claiming that experimenter bias may have influenced the execution of the experiment. This is always a valid concern in scientific experiments, though I personally don't buy it as the claim is made that all packages left Berlin at the same time and I don't believe they're lying. So I consider experimenter bias to be an unlikely source of violated assumptions, but you can't say it's completely without merit, right?
For the reasons you state, the test of statistical significance does not allow us to conclude that the "true" ratio is close to 9:1. The statistical test shows that the ratio is significantly greater than 1. This is the interesting claim, and this conclusion should meet your criterion of "statistical rigour".
"All packages left Berlin via Deutsch Post at the same time on 21.22.12"
It's possible that two such batches straddled a cutoff in the hand-off to USPS, and that might indeed explain the results of the study*, but you present no evidence to support your claim that it was "likely". Care to elaborate?
*Somewhere in the comment above I read that many of the packages arrived at the same time, which would in fact disprove your cutoff-straddle hypothesis, but I can't find the data
Considering that each recipient received one of each package, it seems unlikely that the marked vs. unmarked distinction being grouped unevenly into different trucks can account for the statistically significant results. But it is a valid criticism to warrant concern over the statistical test's assumptions.
Dividing 2 numbers does not result in "statistics"
No, but the Wilcoxon sign rank test with p<0.001 does.
There's also the issue of controlling for sample bias, I'm guessing that the recipients were either existing punters or friends (hence participating in the experiment) and could thus conveniently "lose" their marked packages if it help further the "OMG nasty fundies are persecuting our atheist brethren" agenda.
Each recipient received one of each, shipped on the same day. I find it an unlikely explanation that there was bias on that end.
Heh, I'd also bet good money that you could write up a patent for such a non-causal stereo and get that patent approved.
I thought of one more reason why I still prefer to consider those intermediate frequencies to exist. If you don't--that is, if you just consider the fourier series represented by the samples and nothing else--then it's very easy to run into time-aliasing problems. Without the intermediate frequencies, you're really representing the signal that is made periodic by repeating your N samples indefinitely ([... y y y y y...]). When you filter this signal, the (n>0) response of the filter to y(8) will wrap around to y([1,2,...]). So you could try to use y as above as a filter to chop off the highest frequency component, but then either suffer time-aliasing or explicitly have to zero-pad the signal to avoid it. Of course, your preferred flavor of the math is your own choice. This is mine, but I'd say it's really 6 of one, integral(sin(x)^2, x, 0, 2*pi) of a dozon of the other.
There's probably also an argument to be made about what happens if you frequency-shift the signal. I suspect you'll find those intermediate frequencies causing some headaches in the math if you do it your way, but it's too early in the morning for me to really think that one through. I'll get back to you once I can get some sound out of this true-sinc stereo.
I'm about to head out but I will respond quickly. I can get back at it tomorrow.
There are an infinite number of continuous signals that could be sampled this way.
But only one for an ideal band-limited sampling / reconstruction scheme such as pulse-amplitude modulation.
My main point is that those "intermediate" frequencies do in fact exist and are important for the analog reconstruction of the signal, even if they do not contribute any extra information to your digital signal (and they don't). You are not creating them by using a larger window, that's just a mathematically-equivalent operation to computing more samples of the DTFT. Were you to put y through a DAC, those frequencies will be there in the FT of the analog signal. The samples of the DFT of y produced by zero-padding the 8-sample signal (you read the documentation correctly) before taking the DTFT tell you exactly what the value of the DFT are at those "intermediate" frequencies.
The bottom line argument is that you cannot represent a finite-bandwidth signal with a finite-time signal (using fourier transforms)
However, for the periodic series created by repeating y (which is infinite-time), then your arguments hold. The fourier series represented by the DFT of y is a sum of delta functions in continuous frequency.
So our difference is just that you're considering the periodic signal of y repeated (which is of course how you derive the DFT in the first place), while I'm considering the signal y which is assumed to be zero outside of its support. I'll argue that mine is more relevant for analog signal reconstruction. If I reconstruct my y signal in the analog domain, I don't want that click repeating every 8/(sample freq) seconds out to infinite time (which is truly band-limited and can by further band limited in the way you describe), I want just the one click (and this signal cannot be band limited, analog or digitally, because it is time-limited).
Sorry, no time to proof read. Post may, as always, contain errors.
Where does that figure come from? A CD is a perfect reproduction of the analogue master.
Not quite true; even assuming your master is completely band limited to 22.05 kHz (or also at 44.1 kHz sample rate), you still have quantization error.
In medical tests, people are given a placebo and yet claim to feel better or feel the same effects as people who are given the real medication. These must be the same people who rail against mp3s.
Don't dis the placebo effect, it works (for some limited benefits), even in cases where the subjects were aware that they were receiving a placebo
The most similar analogy would be to say that someone can enjoy lossless music more than lossy music. This could be true even if they can't tell them apart in a blind study. Of course, under these assumptions, they'd also enjoy lossy music more than lossless music if the labels were switched and they believed the labels. It's enjoyed more simply because of what it is believed to be. That may be silly, but hey, who am I to crap on someone's enjoyment?
On the other hand, making the claim that you can tell the difference, i.e. discriminate between then, is more directly challengeable and probably false in most cases.
And I'll bet neither can you. The sawtooth is the sine wave plus a lot of higher-frequency harmonics. But your ear can't detect any of the higher-frequency harmonics of a signal with 22kHz periodicity, so you probably can't tel them apart.
(This is fun; I know we agree on all substantive points but I'm still going to take you on here:-)>
1) This is a misconception. The DTFT only represents samples of the DFT, and you can only work with the DTFT with any real machine with finite computing resources. If you zero out the DTFT samples, you are *not* zeroing out all DFT samples in between them.
Example MATLAB code:
x = [1 0 0 0 0 0 0 0];
X = fft(x);
Y = X; Y(5) = 0; %zero out the highest frequency component
y = ifft(Y);
stem(abs(Y,8)) %Look at the pretty DTFT with zero amplitude at the pi frequency component!
stem(abs(fft(y,256))) %Plot a finer sampling of the DFT. What happened to your perfect cutoff??
2) True, despite the 22 kHz cutoff. f>22 kHz wraps around to the negative frequency region first. That is, w>pi wraps around to the [-pi:0] region before getting back into the [0:pi] region; remember, we have 2*pi periodicity in the DFT, and 0:pi here represents 0:22 kHz. 22:44 kHz is pi:2*pi, which by periodicity is the same as -pi:0. An aliased, rising tone falls continuously from fs/2 to 0 before rising again.
Yes, for both bit depth and sampling frequency. Here are two possible reasons why:
1. Bit depth. Remix wants to amplify a sound in the original mix. At 16 bit depth, you have 2^16 possible values to cover everything from silent to max loudness. If you take a soft sound that uses only some of those values and amplify it, the result suffers from possibly noticeable quantization artifacts. This is like magnifying a small picture to produce a pixelated one.
2. Sample frequency. Remix wants to frequency-shift / pitch-shift a sound in the original mix. Your sampling rate determine the max frequency you can encode, so any audio in a 44.a kHz file has a max frequency range of 22.05 kHz. Say you shift something down by an octave (factor or 1/2); the shifted sound will be cut off at 11.025 kHz.
How much these effects are noticeable in typical mixes is up to the listener...
Two nits to pick:
1) You can get arbitrarily close but you can't get "perfect" frequency cutoff.
2) A 25 kHz tone sampled at 44 kHz gives you a 19 kHz tone. Remember the [-pi:0] (or [pi:2*pi]) frequency range comes first.A 41 kHz tone would get you a 3 kHz tone after sampling.
Otherwise all true, which is why most recording devices do exactly that, sample at a high rate and digitally filter before downsampling to 44.1. But none of that has much to do with whether or not, once you've gotten past the aliasing problem as you say, you can tell the difference between a 44.1 kHz playback and a 96 kHz playback.
First, don't assume the government has your best interests at heart (they don't), and second assume everything you do will be fully and completely monitored without the slightest expectation of privacy.
No, the law says you can't actually insult or defame the king.
He stopped short of insulting the king. It does not matter that he communicated what he felt about the king, what matters is did he actually insult or defame him.
Yes it was clear he was talking about the king.
Here's where you lose me. If it was expressing obvious that he was talking about the king, whether by gesture or even simply highlighted omission, then let's look at why your next premise is or is not true: that it matters exactly what words he said and if he explicitly mentioned the king. Why? I'll grant you that if you try to objectively define the law, then it probably matters. But who says laws have to be enforced or defined objectively? I mean, I personally think they often should for several reasons, notwithstanding the fact that objective definitions usually suck at capturing real-life situation. But who am I to the laws of Thailand?
Obviously to any of us, the laws are screwed up in the first place. But under the screwed-up law, I see no reason an obvious implication, even without direct statement, can't be interpreted legally as communication all the same. I believe this was OP's point.
If I say "President Bush was so [highlighted omission]..... he failed to catch Bin Laden", then I have studiously avoided insulting him.
To me it's pretty obvious what you mean, and thus the insult is complete regardless of the actual words used. Pauses and implication are a part of most languages, and words even including pauses are only one way of communicating. If you want to define insults as strictly of words spoken, then you have not insulted Bush above. But I think that's a pretty useless definition.
Interesting research, seemingly terrible reporting by Engadget. Seems to me that MS did not "[cut] touchscreen lag to 1ms", they simply demonstrated via a test device (which appears to be projecting light from above) what different amounts of lag look / feel like.
Or maybe Paul Shawcross did this in his spare time or break time? While I don't overwhelmingly approve of the speed, thoroughness, or appropriateness of many white house responses, you can't simply assume this response means that other duties were being shirked.
You miss my point though. I understand that your foot is making the same motion, and that you have the same mechanical leverage given the distance between your foot and the gear center.
However, the angle at which the crank connects to the gear is now shifted from the angle between your foot and the center of the gear (relative to horizontal). In a simple free-body diagram, this also makes zero difference, as the angle of connection does not affect torque. But since the crank actually connects at various points along the bolt, the angle of connection could determine the direction of force at these various connection points. This of course has absolutely nothing to with the z shape, and the same putative effect could probably be achieved simply by changing the shape of the connector. Thus I'd agree that that extra metal going into the elbow of the angled crank is really doing nothing.
So all I'm investigating, in thought, is how the time-varying direction of forces applied during pedaling may be shifted depending on how the crank connects to the gear. If you are inclined to believe any of the data produced by this dude, which I realize is suspect, then this graph (available in their gallery) presents data qualitatively in line with my reasoning, and, interestingly, NOT in line with their reasoning (which I think we both agree is grossly incorrect). I'd further note that if I were making up data to fit their explanations, I'd have put in some at least small magnitude changes, not just a phase shift as we see here. So is it possible that there is some (probably very small) effect of angle of connection?
Lastly, what does a phase shift do for total work? Nothing for a given amount of force, I think. But that's where biomechanics (OP) comes in; since the leg isn't equally efficient at all pedal positions, this phase shift could result in a change (positive or negative) in total efficiency.
Again, not a mechanical engineer. I'm just speculating on what could cause differences that might not show up in a simple physics model. All models are imperfect, so I'm trying to challenge the model assumptions. Does this make sense, and do you agree that this is a different issue than you present in your rebutta to my post?
(Not a mechanical engineer)
That's if you measure by foot position. Your legs are indeed doing the exact same thing (in steady state), but I see this as shifting the phase of the angle of your foot relative to the angle of the crank (at it's connection to the gear). In an ideal, simple model, in which the crank was a 0-width, rigid line connected perfectly to the center point of the gear, I can see that that wouldn't make any difference. But since it's not, is some difference possible in this setup?
That's not directly related to the earbud quality though. Using a splitter, you're just putting two (well, four) resistors in parallel. The current will go mostly through the one with lowest resistance (lowest headphone impedance; it's listed on the packaging of some headphones). That's entirely relative between headphones and doesn't tell you much about how the devices perform on their own.
Of course you're right, the bleeding audio comment was more of a "damn kids and your music" than a scientific remark. It would actually depend on everything from the design of the headphones to the shape of the listener's ear and its orientation relative to me. I still think it's a reasonable intuition when it's coming over iPod earbuds, leaky as they may be.
Not all tinnitus is related to the kind of hearing damage under discussion, but for the purpose of this discussion, it's the volume, or, more specifically, the distribution of energy entering your ear. The tricky bit is that that will change for the same song and listener depending on the headphones and external noise. tl;dr = keep the volume down, but I think hanging buds are probably the worst, while well-fitted buds probably aren't so bad.
Mostly, I believe, due to the different frequency response of the headphones (e.g. treble / bass balance, but it gets more complicated that that). Generally, lack of a fit / seal means much less bass, but larger speakers are better at delivering bass that small buds. So if bass is missing but is important to the song, you might turn up the volume. But now the highs are louder than they would have been with fitted buds or larger earphones, and this might cause relatively more damage. As I see it, this is why different headphones make a difference, because they cause you to turn overall volume up or down and that can result in some parts of the spectrum being presented at damaging levels. Ultimately, for any frequency, it's the level and amount of time at that level that does the damage, so if you can be aware of what you're putting into your ears you're already better off.
But, of course, none of that is a guarantee against either hearing loss or tinnitus. It might help your chances though.
It might be true that most listeners aren't subjecting themselves to great damage over their headphones, but even then I think you're drawing too quick a conclusion with too little data. And certainly on the (not-too-uncommon) instances when I can listen to the music of the guy sitting a few seats down from me on the bus through his headphones, I think there's likely some damage going on there.
Note also that the threshold of pain for hearing is often measured well above where damage starts to occur. Maladaptive, maybe, but it seems to be pretty true nonetheless. And if I'm not mistaken, those studies were done before the current research coming out showing that even when loud sounds do not cause permanent shifts in hearing thresholds, permanent nerve damage is still being done. Thresholds for pain are often well above 100 dB SPL (for broadband stimuli), while damage can be done with prolonged exposure well below that.
Lastly, a minor point, but I don't know what the logarithmic nature of the dB scale has to do with any of this. It's just a scale, a way to place arbitrary numbers on physical phenomena.
Warranted or not, you have to admit that controlling (or banning) those tools that are simultaneously easiest and most dangerous may likely reduce deaths in such occurrences. You can still argue for or against banning or other means of control, but don't so confidently state that it wouldn't accomplish anything
I challenge you - and me - and all of us - to figure out why the hell people want to kill kids in a school. Answering that question may be a lot harder, but that's the task we need to be putting our efforts into.
Slashdot Mirror
Is this really appeal to authority? I had to read the definition right now, but my understanding says that it would be if JDPA had done the study and Spy Handler said "they're usually right, so they're right here". But here, Spy Handler is claiming that experimenter bias may have influenced the execution of the experiment. This is always a valid concern in scientific experiments, though I personally don't buy it as the claim is made that all packages left Berlin at the same time and I don't believe they're lying. So I consider experimenter bias to be an unlikely source of violated assumptions, but you can't say it's completely without merit, right?
For the reasons you state, the test of statistical significance does not allow us to conclude that the "true" ratio is close to 9:1. The statistical test shows that the ratio is significantly greater than 1. This is the interesting claim, and this conclusion should meet your criterion of "statistical rigour".
"All packages left Berlin via Deutsch Post at the same time on 21.22.12"
It's possible that two such batches straddled a cutoff in the hand-off to USPS, and that might indeed explain the results of the study*, but you present no evidence to support your claim that it was "likely". Care to elaborate?
*Somewhere in the comment above I read that many of the packages arrived at the same time, which would in fact disprove your cutoff-straddle hypothesis, but I can't find the data
Considering that each recipient received one of each package, it seems unlikely that the marked vs. unmarked distinction being grouped unevenly into different trucks can account for the statistically significant results. But it is a valid criticism to warrant concern over the statistical test's assumptions.
Dividing 2 numbers does not result in "statistics"
No, but the Wilcoxon sign rank test with p<0.001 does.
There's also the issue of controlling for sample bias, I'm guessing that the recipients were either existing punters or friends (hence participating in the experiment) and could thus conveniently "lose" their marked packages if it help further the "OMG nasty fundies are persecuting our atheist brethren" agenda.
Each recipient received one of each, shipped on the same day. I find it an unlikely explanation that there was bias on that end.
Heh, I'd also bet good money that you could write up a patent for such a non-causal stereo and get that patent approved.
I thought of one more reason why I still prefer to consider those intermediate frequencies to exist. If you don't--that is, if you just consider the fourier series represented by the samples and nothing else--then it's very easy to run into time-aliasing problems. Without the intermediate frequencies, you're really representing the signal that is made periodic by repeating your N samples indefinitely ([... y y y y y ...]). When you filter this signal, the (n>0) response of the filter to y(8) will wrap around to y([1,2,...]). So you could try to use y as above as a filter to chop off the highest frequency component, but then either suffer time-aliasing or explicitly have to zero-pad the signal to avoid it. Of course, your preferred flavor of the math is your own choice. This is mine, but I'd say it's really 6 of one, integral(sin(x)^2, x, 0, 2*pi) of a dozon of the other.
There's probably also an argument to be made about what happens if you frequency-shift the signal. I suspect you'll find those intermediate frequencies causing some headaches in the math if you do it your way, but it's too early in the morning for me to really think that one through. I'll get back to you once I can get some sound out of this true-sinc stereo.
I'm about to head out but I will respond quickly. I can get back at it tomorrow.
There are an infinite number of continuous signals that could be sampled this way.
But only one for an ideal band-limited sampling / reconstruction scheme such as pulse-amplitude modulation.
My main point is that those "intermediate" frequencies do in fact exist and are important for the analog reconstruction of the signal, even if they do not contribute any extra information to your digital signal (and they don't). You are not creating them by using a larger window, that's just a mathematically-equivalent operation to computing more samples of the DTFT. Were you to put y through a DAC, those frequencies will be there in the FT of the analog signal. The samples of the DFT of y produced by zero-padding the 8-sample signal (you read the documentation correctly) before taking the DTFT tell you exactly what the value of the DFT are at those "intermediate" frequencies.
The bottom line argument is that you cannot represent a finite-bandwidth signal with a finite-time signal (using fourier transforms)
However, for the periodic series created by repeating y (which is infinite-time), then your arguments hold. The fourier series represented by the DFT of y is a sum of delta functions in continuous frequency.
So our difference is just that you're considering the periodic signal of y repeated (which is of course how you derive the DFT in the first place), while I'm considering the signal y which is assumed to be zero outside of its support. I'll argue that mine is more relevant for analog signal reconstruction. If I reconstruct my y signal in the analog domain, I don't want that click repeating every 8/(sample freq) seconds out to infinite time (which is truly band-limited and can by further band limited in the way you describe), I want just the one click (and this signal cannot be band limited, analog or digitally, because it is time-limited).
Sorry, no time to proof read. Post may, as always, contain errors.
Where does that figure come from? A CD is a perfect reproduction of the analogue master.
Not quite true; even assuming your master is completely band limited to 22.05 kHz (or also at 44.1 kHz sample rate), you still have quantization error.
In medical tests, people are given a placebo and yet claim to feel better or feel the same effects as people who are given the real medication. These must be the same people who rail against mp3s.
Don't dis the placebo effect, it works (for some limited benefits), even in cases where the subjects were aware that they were receiving a placebo
The most similar analogy would be to say that someone can enjoy lossless music more than lossy music. This could be true even if they can't tell them apart in a blind study. Of course, under these assumptions, they'd also enjoy lossy music more than lossless music if the labels were switched and they believed the labels. It's enjoyed more simply because of what it is believed to be. That may be silly, but hey, who am I to crap on someone's enjoyment?
On the other hand, making the claim that you can tell the difference, i.e. discriminate between then, is more directly challengeable and probably false in most cases.
Your 44.1khz sampler can't distinguish them.
And I'll bet neither can you. The sawtooth is the sine wave plus a lot of higher-frequency harmonics. But your ear can't detect any of the higher-frequency harmonics of a signal with 22kHz periodicity, so you probably can't tel them apart.
(This is fun; I know we agree on all substantive points but I'm still going to take you on here :-)>
1) This is a misconception. The DTFT only represents samples of the DFT, and you can only work with the DTFT with any real machine with finite computing resources. If you zero out the DTFT samples, you are *not* zeroing out all DFT samples in between them.
Example MATLAB code:
x = [1 0 0 0 0 0 0 0];
X = fft(x);
Y = X; Y(5) = 0; %zero out the highest frequency component
y = ifft(Y);
stem(abs(Y,8)) %Look at the pretty DTFT with zero amplitude at the pi frequency component!
stem(abs(fft(y,256))) %Plot a finer sampling of the DFT. What happened to your perfect cutoff??
2) True, despite the 22 kHz cutoff. f>22 kHz wraps around to the negative frequency region first. That is, w>pi wraps around to the [-pi:0] region before getting back into the [0:pi] region; remember, we have 2*pi periodicity in the DFT, and 0:pi here represents 0:22 kHz. 22:44 kHz is pi:2*pi, which by periodicity is the same as -pi:0. An aliased, rising tone falls continuously from fs/2 to 0 before rising again.
Your move, good sir
Yes, for both bit depth and sampling frequency. Here are two possible reasons why:
1. Bit depth. Remix wants to amplify a sound in the original mix. At 16 bit depth, you have 2^16 possible values to cover everything from silent to max loudness. If you take a soft sound that uses only some of those values and amplify it, the result suffers from possibly noticeable quantization artifacts. This is like magnifying a small picture to produce a pixelated one.
2. Sample frequency. Remix wants to frequency-shift / pitch-shift a sound in the original mix. Your sampling rate determine the max frequency you can encode, so any audio in a 44.a kHz file has a max frequency range of 22.05 kHz. Say you shift something down by an octave (factor or 1/2); the shifted sound will be cut off at 11.025 kHz.
How much these effects are noticeable in typical mixes is up to the listener...
Two nits to pick:
1) You can get arbitrarily close but you can't get "perfect" frequency cutoff.
2) A 25 kHz tone sampled at 44 kHz gives you a 19 kHz tone. Remember the [-pi:0] (or [pi:2*pi]) frequency range comes first.A 41 kHz tone would get you a 3 kHz tone after sampling.
Otherwise all true, which is why most recording devices do exactly that, sample at a high rate and digitally filter before downsampling to 44.1. But none of that has much to do with whether or not, once you've gotten past the aliasing problem as you say, you can tell the difference between a 44.1 kHz playback and a 96 kHz playback.
First, don't assume the government has your best interests at heart (they don't), and second assume everything you do will be fully and completely monitored without the slightest expectation of privacy.
Yep, sounds like similar service to me
No, the law says you can't actually insult or defame the king.
He stopped short of insulting the king. It does not matter that he communicated what he felt about the king, what matters is did he actually insult or defame him.
Yes it was clear he was talking about the king.
Here's where you lose me. If it was expressing obvious that he was talking about the king, whether by gesture or even simply highlighted omission, then let's look at why your next premise is or is not true: that it matters exactly what words he said and if he explicitly mentioned the king. Why? I'll grant you that if you try to objectively define the law, then it probably matters. But who says laws have to be enforced or defined objectively? I mean, I personally think they often should for several reasons, notwithstanding the fact that objective definitions usually suck at capturing real-life situation. But who am I to the laws of Thailand?
Obviously to any of us, the laws are screwed up in the first place. But under the screwed-up law, I see no reason an obvious implication, even without direct statement, can't be interpreted legally as communication all the same. I believe this was OP's point.
If I say "President Bush was so [highlighted omission]..... he failed to catch Bin Laden", then I have studiously avoided insulting him.
To me it's pretty obvious what you mean, and thus the insult is complete regardless of the actual words used. Pauses and implication are a part of most languages, and words even including pauses are only one way of communicating. If you want to define insults as strictly of words spoken, then you have not insulted Bush above. But I think that's a pretty useless definition.
Interesting research, seemingly terrible reporting by Engadget. Seems to me that MS did not "[cut] touchscreen lag to 1ms", they simply demonstrated via a test device (which appears to be projecting light from above) what different amounts of lag look / feel like.
Or maybe Paul Shawcross did this in his spare time or break time? While I don't overwhelmingly approve of the speed, thoroughness, or appropriateness of many white house responses, you can't simply assume this response means that other duties were being shirked.
You miss my point though. I understand that your foot is making the same motion, and that you have the same mechanical leverage given the distance between your foot and the gear center.
However, the angle at which the crank connects to the gear is now shifted from the angle between your foot and the center of the gear (relative to horizontal). In a simple free-body diagram, this also makes zero difference, as the angle of connection does not affect torque. But since the crank actually connects at various points along the bolt, the angle of connection could determine the direction of force at these various connection points. This of course has absolutely nothing to with the z shape, and the same putative effect could probably be achieved simply by changing the shape of the connector. Thus I'd agree that that extra metal going into the elbow of the angled crank is really doing nothing.
So all I'm investigating, in thought, is how the time-varying direction of forces applied during pedaling may be shifted depending on how the crank connects to the gear. If you are inclined to believe any of the data produced by this dude, which I realize is suspect, then this graph (available in their gallery) presents data qualitatively in line with my reasoning, and, interestingly, NOT in line with their reasoning (which I think we both agree is grossly incorrect). I'd further note that if I were making up data to fit their explanations, I'd have put in some at least small magnitude changes, not just a phase shift as we see here. So is it possible that there is some (probably very small) effect of angle of connection?
Lastly, what does a phase shift do for total work? Nothing for a given amount of force, I think. But that's where biomechanics (OP) comes in; since the leg isn't equally efficient at all pedal positions, this phase shift could result in a change (positive or negative) in total efficiency.
Again, not a mechanical engineer. I'm just speculating on what could cause differences that might not show up in a simple physics model. All models are imperfect, so I'm trying to challenge the model assumptions. Does this make sense, and do you agree that this is a different issue than you present in your rebutta to my post?
(Not a mechanical engineer)
That's if you measure by foot position. Your legs are indeed doing the exact same thing (in steady state), but I see this as shifting the phase of the angle of your foot relative to the angle of the crank (at it's connection to the gear). In an ideal, simple model, in which the crank was a 0-width, rigid line connected perfectly to the center point of the gear, I can see that that wouldn't make any difference. But since it's not, is some difference possible in this setup?
That's not directly related to the earbud quality though. Using a splitter, you're just putting two (well, four) resistors in parallel. The current will go mostly through the one with lowest resistance (lowest headphone impedance; it's listed on the packaging of some headphones). That's entirely relative between headphones and doesn't tell you much about how the devices perform on their own.
Of course you're right, the bleeding audio comment was more of a "damn kids and your music" than a scientific remark. It would actually depend on everything from the design of the headphones to the shape of the listener's ear and its orientation relative to me. I still think it's a reasonable intuition when it's coming over iPod earbuds, leaky as they may be.
Not all tinnitus is related to the kind of hearing damage under discussion, but for the purpose of this discussion, it's the volume, or, more specifically, the distribution of energy entering your ear. The tricky bit is that that will change for the same song and listener depending on the headphones and external noise. tl;dr = keep the volume down, but I think hanging buds are probably the worst, while well-fitted buds probably aren't so bad.
Mostly, I believe, due to the different frequency response of the headphones (e.g. treble / bass balance, but it gets more complicated that that). Generally, lack of a fit / seal means much less bass, but larger speakers are better at delivering bass that small buds. So if bass is missing but is important to the song, you might turn up the volume. But now the highs are louder than they would have been with fitted buds or larger earphones, and this might cause relatively more damage. As I see it, this is why different headphones make a difference, because they cause you to turn overall volume up or down and that can result in some parts of the spectrum being presented at damaging levels. Ultimately, for any frequency, it's the level and amount of time at that level that does the damage, so if you can be aware of what you're putting into your ears you're already better off.
But, of course, none of that is a guarantee against either hearing loss or tinnitus. It might help your chances though.
It might be true that most listeners aren't subjecting themselves to great damage over their headphones, but even then I think you're drawing too quick a conclusion with too little data. And certainly on the (not-too-uncommon) instances when I can listen to the music of the guy sitting a few seats down from me on the bus through his headphones, I think there's likely some damage going on there.
Note also that the threshold of pain for hearing is often measured well above where damage starts to occur. Maladaptive, maybe, but it seems to be pretty true nonetheless. And if I'm not mistaken, those studies were done before the current research coming out showing that even when loud sounds do not cause permanent shifts in hearing thresholds, permanent nerve damage is still being done. Thresholds for pain are often well above 100 dB SPL (for broadband stimuli), while damage can be done with prolonged exposure well below that.
Lastly, a minor point, but I don't know what the logarithmic nature of the dB scale has to do with any of this. It's just a scale, a way to place arbitrary numbers on physical phenomena.
it just as easily could have been a bomb
For many, that's not just as easy
poison gas
ditto
or a well-aimed car
which would likely not have killed over 20 people
Warranted or not, you have to admit that controlling (or banning) those tools that are simultaneously easiest and most dangerous may likely reduce deaths in such occurrences. You can still argue for or against banning or other means of control, but don't so confidently state that it wouldn't accomplish anything
I challenge you - and me - and all of us - to figure out why the hell people want to kill kids in a school. Answering that question may be a lot harder, but that's the task we need to be putting our efforts into.
Absolutely agreed
I am no expert and welcome rebutting data, but the data presented here argue against what you claim.