Yes, these were files generated by Malcolm Hawksford. Pano and I tried doing a double-blind sorting test on some files that Jan randomized and nailed them (albeit with some effort). It is a very subtle effect, to be sure, and the sensitivity of human hearing to phase is VERY dependent on what frequency, how much, and what rate. I *think* (but am not certain) that these were fourth-order all-passed somewhere in the midrange.
I am sure it is, which why it passed under the radar for so many years, but I also think that for some very peculiar cases (= rogue wave condition), phase errors can be very detectable. I didn't chose the ticking clock example randomly: this is the kind of sound that can be affected by such errors.
Even with a single ear, it is still possible to locate the spatial origin of a sharp clicking sound, like a small object dropped on a hard surface (the location accuracy is seriously impaired, but it is still present), but if the phases are wrong, everything is utterly messed and nothing is recognizable.
Well, not my ears anyway. I've have been in the lectronic game long enough to know that anthing faster than about 5 Hz just merges into a hum.
You mean occaisonally written up by quacks and the golden ear brigade who haven't a clue what they are doing. Like all the nonsense you see in audiophile magazines about special $100x speaker cables affecting the sound, etc. Supported by SPICE analysis sometimes! Of course if you have no idea how to use a jack plane, you won't make a piece of wood flat.
It is a fact that the ear is insensitive to phase. There is no doubt.
Those of us who have designed and engineered radio receivers for instance know that the tuned circuits in the RF and IF stages determine what a receiver does with a signal. If MF & HF receivers designed for listening to programmes, the RF and IF circuits are designed for what is called "maximally flat frequency" response, because the ear can readily tell variations from a flat frequency response. A typical high perfoemce radio have many such tuned circuits and together they really mess up the phasing.
However, if the radio is designed to recieve video or data, the pulse shape and therfore the phase response is important. These radios are designed for "maximally flat phase" response.
Maximally flat frequence response filters are called Butterworth filters. Maximally falt phase response filters are called Bessel filters. It turns out that you can't have both in the same circuit. But you can add LCR networks to equalise the phase response of a Butterworth response. Nobody bothers, because it makes no audible difference.
You can do a test yourself. Twist the bass & trebble controls in your stereo system. That does two things: a) changes the frequency response, and b) changes the phase response. You hear only (a), but feed your amp with a square wave and watch the output with an oscilloscope. Apart from with the tone controls in the flat position, you won't see a square wave.
Its not what you don't know...
...its what you "know" that's wrong that bites you
in this case you should look at the evidence, cites, even try the .wav in foobar's abx plugin - good headphones let me get 10 out of 10 on the 2nd harmonic reversal/"inverting the polarity" of the asymmetric waveform
sawtooth rising vs falling sharp edge phase can be heard too
...its what you "know" that's wrong that bites you
in this case you should look at the evidence, cites, even try the .wav in foobar's abx plugin - good headphones let me get 10 out of 10 on the 2nd harmonic reversal/"inverting the polarity" of the asymmetric waveform
sawtooth rising vs falling sharp edge phase can be heard too
What I did find interesting is the physiology, the fact that the sensors fire on the rarefaction (IIRC, I stand to be corrected) rather than the compression peaks in the audio wave. I think the audibility is demonstrated, but I think also that the tones (I listened to them) do not sound very different. So it might still be fair to say that the ear is relatively insensitive to phase. And anyone could be forgiven for not being absolutely up to date with regard to the evidence.
The fastest nerve cells can only fire at 150 to 200 Hz tops. That's close to the bottom octave of music. As with other senses in animals, the rate of firing and the number of nerves firing encodes signal strength. When you listen to a signal 500 Hz or more, clearly that's all the nerve cells can do. Unless you think that nerves can count cycles and fire on every 4th peak or something. There's no mechanism to do that. Even assuming the brain could handle it, which doesn't seem likely.
For very low frequencies, researchers have reported that nerve pathways leading from the cochlea tend to fire in phase with the sound wave but still firing more often for stronger signals. It appears this has evolved so that the brain has additional information to derive the direction of sound, but it isn't very good. That's why in a stereo or multichannel system we can get away with a single sub-woofer, and why TV sets with stereo sound typically had only a single woofer fed from both channels.
For frequencies above roughly around 300 Hz, our sense of direction is much better because the brain can use slight level differences in level between the two ears.
Without thinking, we automatically move our heads about so the brain can process the slight changes in ear-to-ear level difference correlated with head movement to work out direction. This only works where the wavelength is not long compared to the distance between our ears, so directional sense is very good only if frequencies above about 1 kHz are present.
Experiments have been done clamping the head of volunteers so they can't move it. Their sense of direction of sound still works but is very much compromised.
For very low frequencies, researchers have reported that nerve pathways leading from the cochlea tend to fire in phase with the sound wave but still firing more often for stronger signals. It appears this has evolved so that the brain has additional information to derive the direction of sound, but it isn't very good. That's why in a stereo or multichannel system we can get away with a single sub-woofer, and why TV sets with stereo sound typically had only a single woofer fed from both channels.
For frequencies above roughly around 300 Hz, our sense of direction is much better because the brain can use slight level differences in level between the two ears.
Without thinking, we automatically move our heads about so the brain can process the slight changes in ear-to-ear level difference correlated with head movement to work out direction. This only works where the wavelength is not long compared to the distance between our ears, so directional sense is very good only if frequencies above about 1 kHz are present.
Experiments have been done clamping the head of volunteers so they can't move it. Their sense of direction of sound still works but is very much compromised.
Last edited:
ignorance is correctable - but you have to make the effort
only off by an order of magnitude re effective "sample rate" - 4-8K pps max phase synchronous firing rate is common, even in humans a few classes of nerve fibers are hinted at being able to do 2x that - this may include bundle/volley coding of a single rank of hair cells with each individual fiber saturating at ~400 Hz - but the temporal information is shown several places to extend up to 4 k
have looked at my links, tried to educate yourself - the auditory nerve phase locking is decades old now but may be news to the phase audibility deniers
only off by an order of magnitude re effective "sample rate" - 4-8K pps max phase synchronous firing rate is common, even in humans a few classes of nerve fibers are hinted at being able to do 2x that - this may include bundle/volley coding of a single rank of hair cells with each individual fiber saturating at ~400 Hz - but the temporal information is shown several places to extend up to 4 k
have looked at my links, tried to educate yourself - the auditory nerve phase locking is decades old now but may be news to the phase audibility deniers
Last edited:
Jcx, I clicked on your links, Seebeck etc & had a quick read. Nothing there that refutes anything I've said.
Do you have a linke or refrence for you claim that nerves bundled or not, can go at 4 to 8 pps in phase synchonism? That's a rediculous claiim. To encode phase in a sysytem of parallel channels each habing afiring limit of 400 pps (your limit), you'd need some sort of synchonising or voting mechanism, as well a a measn of detecting the phase. There is neither.
Do you have a linke or refrence for you claim that nerves bundled or not, can go at 4 to 8 pps in phase synchonism? That's a rediculous claiim. To encode phase in a sysytem of parallel channels each habing afiring limit of 400 pps (your limit), you'd need some sort of synchonising or voting mechanism, as well a a measn of detecting the phase. There is neither.
It's real easy. I suggest you try it. Read what turns up carefully. You'll find that apart from some wacky science hits, authoritive papers report that in humans, the cells from the cochlea can only phase align to very low frequencies as I said. Depending on which book or paper you read, its variously reporting that the upper limit is around 250 Hz like I said in an earlier post.
I did see an NIH sponsored paper that reports a form of nerve pahse locking with tones well above 250 Hz - with no limit in frequency really. But on moderately carefull reading it turns out they were talking about phase looking to the tone envelope. What this means is that if you are hearing a tone, say 5 kHz or whatever, and it happens to be amplitude modulated by a lower modulating frequency, the nerve firing tends to concentrate at the peak of the modulating frequency if it is lower than about 250 Hz. Its not "phase locking" to the phase of the 5 KHz, its phase locking to the phase of the <250 Hz modulation.
This is as you should expect - nerve firing encodes the stimulus strength.
Googling <auditory nerve phase locking> also turned up some papers that report that birds that hunt in the dark and depend on sound to locate prey, such as owls, have auditory nerves able to phase lock up to 1.5 kHz or so. I note that their heads are much smaller, so they have a greater need. We are not birds.
As Douglas Self said "It would take spectacular incompetence to make an amplifier that accidentally included an all-pass filter". The phase shifts in audio amplifiers are quite small and gradual and there is no evidence that they are audible, even if the ear was sensitive to relative phase.
If you want to worry about phase, go and worry about speakers and crossovers.
If you want to worry about phase, go and worry about speakers and crossovers.
I'll agree that few amps have audible phase shift
but you might want to know how far to trust some outspoken "expert" for the rest of his advice when most here will easily be able to prove phase audibility for themselves
if you had followed my evidence, pointers to other threads, you might have noticed Ethan Winter who was skeptical finally did the abx test after much badgering and did get a positive listening result - he has written about debunking audiophoolishness before http://ethanwiner.com/index.htm had to change his tune on the audibility of harmonic phase when he did the test
I just repeated 10/10 with the 400+800 Hz test files, made new 600+1200 Hz test wav, managed 9/10 on 1st try - 57 year old ears and I have a cold!
foobar2000 + abx plugin, Senn HD600 from my PC, foobar mixer level 5 out of 100, ~70 dB SPL measured with RadioShack digital sound level meter, no coupling fixture though
including the LTspice file for generating the tones
just edit the V1,2 Vsource frequencies, keeping the 1:2 ratio
the .wav files are written out at the end of the sim run to the same folder as the .asc
I did try 800/1600 with no positive discrimination - but again this is on the 1st try, no training, tuning of the test files
but you might want to know how far to trust some outspoken "expert" for the rest of his advice when most here will easily be able to prove phase audibility for themselves
if you had followed my evidence, pointers to other threads, you might have noticed Ethan Winter who was skeptical finally did the abx test after much badgering and did get a positive listening result - he has written about debunking audiophoolishness before http://ethanwiner.com/index.htm had to change his tune on the audibility of harmonic phase when he did the test
I just repeated 10/10 with the 400+800 Hz test files, made new 600+1200 Hz test wav, managed 9/10 on 1st try - 57 year old ears and I have a cold!
foobar2000 + abx plugin, Senn HD600 from my PC, foobar mixer level 5 out of 100, ~70 dB SPL measured with RadioShack digital sound level meter, no coupling fixture though
including the LTspice file for generating the tones
just edit the V1,2 Vsource frequencies, keeping the 1:2 ratio
the .wav files are written out at the end of the sim run to the same folder as the .asc
I did try 800/1600 with no positive discrimination - but again this is on the 1st try, no training, tuning of the test files
Last edited:
I don't necessarily trust outspoken experts, but I do trust science. Any nonlinearity will convert changes in harmonic phase into changes in harmonic amplitude.
We know the ear is nonlinear, so it follows that under some conditions it must be able to hear changes in phase as changes in timbre. The question is just how big the changes need to be before they are audible.
I think they need to be pretty big. The test signals you are using are quite radical phase shifts, you would need a really high order filter to inflict those on music.
We know the ear is nonlinear, so it follows that under some conditions it must be able to hear changes in phase as changes in timbre. The question is just how big the changes need to be before they are audible.
I think they need to be pretty big. The test signals you are using are quite radical phase shifts, you would need a really high order filter to inflict those on music.
Which we have in spades in both 44/16 brick wall filters and in Nth order bandpass boxes. Linkwitz further postulates (see his website) that group delay will be audible for low frequency phase shifts
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.