Ringing must never happen. Thats matter of Q. LR4 is not ringing afaik. I don't imply only sinewaves. They're just the only way to talk spectral analysis. Transient distortion you seem to have in mind is phase distortion. My take is that you can't physiologically hear transient distortion if phase distortion stays below some critical limit.Pan said:
markp, my point was that binaural recordings with headphones are the closest thing to reproducing the proper geometry, and making your ears hear what they would have heard had they been in the recording spot. With any other setup, gross problems from recording method to crosstalk (left-speaker-to-right-ear and reverse) would probably mask any of the smaller effects phase has on imaging. What I intended to say was that binaural recordings and headphones are the best way to study the audibility (indirect, as it influences human aural localization) of phase.
Linkwitz is using same XO at 1440Hz. His hearing tests at 100Hz are dictated by large group delay introduced at that f. Phase difference between ears is still important at 100Hz.Prune said:
Reread the localization paper. Localization is result of stereo - phase difference between ears, and amplitude - closer is louder. If you need proof - switch to mono from single speaker and try find any phase dependant localization cues.
Most probably that as it has several db insertion loss - it is reducing your speaker sensitivity. It is a nightmare of impedance matching.
Quality binaural recordings are impossible to make - they are too highly individual. In addition, whole of our ear canal is participating in aural processing - this would all need to be imitated, and there are no two similar ears on earth. There are though DSP's that imitate head transfer function, making it closer illusion that you are listening to the speakers. That really imitates acoustic interactions of sound waves, so you are back to where you started. Unechoic speaker tests are closest we can make at this point.Prune said:
Though the idea is not bad.
Hmm, references? I don't see how thats possible. More likely reverse: 4/1 wavelengths are needed. At 1/4 of wavelength frequency of the wave is completely ambigous.
Half wl is enough for math - we can always make hat tricks like "we can mirror the halfwave and have a full wave". But for our "spectral analyser" such tricks are forbidden - it must dismantle continuous flood of input noise into spectral composition in real time, and oopsies with retries is not what evolution would like to tolerate. False and confusing data is worse than no data.markp said:
Actually, I think I've heard how halfwave would sound. I was playing with raised cosine modulated tone bursts, tried to shape main frequency with half wl of itself - effectively generating wave of half wl. That didn't sound like the main frequency, more like higher frequency with click. I'll try to generate a wav sample and recheck.
I never said you can get (significant) localization cues from a single driver. However, I see no reason to think that identical phase distortion in each channel of a binaural setup won't have an effect on localization. This is because the distortion is being applied on top of the phase differences that were recorded, thus masking their recorded relationship, which is what the auditory neural circuits extract information from.
I thought it was clear it would be a line crossover, between the source and amplifiers. Check my first post -- obviously I can't drive a couple of kV plasma with a regular amplifier.
Not perfect, but better than anything else. I actually find that binaural recordings done with microphones in real ears (even if not mine) sound better than those done with a dummy head. Signal processing can deal with the issue if the listener's HRTF is known (imagine your DSP equipment has a stored personal profile for each listener). Add to that crosstalk cancellation processing, and you can even use speakers instead of headphones (I've heard proof-of-concept demos that show it can work).
Are you suggesting we are not allowed to change the subject, especially when it's to a related one? We are now talking about phase distortion in general. I mentioned headphones as I proposed phase distortion would be most audible on playback of binaural recordings. This actually has bearing on the crossover discussion, for binaural recordings can be played back on speakers with appropriate crosstalk cancellation DSP (though that only works in a limited sweet spot). Crossover phase nonlinearities, though identical in each channel, could still mask phase localization cues. Of course, driver alignments and room reflections also mess up things, but why not minimize the corrections the DSP needs to do?
I thought you imply that timing shifts between frequencies cause images to be perceived closer or further as thats the only imaging related idea you could think of for XO phase distortions. Thats why I proposed mono test - it'd show that such cues do not exist.Prune said:
I initially suggested avoiding XO in the freq range where our ears are sensitive to phase difference between channels - to avoid unequal phase distortions that are bound to happen in realistic XO implementations. That one is a real risk of being audible.
Yet your comments in this post suggests you assume it is possible that absolutely equal time shift from phase distortion on both channels can somehow ruin phase information originally recorded. This is what I disagree with. Phase distortion is linear distortion - it does not remove nor add any information in the signal. Equal phase distortion of both channels will absolutely not change any phase difference relationships recorded. I'm not sure if this is obvious, so I'd risk to restate here that "phase relationship" has meaning only for single frequency, not the whole complex signal, thus ear is processing phase relationships for multiple but unique frequencies simultaneously. Timing between those different frequencies is not a matter of phase, its a matter of temporal precision. Thats why I mentioned fundamental limits of that temporal precision, and the tests in Koya paper are about the same - they show we tolerate quite large temporal shifts. There are good reasons for that - that happens a lot in nature.
Right. You asked what that comment means. That XO was meant for purists with passive speakers. So you go for active XO and this does not apply to you. I assumed you would make that connection yourself. Why the impression I'm answering the wrong question?
HTF is actually least of an issue. Main trouble is simulating what happens inside the ears. And inear mic's aren't close enough to catch that. Using speakers instead of headphones gives some, but little benefits and adds issues of its own. I can't agree that binaural method is better than anything else - too many unknown variables, including nonlinear behaviour of ears must be dealt with to claim any results from binaural recording more conclusive than with professional speakers in 3D arrangement in unechoic chamber.
An externally hosted image should be here but it was not working when we last tested it.
Well, PC-speakers would be even more practicalPrune said:
The issue is if you'd trust the results. You might want to observe that: http://www.headphone.com/graphs/showGraph.php?graphID=255
ER-4B canalphones for binaural listening.
BTW, Linkwitz has some discussion about equalization for the ER-4S on this page.
BTW, Linkwitz has some discussion about equalization for the ER-4S on this page.
Consider prefering slightly independent testing: http://www.headphone.com/layout.php?specs=Y&topicID=13&subTopicID=64&productID=0020100010
Look at THD. Compare Senns http://www.headphone.com/layout.php?specs=Y&topicID=3&subTopicID=26&productID=0020080650
You don't want to know their phase response. And they don't want you to know it either
Notch filtering to equalize for phase distortion audibility tests??
Look at THD. Compare Senns http://www.headphone.com/layout.php?specs=Y&topicID=3&subTopicID=26&productID=0020080650
You don't want to know their phase response. And they don't want you to know it either
Notch filtering to equalize for phase distortion audibility tests??
I'd expect high-end electrostatic headphones like Stax Omega II to have lower distortion, just like electrostatic speakers distort less than dynamic ones. Also, take a look at these ultra rare ionic headpohnes:
http://membres.lycos.fr/plasmapropulsion/Industrial_issues/Plasmasonic.htm
http://membres.lycos.fr/plasmapropulsion/Industrial_issues/Plasmasonic.htm
Not really. Don't forget that good cans already have orders of magnitude less THD than speakers.Prune said:
http://www.headphone.com/layout.php...raphID[3]=&buttonSelection=Compare+Headphones
Um, now thats scary. 6000 Volts an inch from my brain, with plasma almost in my ears.. I wonder how they defined "loosing clients"..
DIY full range ESL project: http://www.ele.tut.fi/~artoko/audio/speakers/fullrange_esl.html
Quote: "Whole signal chain produces second harmonic distortion about 0.06%@1kHz with about 90 dBSPL." My guess is most of that is transformer distortion. Using transmitter tubes as output devices can drive this directly, and distortion would then be comparable to that of electronics, not drivers.
Step response shown near the end looks good, but pales in comparison to modified Plasmatronics (graph).
Quote: "Whole signal chain produces second harmonic distortion about 0.06%@1kHz with about 90 dBSPL." My guess is most of that is transformer distortion. Using transmitter tubes as output devices can drive this directly, and distortion would then be comparable to that of electronics, not drivers.
Step response shown near the end looks good, but pales in comparison to modified Plasmatronics (graph).
Impressive, however I guess the distortion in the bass would be much higher than the 0.06% at 1k, and higher compared to the most well designed dynamic woofers like TC underhung and similar (just my guess here).
For midrange duty ribbon designer Bo Bengtson claim less than 0.03% THD at 90dB with his new designs which has a very impressing square wave as well. The Accuton diamond tweeters has about 0.01-0.02% THD at 90dB in the high range. I don´t doubt that there are some really good e-stats out there though, but so far I have seen no evidence that they should be the lowest distortion designs. Also I´ve seen a waterfall graph on a ML and that one was not very pretty... and I have never heard a ML that sounded really good to my ears either, have listened to three different desings. The quad is nice though.
/Peter
For midrange duty ribbon designer Bo Bengtson claim less than 0.03% THD at 90dB with his new designs which has a very impressing square wave as well. The Accuton diamond tweeters has about 0.01-0.02% THD at 90dB in the high range. I don´t doubt that there are some really good e-stats out there though, but so far I have seen no evidence that they should be the lowest distortion designs. Also I´ve seen a waterfall graph on a ML and that one was not very pretty... and I have never heard a ML that sounded really good to my ears either, have listened to three different desings. The quad is nice though.
/Peter
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.