Chapter 6
cited by http://www.melaudia.net/zdoc/jml_crossovers_etf04.pdf
no clear conclusion I could find on thresholds, but yes sometimes phase distortion/group delay variation can be heard, sometimes, with some signals, at some frequencies...
headphone listening way more sensitive, except when not...
I would say any "audiophile" headphone targeting flat bass actually achieves better bass smoothness than most really good room+loudspeaker systems
in particular higher quality IEM, say US$100 Etymotic give flat bass response octaves lower than most speaker companies' anechoic chamber's low frequency limits
cited by http://www.melaudia.net/zdoc/jml_crossovers_etf04.pdf
no clear conclusion I could find on thresholds, but yes sometimes phase distortion/group delay variation can be heard, sometimes, with some signals, at some frequencies...
headphone listening way more sensitive, except when not...
I would say any "audiophile" headphone targeting flat bass actually achieves better bass smoothness than most really good room+loudspeaker systems
in particular higher quality IEM, say US$100 Etymotic give flat bass response octaves lower than most speaker companies' anechoic chamber's low frequency limits
Last edited:
Hi,
Helmuth, let me precise #7.
If we want to make an statement about GD we first need to exclude amplitude differences between alignments. Using for example a simple high-Q highpass we can equalize a CB to a similar amplitude response as a BR alignment.
Than the differences in measured GD are negligible small, since the electronic HP adds GD to the CBs mechanical GD. But even though the CB requires more excursion and will show higher THD it´ll still sound more precise than the BR alignment. So either there must be a human perception difference between electronically generated GD and mechanically generated GD, or the GD is not the decicive parameter at all. One should notice a remarkable difference in sonic precision with a rather ´boomy´ BR alignment, corrected by a FIR filter. But afaik this doesn´t apply and all reported impovements due to FIR-filtering took place in the higher registers, but not the lowest octaves. This implies that electronic GD could be sonically transparent at LF. But I wouldn´t bet on that and haven´t seen a report or paper about it.
As long as this isn´t proven, one should also take other sources than the GD into acount as responsible factors.
jauu
Calvin
The editor of a german DIY magazine for example claimed the RMS-value to be the decicive parameter regarding bass precision. The discussion about this is still going on, even though it should be quite clear that the RMS-value is a ´artificial´ parameter, itself depending on several other factors like Membrane-area. As such it could only be a rough indicator but no real or precise factor.
Helmuth, let me precise #7.
If we want to make an statement about GD we first need to exclude amplitude differences between alignments. Using for example a simple high-Q highpass we can equalize a CB to a similar amplitude response as a BR alignment.
Than the differences in measured GD are negligible small, since the electronic HP adds GD to the CBs mechanical GD. But even though the CB requires more excursion and will show higher THD it´ll still sound more precise than the BR alignment. So either there must be a human perception difference between electronically generated GD and mechanically generated GD, or the GD is not the decicive parameter at all. One should notice a remarkable difference in sonic precision with a rather ´boomy´ BR alignment, corrected by a FIR filter. But afaik this doesn´t apply and all reported impovements due to FIR-filtering took place in the higher registers, but not the lowest octaves. This implies that electronic GD could be sonically transparent at LF. But I wouldn´t bet on that and haven´t seen a report or paper about it.
As long as this isn´t proven, one should also take other sources than the GD into acount as responsible factors.
jauu
Calvin
The editor of a german DIY magazine for example claimed the RMS-value to be the decicive parameter regarding bass precision. The discussion about this is still going on, even though it should be quite clear that the RMS-value is a ´artificial´ parameter, itself depending on several other factors like Membrane-area. As such it could only be a rough indicator but no real or precise factor.
I'm not sure I understand you but have you actually tried this out?
My experience is that on such a test with, with F3 between eg 40Hz & 80Hz which covers most HiFi speakers, the BR will sound more precise, musical, faster, bla bla
here's a capture of a listening point measurement.
3 ways 120 Hz-2400 Hz crossover.
no phase correction is applied in this measure.
two curves:
-1: bass-mid with normal polarity (blue curve)
-2: bass driver reverse polarity (red curve -180° phase shift).
room boundary gives more group delay than electrical xover and acoustical load.
i can't hear a difference (exept the crossover region magnitude that shows a littel dip around 120 Hz).
3 ways 120 Hz-2400 Hz crossover.
no phase correction is applied in this measure.
two curves:
-1: bass-mid with normal polarity (blue curve)
-2: bass driver reverse polarity (red curve -180° phase shift).
room boundary gives more group delay than electrical xover and acoustical load.
i can't hear a difference (exept the crossover region magnitude that shows a littel dip around 120 Hz).
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
I also think people often qualify subbas as less precise or boomy. Because they never had sub-bas. Just to at to your conclusion bass with out sub-bas sound more precise.
Lack of reverse null formation at 120Hz suggests poor crossover/speaker setup, and/or use of smoothing in display that is hiding real performance.
I've measured Sennheiser HD650, and get very similar results as headphone.com. Various Etymotic measure similarly at headphone.com:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Droop in squarewave response is entirely phase/GD issue.
This theme is rampant in audio electronics with coupling capacitors forming low frequency high pass filters.
Here is loopback response of E-MU 0404 USB soundcard with square 50Hz square waves:
This is just two amplifiers. Oscilloscope placed between output and input shows intermediate droop. I use DSP to effectively address this.
Pano's first all pass filter adds group delay shelving function that does little more phase damage to that already existent in recording/playback path. In case of "Josie", it's got electric guitars and bass, playing through musical instrument amplifiers, so is already plagued with this type of GD.
I expect that this is why audibility of phase/GD changes are very hard to hear, supporting observations that audible impact occurs with certain music/test signals using a particular playback system.
Pano's second phase 'kink' puts an inflection point in GD with significant second derivative that is highly unnatural.
Blauert's Spatial Hearing states that below 1,6 kHz our hearing evaluates the envelopes. Of course the question is whether only differences between the envelopes of the left and right ear are audible, or if a change in the envelope which is the same for both ears is also audible.
I think that since group delay is audible at 500 Hz, it will therefore also be audible at lower frequencies. Our hearing still works the same at those frequencies. The threshold is probably higher (at least in an absolute sense).
Because our hearing evaluates the envelopes, it would not surprise me if the audibility threshold would be better expressed in cycles. The table by Blauert and Laws does not show sufficient data, but at 500 Hz the threshold is about 1.5 cycles. Perhaps that's the best bet we have for low frequencies.
Nevertheless, we have no hard data so all this is just guessing. Of course, if we make group delay at low frequencies high enough it will be audible, but the big question is how high the threshold is.
I think that since group delay is audible at 500 Hz, it will therefore also be audible at lower frequencies. Our hearing still works the same at those frequencies. The threshold is probably higher (at least in an absolute sense).
Because our hearing evaluates the envelopes, it would not surprise me if the audibility threshold would be better expressed in cycles. The table by Blauert and Laws does not show sufficient data, but at 500 Hz the threshold is about 1.5 cycles. Perhaps that's the best bet we have for low frequencies.
Nevertheless, we have no hard data so all this is just guessing. Of course, if we make group delay at low frequencies high enough it will be audible, but the big question is how high the threshold is.
it was a capture with 24dB/oct LR active crossover at 120 Hz
you're right,raw response shows a nice narrow dip.
But can you hear it? Few seem interested to find out. (Markus, neither can I hear it)
There are all sorts of excuses (not just in this thread) but very little actual testing.
I can do just about any GD/phase on any sort of recording - content below 100Hz would be needed to answer the original question.
It's fun to discuss ideas and concepts, but sometimes the rubber has to hit the road.
There are all sorts of excuses (not just in this thread) but very little actual testing.
I can do just about any GD/phase on any sort of recording - content below 100Hz would be needed to answer the original question.
It's fun to discuss ideas and concepts, but sometimes the rubber has to hit the road.
when bypassing convolver (who correct phase in the lows) during listening
there something a little different,i don't know what.
no better,no worse.spl measurement are the same.
i've done blind test (by masking the mute function ) and enabled/disabled that i didn't know the state of plugin.
it needs to adjust the level exactly the same (with or without convolution ) otherwise,i feel a difference due to spl.
i can not conclude something,but more tests are comming.
there something a little different,i don't know what.
no better,no worse.spl measurement are the same.
i've done blind test (by masking the mute function ) and enabled/disabled that i didn't know the state of plugin.
it needs to adjust the level exactly the same (with or without convolution ) otherwise,i feel a difference due to spl.
i can not conclude something,but more tests are comming.
Last edited:
Barleywater;3305549 [img said:http://graphs.headphone.com/graphCompare.php?graphType=0&graphID[]=856[/img]
Droop in squarewave response is entirely phase/GD issue.
This theme is rampant in audio electronics with coupling capacitors forming low frequency high pass filters.
Here is loopback response of E-MU 0404 USB soundcard with square 50Hz square waves:
View attachment 320629
This is just two amplifiers. Oscilloscope placed between output and input shows intermediate droop. I use DSP to effectively address this.
Nice efforts, How it was ensured that the mic used in measurements
had least distortions for square wave and there was no error. I might be wrong but just my doubt.
Guys ,Just found a link pls read this, Do read it from first page. I am putting link for 4 th page which is conclusion page.
group_delay_4
group_delay_4
As somebody stated, group delay is a graph of the rate of change of phase. I have to wonder if it's the rate of change we are sensitive to, or just the absolute amount of delay (phase shift) at a given frequency? Can you vary GD without varying delay? And then there's the other question about differential phase (or GD) between the channels causing a change in perceived image location. I think the acoustic reactance of virtually any listening room would cause enough reflections (comb filter effects) and even some resonance (psycho-acoustic effect of time blur) at certain frequencies, that GD effects are probably pretty buried.
Having said all that, the first of the three GD examples that Pano put up may have sounded cleaner to me. I could be imagining that. The change was very small. (It was Pano wasn't it?) The latter two examples may have sounded like they had an acoustic environment effect attached to them. But my computer speakers roll off below 100HZ so...
Having said all that, the first of the three GD examples that Pano put up may have sounded cleaner to me. I could be imagining that. The change was very small. (It was Pano wasn't it?) The latter two examples may have sounded like they had an acoustic environment effect attached to them. But my computer speakers roll off below 100HZ so...
That's basically what I've found, too, with my tests. A little different with corrected phase, but not better or worse. It's an interesting idea, tho, so I will also continue to test.
Thanks for listening to the samples, Bob. One of the files has a phase rotation that is 180 deg. at 110 hz. The other two are untouched. I doubt you'd hear it if your computer speakers roll off under 100Hz. Maybe, as there is pretty good phase shift above 100Hz.
I have bin listening to the Pano mp3 files now on the speakers and not by head phone.
I must say I have not much difference between the sound samples. When I have to point the file witch shout be original I would say version L-mp3. I had that experience also on the headphone at first take. Version L seems the music more together.
But when comparing for the second and third time and so on the difference becomes harder to hear.
Am I right Pano is L not modified.
I must say I have not much difference between the sound samples. When I have to point the file witch shout be original I would say version L-mp3. I had that experience also on the headphone at first take. Version L seems the music more together.
But when comparing for the second and third time and so on the difference becomes harder to hear.
Am I right Pano
is L not modified.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.