camplo> I think fluid's plot is showing how much GD is necessary to be subjectively different, the frequency band shown is probably arbitrary and not relevant. They could probably have engineered a speaker where the GD differences manifest at a higher frequency and shown similar results. Insofar as the specific frequency band having any relevance i think Earl has mentioned before that the ear becomes more sensitive to time characteristics above 1khz, which would contradict your wanting to move up in crossover frequency based on fluid's plots. I could be wrong about the plots but I think this is the correct interpretation.
Augerpro, the idea was to move the xo higher above fb to avoid the negative traits near fb....
This is not exactly a discussion of where in the entire spectrum to place a XO
The first chosen xo I had for this project was 150hz iirc lol. So, "moving up" is relative. The lowest technically correct xo I can muster for a tweeter compression driver is about 300hz. The Axi is specd to 300hz but obviously needs a horn to support that....to keep the xo "higher" than fb...we need to use a horn with a lower fb.
The graphs Fluid so timely posted show a trend of gd vs frequency at said impact per graph...they resemble other graphs I've seen, meant to display the threshold of audible group delay vs frequency as seen by creator of the data
This is not exactly a discussion of where in the entire spectrum to place a XO
The first chosen xo I had for this project was 150hz iirc lol. So, "moving up" is relative. The lowest technically correct xo I can muster for a tweeter compression driver is about 300hz. The Axi is specd to 300hz but obviously needs a horn to support that....to keep the xo "higher" than fb...we need to use a horn with a lower fb.
The graphs Fluid so timely posted show a trend of gd vs frequency at said impact per graph...they resemble other graphs I've seen, meant to display the threshold of audible group delay vs frequency as seen by creator of the data
It is posts like this that caused GM to ask if English was your native language
I don't really know what you getting at here. The group delay will increase as the speaker/horn/driver rolls off. If you equalize it or cross it over to another driver that increase can be avoided. Simulating the whole system in Vituix or similar will give a better idea of the systems response as that is what matters.
If the group delay comes from the crossover pushing it higher will make things worse.
The impulses used in the study were both real and simulated and were either from real speakers or designed to resemble possible speakers.
The frequency band is not arbitrary and is a very important part of the research it gives an idea of what amount of group delay at what frequency resulted in no audible difference in the tests.
If the group delay between 300Hz and 1K and above can be kept below 1ms then there should be no audible problem resulting from it.
Between 1 and 2ms in that region then it might be audible, but if it gets above 2ms then it is almost always audible in reversed testing.
Below 300Hz the group delay can get quite high quickly without much audible penalty based on this study.
Indeed, this shouldn't be necessary and if you check my posts you'll see that I have made similar comments, several times in the past, as you do now, both in this thread and elsewhere.
It's just a reflection on certain structural behavior in light of general decency standards, independent of the substantive contributions of both gentlemen.
It's about the disdainful attitude towards (respected) forum members, scientists and others who hold different points of view, which is fuelled by an inherently misguided belief of one's own superiority.
You don't have to go far back in this thread to find some examples.
I'm not sure what you mean by testing other bands. The full frequency response of the impulses was tested and the amount of group delay in those impulses vs frequency plotted and compared to whether the forward or backward versions sounded the same or not.
Some impulses were convolved with another filter to make them "Pink" and the differences in the pink impulses were easier to hear.
Some impulses were convolved with another filter to make them "Pink" and the differences in the pink impulses were easier to hear.
I see. I thought they were just adding GD to a narrow-ish band around 400hz and seeing how much it took to be audible. If that is (essentially) the case, the point I was trying to make was they could just as well have picked 1.5khz (or whatever) and there was nothing special about the 400hz they used. My impression was that camplo was pushed to move the crossover higher based on those plots and their arbitrary selection of the 400hz area to use as in the test.
Actually, that makes sense to me. My Radian 951 tweeter starts with a low shelf, 16.5Khz, Q=.4, -12.8dB. I'll try modeling a crossover with your technique.
If you are able to make a clear overview of all subsystems, eg of a woofer, which is universally valid (ie indisputable), then you can get started at Klippel without applying for a job, because they can't
To my knowledge, no one has succeeded so far.
Wrt some subsystems there are at least 5 or 6 (slightly) different schools of thought/approaches.
Last edited:
Repeat
You are repeating yourself.
Your answer is here:
https://www.diyaudio.com/forums/mul...m-spl-low-distortion-2-a-404.html#post6794226
"The group delay will increase as the speaker/horn/driver rolls off. If you equalize it or cross it over to another driver that increase can be avoided.".
Are you crossing your horn to a lower frequency driver? Use the appropriate crossover and slope. Problem solved.
You are repeating yourself.
Your answer is here:
https://www.diyaudio.com/forums/mul...m-spl-low-distortion-2-a-404.html#post6794226
"The group delay will increase as the speaker/horn/driver rolls off. If you equalize it or cross it over to another driver that increase can be avoided.".
Are you crossing your horn to a lower frequency driver? Use the appropriate crossover and slope. Problem solved.
Last edited:
Mostly it is not a measurement of an impulse response. Most systems (e.g. REW etc) perform a slow (1-10 sec) FR sweep and out from that, calculate impulse response and phase response.
Very few, if any, have a (1) discrete impulse as the measurement stimuli. I would like to see a true impulse stimuli based system as I think the jolt of an impulse is a quite different task than a slowly gliding frequency change in what is almost steady state. I suppose that the DUT need to be a minimum phase system in order for the Fourier calculation to be correct - right? If so, an other reson for why a true impulse measurement would be favourable - it would show/incorporate such deficiency.
Music as we perceive it is mostly impulses rather than "tones". Impulses are hard, steady state tones are easy
//
I'm simply not the only one being fed up with your spreading of nonsense. It's a purely factual matter.
I chose to use the same wording as the originator.
"Simultaneous measurement of impulse response and distortion with a swept-sine technique"
https://www.melaudia.net/zdoc/sweepSine.PDF
I did not intend to imply that the stimulus was a real impulse if that is how you read it.
I think he means "simultaneous calculation" of impulse ....
Its indisputable - no impulse has been offered to the DUT input terminals.
Its just not super stringently expressed I guess...
I know, its hard with real impulses - very little energy, background noise etc etc but just because its hard it should not be avoided or impersonated for a sweep ...
I plan to do the test one day - to see if the sweep impulse and a real impulse will look the same... and EQ after the real impulse and see how that sounds ;-D
//
Its indisputable - no impulse has been offered to the DUT input terminals.
Its just not super stringently expressed I guess...
I know, its hard with real impulses - very little energy, background noise etc etc but just because its hard it should not be avoided or impersonated for a sweep ...
I plan to do the test one day - to see if the sweep impulse and a real impulse will look the same... and EQ after the real impulse and see how that sounds ;-D
//
Last edited: