I'm pretty confident that the pre-ringing shown is not a part of the true impulse response of the system. It looks like a measuring artifact, possibly a sine(x)/x ringing of the excitation impulse? Could be an artifact of linear phase truncation within the measuring/calculating process?
The impulse set to its first non-zero peak shortly after t = 0, as shown, would be the correct setting for non causal, lowest delay measurement (ignore the ringing).
David
The pre-ringing is a result of trying to represent a signal where there is still energy up to the Nyquist point (i.e., fs/2, half the sampling rate). Baseband sampling theory assumes that the spectrum has dropped to zero before you get to fs/2. The speaker isn't without response there though, and neither is the microphone, pretty much no matter what fs is used.
All that can be represented is the energy below fs/2, so the measured result is what you'd get as if your system was also being fed through a zero-delay brickwall filter. The brick wall filter with zero delay is not causal and has pre-ring.
If you first used a "real" brickwall filter (or the best you can do in analog, an approximation) on the signal to avoid the fs/2 issue, that filter will be causal and have delay enough so its "pre-ring" (from the brickwall filter, not from sampling) would actually not occur till time t=0 or later.
Or if your signal energy has dropped gradually to near zero long before fs/2, then the ringing won't likely be enough to be visible and the gradual rolloff (low-pass) imparts delay to push all the impulse response activity into positive time.
In other words, the ringing isn't showing you something that isn't there. It's just not showing you some stuff that is actually there (i.e., all the frequency components above fs/2) which if included would make the impulse response appear causal again.
All that can be represented is the energy below fs/2, so the measured result is what you'd get as if your system was also being fed through a zero-delay brickwall filter. The brick wall filter with zero delay is not causal and has pre-ring.
If you first used a "real" brickwall filter (or the best you can do in analog, an approximation) on the signal to avoid the fs/2 issue, that filter will be causal and have delay enough so its "pre-ring" (from the brickwall filter, not from sampling) would actually not occur till time t=0 or later.
Or if your signal energy has dropped gradually to near zero long before fs/2, then the ringing won't likely be enough to be visible and the gradual rolloff (low-pass) imparts delay to push all the impulse response activity into positive time.
In other words, the ringing isn't showing you something that isn't there. It's just not showing you some stuff that is actually there (i.e., all the frequency components above fs/2) which if included would make the impulse response appear causal again.
Gibb's phenomenon is where they show the construction of a squarewave from any finite number of harmonics. If you synthesize a squarewave by adding up harmonics to the 7th the 9th the 11th, to any quantity actually, there will always be 9% overshoot. It always rings at the up/down transitions. A true square wave filtered would be different and may have no overshoot, but only because the filtered harmonics roll off rather than simply ceasing above a certain number.
I believe the linear phase brickwall impulse, with sin(x)/x ringing, is the same effect.
(Thanks for not being afraid to ask.)
David
Thanks for the excellent explanation David. Again, it's good to have you here. I searched the web for Gibb and acoustics prior to asking, but didn't find what I was looking for.
Dan
Dan
Hah, you probably came up with the guy from the Pirates movies or the Whitehouse press sec. like I did when he posted it 😀
Im glad you asked, I didnt know the reference either.
Im glad you asked, I didnt know the reference either.
I kept getting what I think is the "BeeGee's"(sp?). The guy who sings disco with the high voice. ex: Staying Alive
Dan
Dan
Sorry, no I don't have any information on how it works, nor do I have one. But "sumaudioguy's posts reminded me of a person I met, who was involved with Urei, and possibly Bag End.
Iain.
Last edited:
I wanted to come back to a statement made 100+ pages ago. I thought it was intriguing, then moved on:
Well I have been working on the crossover of my Altec A5 rig (2-way) and just could not get the tonal balance right. Either too bright, too dark, too thin, too heavy, too something - never "just right". Some recordings were OK, others not at all.
Then the other day I had settled on a crossover topology I liked and just on a whim set the pink noise SPL of the 2 halves to the same value. Bingo! Tonal balance fixed. And what is the acoustic crossover point? ~635Hz. Works like a charm. Works so well, in fact that I as doubtful. So I ran some tests in software with pink noise filtered to my system response. Yes - within a dB or less, there is equal energy above and below 635Hz. How about that?
Does this work just because my crossover is at 635Hz - or will this apply to many other systems, too? Further investigation is needed.
Well I have been working on the crossover of my Altec A5 rig (2-way) and just could not get the tonal balance right. Either too bright, too dark, too thin, too heavy, too something - never "just right". Some recordings were OK, others not at all.
Then the other day I had settled on a crossover topology I liked and just on a whim set the pink noise SPL of the 2 halves to the same value. Bingo! Tonal balance fixed. And what is the acoustic crossover point? ~635Hz. Works like a charm. Works so well, in fact that I as doubtful. So I ran some tests in software with pink noise filtered to my system response. Yes - within a dB or less, there is equal energy above and below 635Hz. How about that?
Does this work just because my crossover is at 635Hz - or will this apply to many other systems, too? Further investigation is needed.
635 is about the center of the audible range. sqrt(20*20000) = 632. I think it was Thorsten who I first heard advocating a balance between the high-end and low-end extension -- if the woofer only goes down to 40, the tweeter should only go up to 10K, etc. Just another way of saying the same thing. I think he left out the square root and just did LF*HF = 400000.
Last edited:
Yes, I remember the rule of 400,000. My rig may be close to it. The equal energy thing works well for me so far.
Most folks are not going to want to give up that top end, tho.
Most folks are not going to want to give up that top end, tho.
Funny, it seems to be important, at least for the systems I can speak for and until the present day when a handful of subs go below 35 Hz. Sound is unsatisfactory unless balanced to the rule more or less (or at least, it is bad to add much treble OR bass to a good system unless you add both and better neither than just one).
But here's a question: can anybody think of any rational for that rule? It really seems arbitrary... no I don't mean just the particular numbers (they are obviously arbitrary even with the magic 20-20k - which are themselves quite arbitrary - in the formula). Why should a system seem deficient just because now and then the treble doesn't go to a certain height and now and then (very rarely) the bass seems short of a certain depth?
Seems very logical till you think about it, eh.
Last edited:
I've thought about it and agree. But I don't think it's too strict of a rule. More like a guideline. Gotta say, tho - I think a lot of speakers have way too much treble - like the very top end. I just don't hear the stuff in nature.
So really, I don't know what to think!
BTW, I've been wanted to start a thread on tonal balance, now might be the time.
So really, I don't know what to think!
BTW, I've been wanted to start a thread on tonal balance, now might be the time.
re pitch centre thingy
More bandwidth please, not less.
Regarding Ben's question: Why should a system seem deficient just because now and then the treble doesn't go to a certain height and now and then (very rarely) the bass seems short of a certain depth?
At the LF end, a lack of extension does something to our perception. The principle of the missing fundamental comes in to play. At the HF end, group delay/phase shift, from bandwidth limited devices, seems to alter our perception of instrument timbre and decay. An interesting demo from "speaker Dave" many years ago at Kef, seemed to show, that extended low phase shift sub, perceptually increased the clarity of the low mid range. Either it was the power of suggestion, or else they put something in the tea!
Flak jacket at the ready.
Iain.
More bandwidth please, not less.
Regarding Ben's question: Why should a system seem deficient just because now and then the treble doesn't go to a certain height and now and then (very rarely) the bass seems short of a certain depth?
At the LF end, a lack of extension does something to our perception. The principle of the missing fundamental comes in to play. At the HF end, group delay/phase shift, from bandwidth limited devices, seems to alter our perception of instrument timbre and decay. An interesting demo from "speaker Dave" many years ago at Kef, seemed to show, that extended low phase shift sub, perceptually increased the clarity of the low mid range. Either it was the power of suggestion, or else they put something in the tea!
Flak jacket at the ready.
Iain.
- Status
- Not open for further replies.