A test at home--- Place an all pass filter set at say 7-9kHz, well above the tweeter crossover point, in series with your tweeter with a bypass switch. The coils and caps will be quite small so level should not be an issue and cost minimal. If phase is not important this filter will have little to no effect. Testing first hand gives the most experience.
With all the discussion in the literature about phase one could infer phase is important. Absolute phase may not be. Drivers flipping phase in the pass band is. The test above might be informative.
With all the discussion in the literature about phase one could infer phase is important. Absolute phase may not be. Drivers flipping phase in the pass band is. The test above might be informative.
There are always going to be some people in the audio world that sell products based on subjective experience and have went to great lenght to try and downplay ABX testing.
I say they have an agenda to sell (that is a smart reason too in the business world 😉 ). Colorful words, pretty products and branding has much more impact then removing all extra stimulus and just listening 😉
FWIW, your ghost analogy is funny because I think many audiophiles "Chase Ghosts" daily. My wife always says our 1 year old "doesnt want to give up the ghost" when trying to sleep.......Im not saying audiophiles 1 year olds..its more about "giving up the ghost". 😉
When do you see this?? The greatest phase changes occur when the driver is clearly out of it's useful range. You have a driver flipping phase where it has "flat" frequency response??
Rob🙂
For the all-pass test, why not just put it at line level, then it doesn't matter where the crossover frequency is?
This circuit is 1-pole allpass, as drawn here it gives a 45degree lag by 1.4kHz when the switch is closed. By 20kHz, the response is essentially inverted. You can scale the 10k resistor or 4.7nf cap to change frequency as desired (frequency drops proportional to change in either -- double to 20K resistance cuts frequency in half).
Provides no phase shift when the switch is open. Makes an easy in-out test, just flip the switch open/closed. Use any decent opamp and split supplies.
This circuit is 1-pole allpass, as drawn here it gives a 45degree lag by 1.4kHz when the switch is closed. By 20kHz, the response is essentially inverted. You can scale the 10k resistor or 4.7nf cap to change frequency as desired (frequency drops proportional to change in either -- double to 20K resistance cuts frequency in half).
Provides no phase shift when the switch is open. Makes an easy in-out test, just flip the switch open/closed. Use any decent opamp and split supplies.
Attachments
Hi sumaudioguy.
The all pass filter will affect the phase about a decade or so below
the 7-9kHz point. So it would also change the FR of the loudspeaker
at the crossover point. IMO that is what you would be hearing, not the phase shift. The correct way of doing this would be with the all pass filter before the crossover,or as bwaslo has suggested before the amplifier.
About drivers flipping phase in the passband, I do not think this is even possible.
Regards Tubbe.
The all pass filter will affect the phase about a decade or so below
the 7-9kHz point. So it would also change the FR of the loudspeaker
at the crossover point. IMO that is what you would be hearing, not the phase shift. The correct way of doing this would be with the all pass filter before the crossover,or as bwaslo has suggested before the amplifier.
About drivers flipping phase in the passband, I do not think this is even possible.
Regards Tubbe.
all pass and testing
You are correct of course. The trouble is the useful range is often much less than wished for. See here for a horn tested. http://www.diyaudio.com/forums/multi-way/161627-horn-honk-wanted-68.html#post2160891 This is why it is so critical to know the excess time very accurately so phase and frequency response may be measured. Many drivers flip phase several times while still having "flat" frequency response or follow a constant directivity frequency response which is not flat.
Yes I like the all pass ahead of the amp much better. Please do that if possible or place the all pass between the amp and the speaker for an in between choice. One decade below the all pass frequency there is less than 6 degrees of phase shift. No matter.
If it is tested with a stereo pair it may be desirable to do both channels at once. But if phase does not matter then does it need to be both channels?
See about all pass here: All-pass filter - Wikipedia, the free encyclopedia Look at "Lattice Filter" for passive version.
All you guys who jumped in on this "in house" test I applaud you. Someone will certainly learn something. Thanks! Testing "in house" teaches us much.
=SUM
You are correct of course. The trouble is the useful range is often much less than wished for. See here for a horn tested. http://www.diyaudio.com/forums/multi-way/161627-horn-honk-wanted-68.html#post2160891 This is why it is so critical to know the excess time very accurately so phase and frequency response may be measured. Many drivers flip phase several times while still having "flat" frequency response or follow a constant directivity frequency response which is not flat.
Yes I like the all pass ahead of the amp much better. Please do that if possible or place the all pass between the amp and the speaker for an in between choice. One decade below the all pass frequency there is less than 6 degrees of phase shift. No matter.
If it is tested with a stereo pair it may be desirable to do both channels at once. But if phase does not matter then does it need to be both channels?
See about all pass here: All-pass filter - Wikipedia, the free encyclopedia Look at "Lattice Filter" for passive version.
All you guys who jumped in on this "in house" test I applaud you. Someone will certainly learn something. Thanks! Testing "in house" teaches us much.
=SUM
SUM, I appreciate the idea, but they did far more rigorous and controlled testing here: Chapter 5
than I can do at home. I do agree that doing your own testing is a good idea if you have doubts however so long as its limitations are well understood. I'm quite sure most of us DIY guys do not have this level of understanding on the subject(s) to do better tests than what's available to read. In the link there's a lot more testing than I or any of us could easily do. If you change just one side then it might be audible, but who's going to build a different right and left speaker? So it's not a very useful test.
Anyone care to explain the impulse response I posted? Nice that the polar response is included in the pdf. It seems a more pressing issue to me d/t the available evidence about what's important to measure. Can anyone post the link on the audibility of resonances? The only thing I understand to do with an impulse response is to find the first reflection and gate it out while your doing loudspeaker measurements so the room doesn't contaminate your results which introduces its own limitations.
Here's a nice primer on the various measuring approaches of your loudspeaker's FR:
mh-audio.nl - Acoustic
Scroll to the bottom of the page.
Dan
than I can do at home. I do agree that doing your own testing is a good idea if you have doubts however so long as its limitations are well understood. I'm quite sure most of us DIY guys do not have this level of understanding on the subject(s) to do better tests than what's available to read. In the link there's a lot more testing than I or any of us could easily do. If you change just one side then it might be audible, but who's going to build a different right and left speaker? So it's not a very useful test.
Anyone care to explain the impulse response I posted? Nice that the polar response is included in the pdf. It seems a more pressing issue to me d/t the available evidence about what's important to measure. Can anyone post the link on the audibility of resonances? The only thing I understand to do with an impulse response is to find the first reflection and gate it out while your doing loudspeaker measurements so the room doesn't contaminate your results which introduces its own limitations.
Here's a nice primer on the various measuring approaches of your loudspeaker's FR:
mh-audio.nl - Acoustic
Scroll to the bottom of the page.
Dan
flip my phase
See the on axis graph around 5kHz-7kHz going down and then back up to a peak where the arrow is? Then the first off axis graph going to a minimum at the same place? That is classic phase flipping of the tweeter. On axis and first off axis are out of phase 180° with each other at 6500Hz. One inch dome ehy?
See the on axis graph around 5kHz-7kHz going down and then back up to a peak where the arrow is? Then the first off axis graph going to a minimum at the same place? That is classic phase flipping of the tweeter. On axis and first off axis are out of phase 180° with each other at 6500Hz. One inch dome ehy?
Attachments
I see that, but how does it relate to the impulse graph? Or does it? Is there something that could be identified on the impulse that would let you know you have an audible issue somewhere? IOW, Is there a way you can look at it and identify a problem?
If you are saying that a phase anomaly that causes a response anomaly is audible in normal listening conditions, you won't get any argument from me. I don't think that anyone is doubting response anomalies as being audible. The size and frequency of the anomaly should be in some way proportional to its audibility. IOW a high Q resonance of high amplitude and power should be more easy to discern than a low Q/low amplitude and power.
Anyway, what I'm asking is "what exactly does the impulse response tell us?" Nothing I've read has resolved the issue in my head. How can it be used to design a better speaker? I bet I'm not the only one who doesn't comprehend it exactly.
Thanks again SUM,
Dan
If you are saying that a phase anomaly that causes a response anomaly is audible in normal listening conditions, you won't get any argument from me. I don't think that anyone is doubting response anomalies as being audible. The size and frequency of the anomaly should be in some way proportional to its audibility. IOW a high Q resonance of high amplitude and power should be more easy to discern than a low Q/low amplitude and power.
Anyway, what I'm asking is "what exactly does the impulse response tell us?" Nothing I've read has resolved the issue in my head. How can it be used to design a better speaker? I bet I'm not the only one who doesn't comprehend it exactly.
Thanks again SUM,
Dan
First Im not an expert but I love learning/reading/discussing speaker building.
Honestly, this phase discussion amounts to a whole lot of subjectivity and not enough meat. There are many, many other issues to deal with way before someone remotely gets into the nitty gritty. If the best speaker designers in the DIY and comsumer world are not worrying about it I doubt simple DIYers like us should be worrying about it 😉
If sumaudioguy produces a speaker that sounds better then all others because of these theories then maybe someone will care but until then its all just someone's theory. We have seen how Geddes has changed the DIY world though process on Constant Directivity so who knows.
As for the impulse...
The impulse is used to generate many other responses. Its at the heart of all speaker measurements.
Think of it has the start of your build. You have to measure the impulse of each driver separately using something like HOLM. Gating simply tells you how low you can go for an accurate measurement. The great thing is that most important XOs are above 700Hz so getting gated measurements is pretty easy in most rooms.
You take those impulse measurements and create FR plots, CSDs etc. You import the FR plots into Crossover sim software (Free one like Speaker workshop works great) and you build your XOs from those measurements.
Actually, the research (according to Toole) says exactly the opposite as regards Q -- the lower-Q resonance, even with a much lower amplitude, is the more audible. High Q resonance turns out to be quite hard to hear. Skeptical about that against-intuition statement, I wrote a quick PC application to do this test myself, processing various music and noise signals. It certainly turned out to be true for my ears.
Which is kind of unfortunate for people (like me) who write software to do the usual CSD waterfall plots -- those are good at showing the kinds of resonances you are less likely to hear (and not so good at the ones that are clearly heard).
As well, ue to the nature of music, there is usually less energy available inside the box to excite a high Q panel resonance.
dave
dave
Basically knowing very little and absolutely nill of substance about the impulse response, I did a whole bunch of trial and error work and was able to get this:
That's the best I've done. I wish I would have saved all the data to get me there. I realize that the impulse is the start of the whole deal, but is it possible for me or anyone to view that with their eyes and make some useful generalization about how that speaker will sound based on them alone. The polar response will definitely tell me what I'll hear--at least it has so far. That's why I'm confident in its importance.
I believe BudP mentioned changes in CSD and it's audibility, but has come back for follow up on that. I would think it should be audible, but to what degree I'm unsure. Anyone know of any tests on that?
I'm sure you know, but any gating also reduces your resolution and you can see its effects in my posted graph.
Thanks Doug,
Dan
Thanks Bill, I knew something wasn't correct when I was writing it.🙂 Now that makes sense.
That second statement is intriguing.
So, loudspeaker polar response is readily audible, low Q resonance is more audible than high Q, CSD doesn't show low q resonance well................ are we getting somewhere?
Still fuzzy on impulse,
Dan
Last edited:
Doug 20, please see post #153 of this thread for the original. Saw no reason to repost the whole thing.
dantheman, why would you ever discourage anyone from doing testing at home?
Can you hear the phase flip, yes. Play female vocal with guitar and piano or violin. Move on and off axis of the tweeter in the direct sound field. It should be pretty easy to locate the change in sound as you move 15 degrees. Covering or blocking one ear can help a lot.
Time, frequency, and energy are the three domains. Ignoring one in favor of another just does not seem right.
I have no idea how anything relates to the impulse graph because the impulse graph is such a mess it is like looking for a mud ball in a mud hole. No insult intended please! Until the impulse is a lot cleaner it means nothing to me.
My speakers are only custom installs as each system is calibrated on site so there is little chance to hear one for the average person. Will post "public" locations if they become available. Yes I build some number of them but not a bunch. The array is constant directivity from 200Hz up.
From a psycoacoustics standpoint when that tweeter flips phase the ear brain only uses the information above that frequency for energy and not for time or formation of image. Frequency balance yes, image no. To hear the effect one would need to listen to a speaker which stays in phase and one that flips phase. At that point it becomes grossly obvious. But that means in a typical case two different tweeters or a test system with an sharp all pass which is not valid either. The phase flip of any driver has so far eluded all my attempts to fix using some pretty sophisticated DSP. Phase can be flip but not like a driver changing radiation from piston to drum head mode where the outside moves forward (tweeter now) and the center moves back. I have not tried a FIR filter yet and that may be the answer because of the zero phase shift in the passband is possible so two adjacent passbands could have opposite phase and fix the flip.
When the signal is in the region of the phase flip directivity and coherence are non-existent as a tremendous amount of lobes appear on a direct radiator like a dome tweeter. This qualifies as FUBAR at that frequency.
Here is another possible way to have fun... Put a sine wave at 3000Hz through the tweeters and then stand back with your eyes closed and point at the apparent sound source. Next open your eyes and check your aim. Do this at several frequencies including where the phase problem is. Now I know that is pretty high for directional information so use both speakers because the sine wave sound should always come from exactly between the two speakers at every frequency with a mono signal. If the sound does not come from exactly between the speakers there is a big problem at that frequency. Another easy test to do at home. Computer based oscillators are free at Merchand. Only use one channel of the sound card with a "Y" cable to not rely on the L&R of the sound card.
Yes that is right, low Q means affecting a wide band. High Q means affecting a very narrow band.
Actually I encouraged it so long as its limitations are known. Reread my post if you don't believe me. There is much more complete information available at this site: Table of Contents than would be reasonable to do at home. I'd encourage doing any testing anyone can do at home. Test to the maximum of your ability.
Thanks for the comment on the impulse. I appreciate the effort.
Dan
- Status
- Not open for further replies.