Thanks for the advice's Gabdx
Would you mind to elaborate on the quoted paragraphs ?
Do you mean that you listened to a selection of tones (sine-wave?) in free air for each driver that was part of your project before measuring them to identify peaks & dips? Please describe how was your setup to listen to it.
I can identify when there are irritating tones in music but it's more difficult to notice dips, I am not sure my ears are trained enough to do that with a tone generator, never tried that in fact.
My pleasure,
ARTA has the bar graphs which clearly shows a problem around a group of frequences.
You can reach the same conclusions by ear using Online Tone Generator - generate pure tones of any frequency, I like to browse frequency and listen for resonances, and harmonics. (Headphones are terrible, you can hear many undertones with headphones browsing tones.)
Anything that you find irritating in music when you hear a key or sound, you find the dominant pitch, then listen to the tone of the same pitch with the generator, and confirm the anomaly by ear.
If you hear it and you can associate a frequency number or group to it, proceed to next step.
Now look at the measured response, to see if it is visible, is it also a room mode? is it floor bounce? or some odd reflections? or a real response curve problem that needs to be corrected.
If you can see a problem, correct the response or redo the Xo until this anomaly is removed from the response. Then you do A/B comparison. To be sure you are going in the right direction.
--------------------The reason for this way of correcting the response is that there are too many visible problems with the response of a speaker in a room. It is too overwhelming to address all of them and you will never be happy with the result if you do. for example, if you flatten the sound too much you create a dead sound with more problems, XO resonances of components, loss of dynamics, veiled sound etc.
So, this is the process, after that you can send the new desired response curve to the simulator.
Just inputing a straight line in the simulator or a slopped with bass boost wont give you a good result.
The simulator needs very accurate curves of impedance/response with calibration mic of each driver IN the final enclosure exactly like the finished product!!!!
Any pre-simulations, or modification to the enclosure null any simulation.
ARTA has the bar graphs which clearly shows a problem around a group of frequences.
You can reach the same conclusions by ear using Online Tone Generator - generate pure tones of any frequency, I like to browse frequency and listen for resonances, and harmonics. (Headphones are terrible, you can hear many undertones with headphones browsing tones.)
Anything that you find irritating in music when you hear a key or sound, you find the dominant pitch, then listen to the tone of the same pitch with the generator, and confirm the anomaly by ear.
If you hear it and you can associate a frequency number or group to it, proceed to next step.
Now look at the measured response, to see if it is visible, is it also a room mode? is it floor bounce? or some odd reflections? or a real response curve problem that needs to be corrected.
If you can see a problem, correct the response or redo the Xo until this anomaly is removed from the response. Then you do A/B comparison. To be sure you are going in the right direction.
--------------------The reason for this way of correcting the response is that there are too many visible problems with the response of a speaker in a room. It is too overwhelming to address all of them and you will never be happy with the result if you do. for example, if you flatten the sound too much you create a dead sound with more problems, XO resonances of components, loss of dynamics, veiled sound etc.
So, this is the process, after that you can send the new desired response curve to the simulator.
Just inputing a straight line in the simulator or a slopped with bass boost wont give you a good result.
The simulator needs very accurate curves of impedance/response with calibration mic of each driver IN the final enclosure exactly like the finished product!!!!
Any pre-simulations, or modification to the enclosure null any simulation.
Last edited:
This were the responses that I obtained at around 2 m if I recall.... after the design was completed, the tweeter is 0.5 db too high in the graph, just a 1 ohm higher resistor to fix.
This is with 6 db smoothing! a lot, otherwise it is a sawtooth graph.
Green = my 2 way speaker design
Red = professional speaker
Blue = professional speaker + Rel Sub T/9i - 10" Home Subwoofer | REL Acoustics
It is easy to see the room resonances : 13, 26, 49.
I was able to have a sub-boosted bass in the 60 - 120 region with no bumps after, except the 600-700 not so bad, too much to pay in terms of speed to tame it more. my 'ear' problem was in the 650 - 900, you can see it almost matches the professional speakers there, I cannot ask for better,
from 900 to 2khz you can see the very accurate but 1 db maybe lower response, maybe I didn't properly matched the volume there by only adjusting at 1khz, the green line should be a little higher I suppose. So yes, a 900 to 2khz that equals what I consider the ultimate reference speakers (I own them). From 2khz to 4khz this is the ability to hear language and this is to me an improvement at the cost of a little bit more fatigue and revealing of high distortion music, it is around 1db lower than in graph because of tweeter is not measured with final resistor... and the rest is very good.
Not hard to see the advantages of DIY over even top of the line commercial products.
This is with 6 db smoothing! a lot, otherwise it is a sawtooth graph.
Green = my 2 way speaker design
Red = professional speaker
Blue = professional speaker + Rel Sub T/9i - 10" Home Subwoofer | REL Acoustics
It is easy to see the room resonances : 13, 26, 49.
I was able to have a sub-boosted bass in the 60 - 120 region with no bumps after, except the 600-700 not so bad, too much to pay in terms of speed to tame it more. my 'ear' problem was in the 650 - 900, you can see it almost matches the professional speakers there, I cannot ask for better,
from 900 to 2khz you can see the very accurate but 1 db maybe lower response, maybe I didn't properly matched the volume there by only adjusting at 1khz, the green line should be a little higher I suppose. So yes, a 900 to 2khz that equals what I consider the ultimate reference speakers (I own them). From 2khz to 4khz this is the ability to hear language and this is to me an improvement at the cost of a little bit more fatigue and revealing of high distortion music, it is around 1db lower than in graph because of tweeter is not measured with final resistor... and the rest is very good.
Not hard to see the advantages of DIY over even top of the line commercial products.
Attachments
Bummer.
-well, treat it as an opportunity!
When moving-onto mock-up for testing the transition between the Fostex and Gradient, try it without any tweeter.
If you need more dispersion at higher freq.s then you can start looking for an option, though there aren't that many at lower prices that are planar dipoles.
Hi Gabx,
Green is indeed more consistent than the two others, nice job with that!
I have played a bit with the online frequency generator, and it's close to unbearable for all higher frequencies, especially the 2.5k to 3.5k, around 15k it's barely discernible, and I don't get the top end even with headphones
And worst it's fast getting to my nerves, not my cup of tea, I don't know how you do it.
..err, hopefully not ears as defunct as those drivers!
Haha I hope not!
I put the LS Plan back together but to no avail, I think there is a dent in the diaphragm that is causing that issue with the impedance, with a heavy hearth I scraped it because even if the that diaphragm was not the issue, the magnets where in bad shape. And worst I had a look at the other one and I could see that the aluminum was going away from the magnet too.
So in the long run they would have failed anyway
Btw in case I am not satisfied with only the Fostex, what do you think of this driver as a replacement or in general ? :
Gradient AT1-R AIR Motion Transformer buy at hifisound.de
It's a 4 ohm unfortunately, but it's a potential dipole
O... well I didn't mentioned that it works to pin point problem within a range...
If you sweep from 20 to 20k the higher fq will be too loud and can be unhealthy. I adjust the volume to be comfortable within a range, like 1300 to 1400, then I try to hear the suspected problem that I suspect is in 1350....
There is a sweep tone software somewhere which follows the natural hearing curve of music, it's not that great.
To resume, use the generator at comfortable listening volume to confirm problems within a restricted range, like you cant compare 200 hz and see if there is a rise at 2 khz, it will not work...
But in a finished XO you should hear a very smooth transition with nothing sharp/irritating/vibrating/quiet passages.
If you sweep-tone and the perceived volume decreases then back to normal at some points of the tests, then you have a major dip problem
If you sweep from 20 to 20k the higher fq will be too loud and can be unhealthy. I adjust the volume to be comfortable within a range, like 1300 to 1400, then I try to hear the suspected problem that I suspect is in 1350....
There is a sweep tone software somewhere which follows the natural hearing curve of music, it's not that great.
To resume, use the generator at comfortable listening volume to confirm problems within a restricted range, like you cant compare 200 hz and see if there is a rise at 2 khz, it will not work...
But in a finished XO you should hear a very smooth transition with nothing sharp/irritating/vibrating/quiet passages.
If you sweep-tone and the perceived volume decreases then back to normal at some points of the tests, then you have a major dip problem
I wish you did, that would have spared me that short torture moment
If you sweep from 20 to 20k the higher fq will be too loud and can be unhealthy. I adjust the volume to be comfortable within a range, like 1300 to 1400, then I try to hear the suspected problem that I suspect is in 1350....
There is a sweep tone software somewhere which follows the natural hearing curve of music, it's not that great.
To resume, use the generator at comfortable listening volume to confirm problems within a restricted range, like you cant compare 200 hz and see if there is a rise at 2 khz, it will not work...
But in a finished XO you should hear a very smooth transition with nothing sharp/irritating/vibrating/quiet passages.
If you sweep-tone and the perceived volume decreases then back to normal at some points of the tests, then you have a major dip problem
I see what you mean and I guess not everyone is sensible too the same frequencies, but to the ears using a tone I find it's still difficult to evaluate what is or isn't acceptable, but that may be just me.
You are right on that, we don't hear the same things. Some of us are sensitive to distortion, some prefer a raise treble, that is in great part due to different musical tastes and recordings which are mastered with opposite philosophies. That is why when you design your speakers you do not listen to your favorite music all the time to voice it. because it risks to sound bad on all the other music.
I use the tones only when I know the pitch of something that I find irritating playing some music, for ex, if the violins sound thin or subdued, or if that particular note on the song sounds forced or if the :yelling: of the soprano is my speakers problem or the phase or it is really the recording.
So the only way to make sure is to remember that pitch or group, then test it without music with pure tones, if you hear it then, it needs to be addressed because it is not the recording's fault.
It's going to be your speakers, so take the time to voice it right, I hope my explanation helped.
I use the tones only when I know the pitch of something that I find irritating playing some music, for ex, if the violins sound thin or subdued, or if that particular note on the song sounds forced or if the :yelling: of the soprano is my speakers problem or the phase or it is really the recording.
So the only way to make sure is to remember that pitch or group, then test it without music with pure tones, if you hear it then, it needs to be addressed because it is not the recording's fault.
It's going to be your speakers, so take the time to voice it right, I hope my explanation helped.
Thanks, yes it help
I have often the feeling when I listen to (try to) rock musics from the 70ies, I feel it's all in the middle and nothing up or down, or some electro that are the opposites.
I try to select the musics that I can recognize the sound, wind or string instruments, voices. Sometimes it's just about recordings of noises from life where I can notice issues because it's a relatable reference to me.
It could work, but because of the efficiency AND the 4 ohms, it substantially limits your crossover options.
It's also expensive!
-look to the other options I mentioned on the multi-way thread.
It could work, but because of the efficiency AND the 4 ohms, it substantially limits your crossover options.
It's also expensive!
-look to the other options I mentioned on the multi-way thread.
Thanks I'll go there have a look
Please help me just clarify if I correctly done my home work :
Now everything I will write about has to be validate with actual measurement of the frequency response but bear with me to see if the theory is allright
The AXP-08 is listed as 93.2 but will experience a roll-off due to baffle step in the frequency that it will be used, and at the frequency of the crossover should be aligned with the Fostex before the bump in mid-range
The FF85WK even if listed at 86.5dB the frequency range where it will be used is around 90dB so it should be able to join the Gradient between let's say 300hz or 400hz
Now if a tweeter was added it would have to be joining the Fostex at around 2khz or 3khz also close to 90db
So we are aiming for a 90dB efficiency tweeter in 8ohm, so If I remember well a 4 ohms means +3dB and that is the problem no ?
Is it a bad idea to attenuate a tweeter to meet lower, do we loose quality by doing so ?
PS: Yes that AMT is expensive but at intertechnik it was listed at €89, it's still not cheap but acceptable.
-about 89 db 1 watt 1 meter, maybe 90 db. Around 400 Hz. Really not much different than the design that used the Gradient (already mentioned).
click on "Der Lautsprecher Club" under "BAUBERICHTE" (and scroll down considerably looking for the graph seen below here):
Club23 | Lautsprecher Intertechnik - Shop
You'll need to attenuate the tweeter if you use it in a standard configuration - and yes, resistance subjectively "dulls" the result. But better resistors do it much less, and DIY'ed resistors almost none at all. (..same thing with cap.s, but a bit less so, though DIY is impractical there.)
click on "Der Lautsprecher Club" under "BAUBERICHTE" (and scroll down considerably looking for the graph seen below here):
Club23 | Lautsprecher Intertechnik - Shop
You'll need to attenuate the tweeter if you use it in a standard configuration - and yes, resistance subjectively "dulls" the result. But better resistors do it much less, and DIY'ed resistors almost none at all. (..same thing with cap.s, but a bit less so, though DIY is impractical there.)
Attachments
Last edited:
Hey Scott, sorry to intervene, I find no drastic difference in "quality" and dullness? from 0 ohm to 10000 (exaggeration) ohm on the tweeter.
If you need parts in the XO, put them in: if you give me the choice between placing a 6uf 'audiophile' grade cap in the tweeter OR a 5.2uf cheap polar electrolytic of the CORRECT value, I would not hesitate to use the 5.2uf. The correct VALUE is important, not the part itself or its supposed quality.
No difference either between 0.5 watt no name bin resistor, 2 watts tantalum resistor, and 10 watt cement wire wound. The most important for resistors is 0 induction, 0.5Watts non-inductive is better than 10Watt inductive, the 10 watt will heat way more probably.
The resistors in the signal path is the same typical audiophile non sense than capacitors, gold binding posts, silver wires, inductors (supposedly reduce the 'speed' lol?) etc.
-and yet I've heard audible differences, along with others. (ALL ELSE EQUAL.)
Of course I also often hear differences in amplifiers (depending on the partnering loudspeaker), and even cabling in some circumstances.
So while it might be nonsense for you, it might not be for everyone.
The same is also true with the series notch filter at lower freq.s in a bass-reflex design. I tend to hear a much clearer bass presentation with a subjective half-octave of low-freq. extension (..unless the partnering amplifier has a power supply with huge current capability.) This is of course despite the fact that the room's effects dominate freq. response at these freq.s..
I should also note that a perceived negative result in one design might well be beneficial in another. Ex. in some cheaper systems that use cheap compression drivers the "dull" result was preferable (despite the freq. response not really changing).
Of course none of the above is saying that the fundamentals for the design is somehow less important than the "tweaky" bits: that would be absurd. Nor is it to say that those "tweaky" bits are necessarily true for any given person, it's just my experience (and in some cases, others as well), and it could well be for Zoltan.
(ALL ELSE EQUAL.)Of course I also often hear differences in amplifiers (depending on the partnering loudspeaker), and even cabling in some circumstances.
So while it might be nonsense for you, it might not be for everyone.
The same is also true with the series notch filter at lower freq.s in a bass-reflex design. I tend to hear a much clearer bass presentation with a subjective half-octave of low-freq. extension (..unless the partnering amplifier has a power supply with huge current capability.) This is of course despite the fact that the room's effects dominate freq. response at these freq.s..
I should also note that a perceived negative result in one design might well be beneficial in another. Ex. in some cheaper systems that use cheap compression drivers the "dull" result was preferable (despite the freq. response not really changing).
Of course none of the above is saying that the fundamentals for the design is somehow less important than the "tweaky" bits: that would be absurd. Nor is it to say that those "tweaky" bits are necessarily true for any given person, it's just my experience (and in some cases, others as well), and it could well be for Zoltan.
Last edited:
Do you speak German or google translate the hell out of it ?
err, google translate..
I can barely "operate" my own "native" language!
I blame the school system for that: at those "formative years" (I mean they are called "formative" for a reason, a MAJOR portion of which is learning to communicate properly). Most states in the US don't have other languages to learn until High School, which for someone like me was worse than useless (..because not only could I not (effectively) learn a language at that point, it was taking-away from other classes that I could have benefited from - and later limited my college selection.)err, google translate..
I can barely "operate" my own "native" language!
Well Scott, we are much alike! standard education failed me too and most everything I know is self taught. For English it was my thirst of understanding what was written in those Dungeon & Dragon books, they weren't translated at the time so I started with an English/French dictionary, after that it opened the world to me
I sometimes regret I am not better with math, many time it would have served me well.
You have proven here that you are more than competent at communicating your knowledge
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.