RoHo Posted December 3, 2019 Posted December 3, 2019 I have custom made speakers. They are sealed box 2 ways with a 6 inch wide-band driver and Fountek ribbon tweeter. They roll off at about 80 Hz and bass is handled by a subwoofer. When building the speakers the builder gave me the choice of two crossovers: a conventional 3rd order at about 3kHz or a simpler 1st order. We went with the 1st order which I've been happy with for a number of years. Eventually I came to realise that there is quite a bit of harshness in the treble, mainly evident on female vocals, which I hypothesise is due to the woofer cone break-up being audible in the 5-10K region. So I decided to change to the 3rd order crossover, using the circuit drawn up by the builder. The crossover is internal and accessible by taking the base off the speaker which is rather tedious but I new "it was time". So out with the old, in with the new, hook up the measuring gear and see how it looks: Not too bad and certainly flatter than the old crossover which had a dip in the lower treble as well as the cone break-up peaks I mentioned earlier. So time for a listen, hit play and.......AARGH! Awful, over-prominent high treble. Percussion and female vocals had an awful tizzyness, unlistenable. Hmmm....figuring the tweeter was to blame I covered it with some cardboard and listened to the woofer. Not too bad. Less sibilance than previously though still noticeable. So I figure I've got a problem with the tweeter level. I can't be bothered pulling out the crossover and changing the L-pad so I decide to improvise using my vast knowledge of the attenuation factors of household fabrics: Looks great doesn't it? And it sounded a whole lot better, much better than I expected. So I took another measurement: Purple is as before, green is with "acoustic attenuation". Doesn't look that remarkable, just a lower level but the difference in audio quality was huge. So I live and learn. Any comments? A quick look at the graph tells me I need to add another 4-5 dB of attenuation to the tweeter. Sounds about right? While I'm at it, here's a measurement of the woofer up close with the new, 3 rd order, crossover. The cone break-up is shown at about 8-10kHz. The question is -is this at a low enough level to be inaudible or should I consider adding a notch filter? Any comments welcome. If your interested I can also display all the pics of the drawn out process of changing the in-room bass response from bad to ...slightly less bad!
davewantsmoore Posted December 3, 2019 Posted December 3, 2019 13 minutes ago, RoHo said: While I'm at it, here's a measurement of the woofer up close with the new, 3 rd order, crossover. This measurement shows the woofer rolling off about "2nd order" ie. about 12dB/octave. 13 minutes ago, RoHo said: The question is -is this at a low enough level to be inaudible or should I consider adding a notch filter? No... and.... that might be one way to deal with it. Although given the above comment, something could be "wrong" more generally, and so maybe "back to the drawing board" is better. (It's hard to say too much definitive, without more info to go on, so I don't want to lead you too much). Quote AARGH! Awful, over-prominent high treble I don't really see that in your measurements. To me it looks perhaps a little hot around 1-2Khz.... but the tweeter level looks ok. Do you have a measurement of the response with the old crossover?
Ittaku Posted December 3, 2019 Posted December 3, 2019 What components do you have in the tweeter crossover? Going to a third order crossover would involve much higher capacitor values which would generally mean much lower quality components. A good crossover potentially can cost as much as the drivers when done well
RoHo Posted December 3, 2019 Author Posted December 3, 2019 11 minutes ago, davewantsmoore said: This measurement shows the woofer rolling off about "2nd order" ie. about 12dB/octave. No... and.... that might be one way to deal with it. Although given the above comment, something could be "wrong" more generally, and so maybe "back to the drawing board" is better. (It's hard to say too much definitive, without more info to go on, so I don't want to lead you too much). I don't really see that in your measurements. To me it looks perhaps a little hot around 1-2Khz.... but the tweeter level looks ok. Do you have a measurement of the response with the old crossover? Woofer XO is .55mH, 10uF, .16mH. There was notch filter at the break- up frequency on the old XO so I was going to re-use that as a quick trial. I'll hunt up a plot of the response with the old XO. The lower-mid treble was recessed and upper treble more prominent so the aim was for a flatter overall response.
RoHo Posted December 3, 2019 Author Posted December 3, 2019 20 minutes ago, Ittaku said: What components do you have in the tweeter crossover? Going to a third order crossover would involve much higher capacitor values which would generally mean much lower quality components. A good crossover potentially can cost as much as the drivers when done well Yeah, the new tweeter caps are 4.7uF and 10 uF. Luckily I had some Auricaps from my previous speakers lying around.
Ittaku Posted December 3, 2019 Posted December 3, 2019 (edited) 14 minutes ago, RoHo said: Yeah, the new tweeter caps are 4.7uF and 10 uF. Luckily I had some Auricaps from my previous speakers lying around. Auricaps are fine quality metallised polypropylene so I think that's one less thing you have to worry about. The only other issue is whether the crossover really is doing what you think it's doing as third order crossovers are really sensitive to the impedance being off from that chosen in the design and there may be more overlap and wrong phase alignment between the woofer and tweeter components. If you need tweeter attenuation, ribbon tweeters are quite directional and simply making them more off-axis can tame them. Edited December 3, 2019 by Ittaku
mwhouston Posted December 3, 2019 Posted December 3, 2019 The tweeter is probably easy to get out. Put a 3db pad across it. I solder these on directly. You only need two resistors if you want values I can help. And how to attach,. May have a better WAF than a green towel.
afa Posted December 3, 2019 Posted December 3, 2019 What model Fountek ribbon tweeter are you using and could you post a schematic of the new crossover you are using.
RoHo Posted December 4, 2019 Author Posted December 4, 2019 13 hours ago, Ittaku said: Auricaps are fine quality metallised polypropylene so I think that's one less thing you have to worry about. The only other issue is whether the crossover really is doing what you think it's doing as third order crossovers are really sensitive to the impedance being off from that chosen in the design and there may be more overlap and wrong phase alignment between the woofer and tweeter components. If you need tweeter attenuation, ribbon tweeters are quite directional and simply making them more off-axis can tame them. Tying phase and impedance into crossover design are beyond my (current) knowledge, which is why I'm trying some simple changes first. Looking back , I found myself listening more in a higher chair, above the tweeter axis for the reason you mention - reduced tweeter output 13 hours ago, mwhouston said: The tweeter is probably easy to get out. Put a 3db pad across it. I solder these on directly. You only need two resistors if you want values I can help. And how to attach,. May have a better WAF than a green towel. I am going to change the L-pad resistors this weekend - maths and electronics I can handle, but don't ask me to do any woodwork! You don't like the green? The other speaker has contrasting black/white teatowel. Looks very jazzy.
Ittaku Posted December 4, 2019 Posted December 4, 2019 2 minutes ago, RoHo said: Looking back , I found myself listening more in a higher chair, above the tweeter axis for the reason you mention - reduced tweeter output Try pointing them straight ahead then, instead of at your listening chair, then experiment with more and more toe in till you get the balance you like. Free mod! 1
RoHo Posted December 4, 2019 Author Posted December 4, 2019 (edited) 13 hours ago, afa said: What model Fountek ribbon tweeter are you using and could you post a schematic of the new crossover you are using. Ribbon is a JP3.0, a pure aluminium ribbon model. Can't post a diagram ATM but the tweeter crossover is C(4.7u), L(0.22m), C(10u) then the L-pad (4.1R, 6.8R). Edited December 4, 2019 by RoHo
davewantsmoore Posted December 4, 2019 Posted December 4, 2019 14 hours ago, RoHo said: There was notch filter at the break- up frequency on the old XO so I was going to re-use that as a quick trial. Depending on the strength and Q of the filter .... is may have been making a significant contribution to the rolloff. 14 hours ago, RoHo said: Woofer XO is .55mH, 10uF, .16mH. I see. That's higher than I expected (eg. filter is 30db down at 10khz) ... but does explain the response shown. 14 hours ago, RoHo said: The lower-mid treble was recessed and upper treble more prominent so the aim was for a flatter overall response. This may have been done on purpose. The ribbon tweeter has a wide response at LF ..... so for the overall "power response" to be balanced, they may have left a dip in the axial response..... and so making the axial response flat(er), puts more (too much) energy into the "few khz" range. To really understand the speaker, you need to look at frequency response vs horisontal angle to see the total energy.
RoHo Posted December 4, 2019 Author Posted December 4, 2019 3 minutes ago, davewantsmoore said: Depending on the strength and Q of the filter .... is may have been making a significant contribution to the rolloff. I see. That's higher than I expected (eg. filter is 30db down at 10khz) ... but does explain the response shown. This may have been done on purpose. The ribbon tweeter has a wide response at LF ..... so for the overall "power response" to be balanced, they may have left a dip in the axial response..... and so making the axial response flat(er), puts more (too much) energy into the "few khz" range. To really understand the speaker, you need to look at frequency response vs horisontal angle to see the total energy. Good information, Dave, thanks. Your mention of the horizontal dispersion pattern gets me thinking about the reason I posted the first two plots. If that is the response at the listening position, a sum of all the reflections, don't these plots reflect the interaction of the polar response with the room ? Given that I'm happy to optimise the system for this set listening position. So the questions are 1) is this FR unacceptable? And 2) is it worth redesigning the crossover?
davewantsmoore Posted December 4, 2019 Posted December 4, 2019 26 minutes ago, RoHo said: If that is the response at the listening position, a sum of all the reflections, don't these plots reflect the interaction of the polar response with the room ? It's a good question... it depends on how you have REW configured.... but default settings you are seeing reflectons from the room the data. However. In short, your microphone doesn't listen quite the same as your ear does. 26 minutes ago, RoHo said: So the questions are 1) is this FR unacceptable? The one you posted previously looks OK. ..... but seeing the response curve at other angles, helps to understand what is really happening. 26 minutes ago, RoHo said: And 2) is it worth redesigning the crossover? That's hard to answer. If you liked the sound well enough of the old XO ..... then have a look at it's response under the same measurement conditions. If it's substantially different to what you've got now.... then you may want to question your goals. Given the substantial change in directivity that these two drivers will have at ~3khz, then that does make some justification for overlapping the drivers substantially (ie. using shallow rolloff slopes).
RoHo Posted December 4, 2019 Author Posted December 4, 2019 1 hour ago, davewantsmoore said: Given the substantial change in directivity that these two drivers will have at ~3khz, then that does make some justification for overlapping the drivers substantially (ie. using shallow rolloff slopes). That makes a great deal of sense, thank-you.
RoHo Posted December 4, 2019 Author Posted December 4, 2019 As promised REW plots showing the old and new crossovers Red-new 3rd order XO Blue-old 1st order XO Green-nearfield woofer response with 1st order XO
b0dhi Posted December 6, 2019 Posted December 6, 2019 Would be helpful to see phase response in these graphs. Are you doing phase compensation? Would also be good to see tweeter nearfield to see how much of that green curve's top octave is bleed-in from the tweeter. I once had a similar problem with harshness in female vocals introduced with a change in crossover, which was resolved by re-adjusting phase compensation.
RoHo Posted December 6, 2019 Author Posted December 6, 2019 Below are the nearfield woofer and tweeter and the combined sum at the listening position. All with the "new" 3rd order XO. Understanding phase manipulation is beyond me but I can display whatever you like. Which REW plot would you like? One problem with REW is that there are dozens of functions to choose from. I'm sure the XO designer took phase into consideration, can I tell from the component values? My next step will be to add the notch filter that was on the woofer with the 1 st order XO (0.04 mH, 6.8 uF, 10R) to see if that further reduces the 8-10k peaks. If it improves things at all I will be happy. It does sound pretty good as is.
Recommended Posts