Thanks for the info.
In theory using ideal resistive looking drivers, you can achieve the same result with either 1 or 2 resistors just by adjusting the reactive components.
As for using real drivers, I'm sure that you could achieve the same result either way if you were willing to add components as needed.
So with that said I'd like to run a simulation, using ideal drivers except for the tweeter impedance variations.
In theory using ideal resistive looking drivers, you can achieve the same result with either 1 or 2 resistors just by adjusting the reactive components.
As for using real drivers, I'm sure that you could achieve the same result either way if you were willing to add components as needed.
So with that said I'd like to run a simulation, using ideal drivers except for the tweeter impedance variations.
Last edited:
I simulated with 2.2mH, 24uF, 0.76R and 15R (dotted traces).
Then I removed the 15R and increased the 0.76R to 0.9R for the same attenuation.
Then I removed the 15R and increased the 0.76R to 0.9R for the same attenuation.
It's very interesting to see how it affects each driver's output in opposite ways. I don't know what it means yet but it's something to learn about.
The big dips in FR I assume are related to distance from the speaker and maybe nodes in the room.
Is it logical to think that adding to the value of inductance would lower the woofer's output under the tweeter and allow getting rid of the resistor altogether?
What question should I ask next?
The big dips in FR I assume are related to distance from the speaker and maybe nodes in the room.
Is it logical to think that adding to the value of inductance would lower the woofer's output under the tweeter and allow getting rid of the resistor altogether?
What question should I ask next?
You are forcing your AMT too low. With your current values of 2.2mH / 24uF the XO frequency is in the vicinity of 700 - 800Hz.
If you go back to your original 1.3mH, but 22uF, plus 1 - 2R series resistor, you'll be much closer to 1kHz.
And that should be your starting point.
If you have or can measure the parameters of your drivers and enter them in Basta! you'll be able to model your entire speaker very quickly.
Basta does not do Series XO, but does 1st order Butterworth (Q = 0.7) and you use those values in your XO without a problem.
Working directly from the previous simulation with the original overlay, I was able to get the single resistor version closer using 1.8mH and 19u5F.
Thanks to both of you .
The graphics are giving me a new perspective on terminology. So in the graphic immediately above , Fc would be about 2300 Hz.?
As I already had the original combi yielding 1.3mH , I went back to that and followed Stanislav's recommended value of 22uF (actual 22.6uF) , with single series 1.25Ω . Not bad though some music I like is still a bit irritating and there's often a kind of fuzzy burr around some horn tones that I guess is resonance. I'll try increasing the pad resistor next.
PS. AllenB - Is that simulation software you are using something you'd recommend?
Thanks
The graphics are giving me a new perspective on terminology. So in the graphic immediately above , Fc would be about 2300 Hz.?
As I already had the original combi yielding 1.3mH , I went back to that and followed Stanislav's recommended value of 22uF (actual 22.6uF) , with single series 1.25Ω . Not bad though some music I like is still a bit irritating and there's often a kind of fuzzy burr around some horn tones that I guess is resonance. I'll try increasing the pad resistor next.
PS. AllenB - Is that simulation software you are using something you'd recommend?
Thanks
Attachments
I think there are a few issues at play here.
You are running both drivers in their non-linear range.
The AMT cannot play low, 1kHz should be the minimum, better 1.5 - 2kHz.
The 12 incher cannot play high, starts beaming probably at around 500Hz.
Due to the nature of the Series XO you don't see a hole in the response, but your ears tell you the sound is not right.
Solutions:
1. make it a 3-way with a wide range covering 300 - 3000Hz
2. use a different, smaller LF driver - some of Mark Audio units are quite suitable for that.
I think AllenB uses XSim, a little xover simulator by Bill Waslo of Liberty Instruments fame.
It's a very good piece of SW but it has its limitations.
Can only model XO based on driver's imported response, does not take into account the whole speaker - acoustic loading, shape and size of box and baffle, location of drivers on the baffle and so on.
That's why I suggested Basta, which does all this, but does not do Series XO, cannot import drivers responses.
Using both programs as a combination works quite well.
You are running both drivers in their non-linear range.
The AMT cannot play low, 1kHz should be the minimum, better 1.5 - 2kHz.
The 12 incher cannot play high, starts beaming probably at around 500Hz.
Due to the nature of the Series XO you don't see a hole in the response, but your ears tell you the sound is not right.
Solutions:
1. make it a 3-way with a wide range covering 300 - 3000Hz
2. use a different, smaller LF driver - some of Mark Audio units are quite suitable for that.
I think AllenB uses XSim, a little xover simulator by Bill Waslo of Liberty Instruments fame.
It's a very good piece of SW but it has its limitations.
Can only model XO based on driver's imported response, does not take into account the whole speaker - acoustic loading, shape and size of box and baffle, location of drivers on the baffle and so on.
That's why I suggested Basta, which does all this, but does not do Series XO, cannot import drivers responses.
Using both programs as a combination works quite well.
Any crossover simulator is fit for that type of comparison. These days I often use XSim or Vcad to answer questions of this nature. The plot above uses Vcad.
What's important is that you include anything that's relevant.. in other words are careful to simulate the right thing. Any questions feel free to ask. As Stanislav mentions..
The less the simulator does, the more I do beforehand or concurrently using support software and measurements. I can therefore use XSim to work on a complex result. People have been exceeding the capabilities of Vcad since before it was written.
Doing some reading I gather that the true nominal size of a woofer should be stated as the diaphragm diameter plus half the surround. I casually called this a 12" woofer but measuring as above it's only 8.5".
I chose this area to work in as the woofer's owner told me not to cross it above 1600Hz and I thought it likely that the closer to the upper limit the steeper the crossover needs to be.
Searching for user experience with the AMT , I found one user got results they were after crossed at 800Hz. (though I did see other posts saying that was too low. ) Looking at the parts I have on hand 1KHz seemed a place to start.
Incidentally I can really hear that the series resistor has a big effect on the entire system. I got it to 3.3Ω and it was sounding a lot better but I could hear some high pitched sounds on certain transient peaks. Upping it to 4Ω sounds very good until those loud peaks start to shout.
So the next step here would be to sweep each driver individually without crossover?
Related to that - the woofer seems to be getting a little better in the bottom end and wonder if it's still breaking in. Are drivers usually run for a bit before testing for crossover design?
Your generosity is always appreciated a lot.
Thanks
H.
Hi H.
You are correct, the cone area of a driver is never its overall diameter. All of my 12" drivers, and I have plenty, have a cone that measures 25 - 26 cm diameter, slightly more than 8.5".
I understand you desire to cross as high as possible.
And yes, the resistor will have a big effect on the entire system's sound. But that resistor attenuates only the HF driver's output and what you hear is most likely the break up of the LF driver.
There's one area you have not said anything about, how are you addressing Baffle Step?
Because every speaker goes through that transition.
And with a 1st order series xo it's very difficult to deal with that unless you cross at the Baffle Step frequency.
Which in your case is around 300 - 350 Hz.
Yes, you read that right, 300 - 350 Hz.
If you wonder how did I get to that figure the formula is this:
115 / WB = Fbsc
where WB is the width of your baffle
115 is the speed of sound divided by 3 ( source David Weems & Peter Comeau)
If you choose to switch to a parallel xo you have the choice of correcting for BSC with some additional circuitry. Interesting reading here:
https://www.diyaudio.com/community/threads/baffle-step-compensation-methods.230342/
All in all, I am convinced your idea of a 12" LF and an AMT crossed at around 1kHz with a series xo will not work.
Introducing an MF driver to cover 300 - 3000 Hz will be the easiest way to deal with all the issues.
As Allen says, keep asking questions, I'll try to answer as best as I can.
Cheers
Stan
You are correct, the cone area of a driver is never its overall diameter. All of my 12" drivers, and I have plenty, have a cone that measures 25 - 26 cm diameter, slightly more than 8.5".
I understand you desire to cross as high as possible.
And yes, the resistor will have a big effect on the entire system's sound. But that resistor attenuates only the HF driver's output and what you hear is most likely the break up of the LF driver.
There's one area you have not said anything about, how are you addressing Baffle Step?
Because every speaker goes through that transition.
And with a 1st order series xo it's very difficult to deal with that unless you cross at the Baffle Step frequency.
Which in your case is around 300 - 350 Hz.
Yes, you read that right, 300 - 350 Hz.
If you wonder how did I get to that figure the formula is this:
115 / WB = Fbsc
where WB is the width of your baffle
115 is the speed of sound divided by 3 ( source David Weems & Peter Comeau)
If you choose to switch to a parallel xo you have the choice of correcting for BSC with some additional circuitry. Interesting reading here:
https://www.diyaudio.com/community/threads/baffle-step-compensation-methods.230342/
All in all, I am convinced your idea of a 12" LF and an AMT crossed at around 1kHz with a series xo will not work.
Introducing an MF driver to cover 300 - 3000 Hz will be the easiest way to deal with all the issues.
As Allen says, keep asking questions, I'll try to answer as best as I can.
Cheers
Stan
Last edited:
Showing a generic baffle step compensation circuit (cross=1k, baffle step=500Hz). YMMV.
You guys are starting to look like wizards.
Based on what I get from the formula in Stanislav's posted Peter Comeau article . With baffle 356mm x 685mm , Baffle step starts at 344/.712 = 483Hz. and F3 is at 115/.356 = 323Hz. Have I got that right?
Haven't had time for any follow up but hope to get something to post tomorrow or next day ( will try the circuit AllenB posted and post test.).
One other question about the baffle step thing is re: height above the floor. The boxes are on stands too tall for them and how does calculation for the baffle step relate to that. I plan to try some milk crates to bring the woofer down closer to ear height and expect that will get a bit more low end.
BTW, Is there any source for whole series of Comeau's Speaker Design articles? The one posted above is very easy to follow. If the rest are similar it would be very helpful.
Thanks !
Based on what I get from the formula in Stanislav's posted Peter Comeau article . With baffle 356mm x 685mm , Baffle step starts at 344/.712 = 483Hz. and F3 is at 115/.356 = 323Hz. Have I got that right?
Haven't had time for any follow up but hope to get something to post tomorrow or next day ( will try the circuit AllenB posted and post test.).
One other question about the baffle step thing is re: height above the floor. The boxes are on stands too tall for them and how does calculation for the baffle step relate to that. I plan to try some milk crates to bring the woofer down closer to ear height and expect that will get a bit more low end.
BTW, Is there any source for whole series of Comeau's Speaker Design articles? The one posted above is very easy to follow. If the rest are similar it would be very helpful.
Thanks !
Last edited:
The cabinet height does make a difference but since the sides are usually closer to the driver, and within the driver nearfield, they have the greater effect. The sides also have the greater share of the angular coverage (radially out from the driver along the baffle). These reasons are why calculations tend to focus on width.
Hi H.,
Yes, you've got that right. Don't pay too much attention to the exact figures, 300 - 350 Hz range is just fine.
Allen's circuit diagram above is the only sensible way of doing BSC in a series xo when not crossing at the baffle step. Nicely done.
Don't know about a source for Peter Comeau's articles, but I have copies of 10 of them from the now I think long gone Hi-Fi World from 2006.
For the benefit of the whole community here I am going to post them in this thread, with the risk of not playing by the rules.
Cheers
Yes, you've got that right. Don't pay too much attention to the exact figures, 300 - 350 Hz range is just fine.
Allen's circuit diagram above is the only sensible way of doing BSC in a series xo when not crossing at the baffle step. Nicely done.
Don't know about a source for Peter Comeau's articles, but I have copies of 10 of them from the now I think long gone Hi-Fi World from 2006.
For the benefit of the whole community here I am going to post them in this thread, with the risk of not playing by the rules.
Cheers
While I understand the rudiments of R, C and L, this simple looking addition is more than I can grok without a little orientation. I notice that Stanislav - who clearly knows his way around - was impressed by your solution too, so can you say something about the thought process one might follow that might lead to that sort of circuit?
I'm meeting up with a friend on the weekend for some inductive horse trading so I can try out your circuit. Will post how it turns out. Thanks Allen.
Thanks VERY much for posting those. I did manage to find the Hi-Fi World issues archived online but the scans were not very high rez. and when converted to pdf the text and image quality is poor. Yours are infinitely better . Thanks !
It is in series with the crossover and it's effect covers both drivers.
The basic function is to apply series resistance at higher frequencies. The inductor removes itself from the circuit at higher frequencies since it's reactance exceeds the level of the resistor. At low frequencies the inductor removes the resistor from the circuit by shorting it. The value of inductance changes the frequency and the value of resistance changes the level.
It models well in a simulation where ideal impedances are used, which overall remain smooth with changing frequency. A practical series crossover benefits naturally from some impedance averaging already but it's possible you may need to tweak this in practice, as with any crossover.. perhaps by adjusting the tweeter level resistor or in some cases by adding some impedance compensation.
It can, of course, be simulated if you provide sufficient data. If you want to experiment with levels and frequency and don't have a range of component values, you can try these with EQ using a shelving filter.
The basic function is to apply series resistance at higher frequencies. The inductor removes itself from the circuit at higher frequencies since it's reactance exceeds the level of the resistor. At low frequencies the inductor removes the resistor from the circuit by shorting it. The value of inductance changes the frequency and the value of resistance changes the level.
It models well in a simulation where ideal impedances are used, which overall remain smooth with changing frequency. A practical series crossover benefits naturally from some impedance averaging already but it's possible you may need to tweak this in practice, as with any crossover.. perhaps by adjusting the tweeter level resistor or in some cases by adding some impedance compensation.
It can, of course, be simulated if you provide sufficient data. If you want to experiment with levels and frequency and don't have a range of component values, you can try these with EQ using a shelving filter.