Jump to content

Crossover mods for the AR4x II


Recommended Posts

What is PCD?

http://techtalk.parts-express.com/showthread.php?t=217994

PCD is Jeff Bagby's Passive Crossover Designer, one of the tools from the FRD Consortium:

http://www.pvconsultants.com/audio/frdgroup.htm

The latest version of PCD may be found here:

http://audio.claub.net/software/jbabgy/jbagby.html

Based upon the data I acquired and entered, here's how it models what we have at -14°:

post-102716-1277868822.jpg

Link to comment
Share on other sites

  • Replies 88
  • Created
  • Last Reply

My highpass components are:

Solen 20 uF || Dayton 2 uF

Jantzen 0.25 mH, 18 Ga., 0.22 DCR Aircore

Dayton 10 uF

Series notch across the L-pad is:

Dayton 4.3 uF

Jantzen 0.025 mH, 18 Ga., 0.06 DCR Aircore

Dayton 3.3 Ohms 10W

PCD is telling me a better combination for the notch would be 10uF, 0.01 mH, and 6 Ohms, but I haven't tried that.

Link to comment
Share on other sites

Series notch across the L-pad is:

Dayton 4.3 uF

Jantzen 0.025 mH, 18 Ga., 0.06 DCR Aircore

Dayton 3.3 Ohms 10W

PCD is telling me a better combination for the notch would be 10uF, 0.01 mH, and 6 Ohms, but I haven't tried that.

It's a parallel notch, actually, in series with and before the L-pad.

Link to comment
Share on other sites

My highpass components are:

Solen 20 uF || Dayton 2 uF

Jantzen 0.25 mH, 18 Ga., 0.22 DCR Aircore

Dayton 10 uF

Series notch across the L-pad is:

Dayton 4.3 uF

Jantzen 0.025 mH, 18 Ga., 0.06 DCR Aircore

Dayton 3.3 Ohms 10W

PCD is telling me a better combination for the notch would be 10uF, 0.01 mH, and 6 Ohms, but I haven't tried that.

Your component values seem right but your crossover is still somehow different than mine.

My suggestions:

Don't model it a -14 degrees, thats not what you want. Somehow my network works at 0 degrees.

You need to get the overlap and phase to look like my curve from post #33. Try larger inductor and/or more inductor DCR. My inductor was about 0.25 DCR, but maybe a lossy core is upping the apparent Z at 1k (could also help the low Z dip) Might try non L-Pad in the model and see if that changes anything--it will raise the network termination Z. (Okay, probably wrong direction.)

Rather than searching for the angle of maximum reverse phase cancelation, I would set the model to your desired axis and optimize until you pulled the section phase curves to where you want them, i.e. phase overlap at the desired axis. That was why I went from 2nd order to 3rd, it got the tweeter phase where I wanted it.

I'm not wild about your HF tweeter dip circuit, it pulls down the surrounding region. Without it you were flat to 12k. Maybe your other values will reduce the width?

Davd

Link to comment
Share on other sites

Your component values seem right but your crossover is still somehow different than mine.

Yes, I posted that for assurance we are using the same component values. It'd not be unusual for me to have that part wrong.

I'm using all Parts Express stock parts so that anyone can build and expect the same result.

Still, if we can work out the best nominal values, anybody that can recap a system would be able to build their own pair.
As Laurie Fincham (KEF) used to say, anyone can design one perfect speaker, it's building multiples (in production) that is the challenge.

I haven't rounded up the supply of woofers here yet, but if I had a bunch more tweeters, I could design to the actual average of their frequency response, phase, and impedance.

[it'd be kinda labor intensive verifying the outcome, tho.... ;) ]

Link to comment
Share on other sites

I would set the model to your desired axis and optimize until you pulled the section phase curves to where you want them, i.e. phase overlap at the desired axis.

I'm at +0.67° with tweeter #6. We'll see how the others do with this:

post-102716-1277962576.jpg

post-102716-1277962614.jpg

post-102716-1277963199.jpg

Link to comment
Share on other sites

I'm at +0.67° with tweeter #6. We'll see how the others do with this:

What is the filter dip at 1100? Is your topology right?

The overlap is looking better and the summed response is "familiar".

David

Link to comment
Share on other sites

I always wondered how that turned out, since there's no follow-up. So how does the "truly awful dispersion" measured 30 degrees off-axis compare with what can be bought new in the same market range the 4x originally sold in?

Murphy's comments in his AR-3a page are also interesting ("Still, it sounds better than you might think.") At first reading I thought perhaps it meant he didn't listen to them before he measured them, but that's not what he did with the 4x.

Link to comment
Share on other sites

What is the filter dip at 1100? Is your topology right?

To resolve the impedance issue, I had thought to use a tank filter as the second pole in the highpass, but AR's original 20uF plus a series RLC notch got it above 4 Ohms. See the topology on the tweeter filter page I posted there.

The resonance frequency of the notch is in the tweeter resonance region; I can "dial" the phase around by varying the "Q" with the R value, minimally affecting the response. Same with the lowpass; it's got one, too. More work to do optimizing all of that, but it's accomplishing the objective. Here's the +/- 20° vertical polar response with the forward axis now between the drivers. Compare to what it was last week:

http://www.classicspeakerpages.net/IP.Boar...ost&id=5786

post-102716-1278011703.jpg

Link to comment
Share on other sites

So how does the "truly awful dispersion" measured 30 degrees off-axis compare with what can be bought new in the same market range the 4x originally sold in?

Domes do it better, but until very recently, cannot play this low; this was a major hurdle for vintage 2-ways. Kloss went after it with the "fried egg" tweeter, a hybrid mid/tweet in Advents.

Look back to Allison & Berkowitz's observations with respect to AR4x dispersion. From our measurements here, AR4x is effectively "done" in the top octave at +/- 22.5°:

http://www.classicspeakerpages.net/IP.Boar...ost&p=85148

Link to comment
Share on other sites

Thanks. I've actually only heard 4-series speakers a few times over the years. My "low end" AR speakers are the 6's in my office, and they have the Allison cone/dome combo tweeter (same as 4xa). It's also narrower than the domes in my 3a and 2ax, but for a tiny little 12x12 office it hasn't been an issue.

Link to comment
Share on other sites

Vertical polar response 0° to +15°, the range of typical use, and frequency response of all six tweeters using this Zilch-design crossover, measured at a nominal +0.7° axis with grille, L-pad set to 3:00.

Which pair would YOU give to Ken?

[Heh, heh.... ;) ]

post-102716-1278023396.jpg

post-102716-1278023418.jpg

Link to comment
Share on other sites

Here's the forward axes of the six tweeters (the order remains the same, but the axes on average moved up 14.125°,) and the lowpass section of the crossover.

Unless someone has something more for me to resolve with this, I'm moving on to optimizing the AR4x eWave variant.... ;)

post-102716-1278040811.jpg

post-102716-1278040828.jpg

Link to comment
Share on other sites

Here's the forward axes of the six tweeters and the lowpass section of the crossover.

Unless someone has something more for me to resolve with this, I'm moving on to optimizing the AR4x eWave variant.... ;)

Z,

What are your final crossover component values? Did they end up differing much from Dave's?

Roy

Link to comment
Share on other sites

What are your final crossover component values? Did they end up differing much from Dave's?

I wouldn't call it "final" yet, but the topology and components are different.

I'm at ~$70 apiece by current Parts Express list prices:

http://www.classicspeakerpages.net/IP.Boar...ost&id=5813

http://www.classicspeakerpages.net/IP.Boar...ost&id=5820

eWaving them instead will be ~$40 each more....

Link to comment
Share on other sites

I've started playing with PC Designer. Any chance you could send me some of your driver response and Z curves?

Here is some background info on crossover design software for those that may be following along.

To design a crossover by software you need bring together a lot of information at the same time. You start with the raw material of the woofer and tweeter. You'll need the frequency response of the individual elements on axis, on the intended cabinet. If the modeling software gives off axis curves then you need those measured at your intended angles (most designers model on axis only and then measure the prototype off axis).

You need the impedance curves. The crossover you simulate isn't loaded by a resistor, it drives the impedance curve of each driver so a woofer and tweeter impedance curves are required.

You need to define the layout of the system. This is where the tweeter lies on the baffle and where the woofer is positioned below it, also relative depth of the two units. This information determines the relative phase response of the units both on axis and off axis. As an example, if the units are far apart then a small swing of the "microphone" up or down will move one unit closer and the other unit farther from the microphone by a fair amount. This is what creates nulls at the crossover frequency, as Zilch is describing. A design ideal is to get the best response on some nominal "reference axis" and have the inevitable nulls evenly spaced above and below that axis (as far apart as possible). The software needs to simulate that response for any observation angle you want to define. The relative movements of microphone position are advancing or retarding the phase response of the individual units and therefor modifying how their individual responses add in space. Also, the depth of the units will determine their delay or phase shift, unless that was part of the original response curve (different programs handle this differently).

You need to define a crossover topology. These are generally "ladder networks" with a common ground and then a number of series and shunt elements. First order, second order, third order, each just add another element to the sequence. Also add any equalizing dip circuits or impedance modifying circuits (Zobel, etc.) Also, if you want an accurate simulation, you will have to include parsitic crossover losses as extra resistance.

With all this, the software starts calculating the node to node impedance then voltage drop of all the circuit elements as loaded by the impedance curve of the driver, This gives the voltage curve of the network at the woofer or tweeter that modifys the raw unit response that you entered. It generally rolls off the response as desired and adds extra phase shift at the band edges. The response, with network, of the woofer, and the same for the tweeter, are added with the air path delays defined by their phase curves and/or depth and the geometry (location on the box and listening angle) that you defined, accounted for, all as observed at the microphone location you defined.

The end result is the full simulated frequency response of the system. Any less than this involved approach will not give any meaningful results.

Most software includes optimizers that let you specify an intended response shape (acoustical response, the combination of driver and network) and will automatically nudge the crossover values around until the section hits that shape (or the values blow up and the optimizer crashes!)

I believe KEF was the first company to use this approach with proprietary software in the early 80s. The "Acoustic Butterworth" models were the result. A lot of us used XOPT (a Peter Schucks program) and perhaps LEAP crossover shop.

Without them we wouldn't be able to do the ever more sophisticated crossover networks used by "good" designers these days. They make better direct field summing very easy to visualize.

David

Link to comment
Share on other sites

I've started playing with PC Designer. Any chance you could send me some of your driver response and Z curves?

I expected you'd be tempted. ;)

The files for the seven EconoWave designs completed thus far this year are in the Tech Talk thread:

http://techtalk.parts-express.com/showthread.php?t=215536

That's kinda the chronicle of my PCD adventures to date; it has grown to become that forum's most-viewed thread. Jeff fixed most of the bugs I found in his latest update, V7, though I haven't tried it yet. I'm still using V6.2 with workarounds.

Here are the files for the stock AR4x, woofer #1 and tweeter #6. My standard protocol is to measure at the midpoint between drivers from 44". HF usually includes an L-pad set at 1:00. Phase is CLIO autocalc minimum, and I also do an FRD with both drivers together. I load that in as an overlay in the main window to get a first approximation of the driver "z" offsets. Delete ".txt" to use the files in PCD:

AR4x_HF_6.frd.txt

AR4x_HF_6.zma.txt

AR4x_LF_1___HF_6.frd.txt

AR4x_LF_1.frd.txt

AR4x_LF_1.zma.txt

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...