|
Home » Sponsored » Pi Speakers » 4Pi crossover study
| 4Pi crossover study [message #77204] |
Thu, 18 July 2013 21:17  |
Joe Sever
Messages: 11
Registered: May 2009
|
Chancellor |
|
|
After having spent several months biamping with MiniDSP, I have for several reasons decided to go the passive route (or at least away from MiniDSP). Given that my "design" is inspired by the 4Pi, along with the flexibility of its crossover, I figured I'd give Wayne's crossover a shot. However, despite some reasonable success with the MiniDSP, the 4Pi crossover is not doing what I expect (or at least would like) it to do. A 45 SET amp supplies the input to the crossover.
The speakers:
5.5 cu. ft. box tuned to 39.5Hz
Omega Pro-15A
Radian 475PB on SEOS-12 waveguides
Crossover details:
R1 = single 30 ohm 10W resistor
R2 = single 13 ohm 10W resistor
Rs = none
C1 = none, though several values from 0.33uF to 6.8uF have been tried
Here's the problem: Measured 1M from the speaker baffle and midway between the LF & HF drivers, the frequency response rolls off in a linear fashion, beginning at 6KHz or so and ending about 12dB down at 20KHz. I would expect this without the compensation EQ in place, but not with it. I think the attenuation around 6KHz is correct - it's just the lack of rolloff compensation that has my stumped. To rule out a bad driver, I measured the other speaker with the same results.
I know that Wayne has compared the SEOS-12 and the H290C side by side on the DE250, and the SEOS-12 showed a steeper rolloff than his horn, but not to the degree that I'm seeing.
Here are what I see as my options:
1) Buy Wayne's horns and hope that that alone will solve the problem.
2) Try a different driver. I've tried the DE250 in the past and didn't like it. I have a Faital HF10AK on hand that I haven't listened to yet.
3) Scrap the 4Pi crossover and try something else.
I'm leaning toward #1, but deep down I don't believe that's the answer. There has to be something fundamental I'm missing, but I've been over the circuits and Wayne's various docs and I'm at a loss.
Even with its current shortcomings, this crossover blows the MiniDSP out of the water, so I'm hoping that with some tweaking I can come up with something I can live with.
Thanks in advance for any help anyone can lend.
|
|
|
|
| Re: 4Pi crossover study [message #77207 is a reply to message #77204] |
Thu, 18 July 2013 22:18  |
 |
Wayne Parham
Messages: 18787
Registered: January 2001
|
Illuminati (33rd Degree) |
|
|
The π crossover will not work very well on the SEOS12 because SEOS horns have too much ripple. Designers using this horn usually incorporate notch filters to get decent response. To me, this is a deal breaker. That's why I did not adopt that horn in my designs.
You can use notch filters to smooth the response, but I don't think it sounds right on any horn I've ever used with this same problem. I walked away from horns like that years ago for this very reason.
The SEOS ripple is about 5dB, and it is definitely audible. So the crossover needs to do something about it. When Erich asked about the notch filters in the crossovers his designers were using, Bill Waslo said this:
"It's those Inductor-Capacitor-Resistor (LCR) strings that go across the CD driver. Can't just take them out, other stuff would need to be adjusted to compensate or it would sound awful. I did run some designs without those LCR, but really think they should be left in. One of the bumps they deal with is at 2kHz, which is a terrible frequency to have a bump at (near where ears are most sensitive)."
In contrast, the H290C is smooth, only having about 2dB ripple. It doesn't need notch filters to sound good. This is a requirement for using the π crossover, because the only response tailoring it does is to conjugate mass-rolloff. It is extremely flexible in this regard, provding a way to set the damping in the crossover region, which sets the SPL in the first couple octaves independently of the augmentation used for mass-rolloff compensation. But it doesn't provide any mechanism for mitigating response ripple.
More information on the π crossover:
One more thing: When you work with a horn/waveguide, you have a funny impedance curve that can sometimes interact with the crossover and create peaks, even if the horn/waveguide response is smooth when connected directly to the amp. This is especially true when padding is used, as is usually the case. But the impedance peaks can easily be damped with a snubber resistor, much the same way a Zobel is used on woofers. You see this in some waveguides, like the QSC and the H290C. I really like it when even those impedance peaks are below the passband, but it isn't really a problem if it's above crossover, it's just that you'll need a snubber resistor or Zobel to flatten the impedance curve.
But when you use a horn that has a lot of ripple, you either have to use a bunch of notch filters, or you have to accept the ripple. Neither one is acceptable to me. I would rather use a horn/waveguide that has smoother response.
H290C compared with SEOS12
|
|
|
|
|
|
| Re: 4Pi crossover study [message #77210 is a reply to message #77208] |
Fri, 19 July 2013 07:08 |
Joe Sever
Messages: 11
Registered: May 2009
|
Chancellor |
|
|
Thanks for the reply, Bruce.
Two other possibilities have occurred to me. One is, as you suggest, that my mic is in a null - I'll check this evening. I'm actually hoping this is the case, though I think not given the linear nature of the dropoff and that it gets worse at the upper extreme of the response.
The other is that the amp's OPTs are rolling off. I really doubt the latter as others have measured them as ruler flat.
FWIW I'm using REW, with a calibrated Dayton EMM-6 mic through an MXL USB Mic Mate. I've been measuring in my small room with the speaker in its normal listening position; I really need to lay it on its back, though that may just clean up the hump and dip (floor bounce?) that I see well below the crossover point.
Wayne - Many thanks for your input.
While I appreciate that the SEOS may be introducing ripple on top of my other issue(s), my approach is to address one issue at a time. I'm certainly not married to this particular horn.
My primary issue at the moment is that the mass rolloff doesn't appear to be touched by the compensation circuit - it's an obvious 12dB dropoff from 6K-20K. My fear is that while another horn may be a better bet in the long run, I'll still be left with this weird rolloff to deal with.
I'll post some images of the measurements - maybe something will jump out that I'm not adequately describing. And though I hate to disturb my pretty soldering, I should probably also measure without the compensation circuit in place to see what, if anything, it's doing.
|
|
|
|
| Re: 4Pi crossover study [message #77211 is a reply to message #77210] |
Fri, 19 July 2013 12:12 |
|
Wayne Parham
Messages: 18787
Registered: January 2001
|
Illuminati (33rd Degree) |
|
|
I think it is smart to tackle one issue at a time. The way you're going about it is good. It's just that I also think if you get the response right for the SEOS, it won't be right for the H290C and vice versa. The biggest difference is the ripple, but the rolloff slope is different too.
A horn with constant directivity will present the response of the driver, unmodified. So mass-rolloff is there. However, if the horn either increases directivity or increases efficiency as a function of frequency, then that will provide some acoustic equalization. Also, some drivers don't have as much mass-rolloff as others.
High-end rolloff is largely a diaphragm thing, not just mass but also flex modes. Sometimes you'll see mass-rolloff countered by damped breakup, and this can increase the top end. Where there isn't enough damping response is rough, but if the diaphragm has enough internal self-damping, response is smooth and it just looks like the driver has extended output. In any case, the top-end response of some drivers shows more rolloff than others.
What the π crossover allows you to do is to set the output in the first octave independently of the output in the top two octaves. It also lets you set the amount of augmentation in the top two octaves.
Here is the crossover for the four π speaker, which we can use as an example. All π Speakers that have a waveguide/horn use this same basic circuit topology.
Top-octave compensation is set mostly by the R1/C1 values, but it is also affected by voice coil inductance and the amount of damping from the Rs resistor, if used. Voice coil inductance tends to increase rising response. A shunt value in the Rs position tends to reduce rising response. And of course, the use of capacitor C1 increases rising response, and also sets where it begins.
The optional shunt value Rs is used for horn/driver impedance control. If the impedance peaks are all below crossover, then resistor Rs isn't needed. But if there are impedance peaks above crossover, then they will need to be snubbed by a resistor in position Rs. It's like a Zobel. The value is generally around 2x the DCR of the compression driver, which prevents the Zmax from rising too high. One could also use an inductor in series with Rs, and then use a lower value resistor. But I find the resistor by itself is sufficient.
The bottom octave shelf is created mostly by the value of R2. When it is high, the damping is reduced, boosting output near crossover. This is what creates the shelf. Where the shelf isn't needed, the R2 resistor should be set low, to provide more damping. Increased damping (using lower R2 value) can also be used to counter an impedance spike, similarly to Rs. But R2 is providing damping for the crossover, not the driver. So it acts differently than Rs, in that it isn't modifying the driver's impedance. Being ahead of R1/C1, resistor R2 is more effective at modifying the transfer function within the crossover region, whereas Rs is more effective at modifying the curve above the crossover region. Again, Rs is used just like a Zobel.
Balanced with all this, the R1/R2 ratio sets the overall attenuation. You have to set R1/R2 values that give the right attenuation for matching the tweeter output to the midwoofer, while simultanously setting R2 for the right bottom octave shelf, and R1 and C1 for the right top-octave augmentation.
This is the basic transfer function we are looking for. It may need a little more or less overall attenuation, or a little more or less top-octave augmentation. It may need more output in the bottom shelf, or maybe a little less. You may need to use impedance compensation, or you may not, depending on the horn and driver, whether or not it has impedance peaks above crossover. But in general, this is what we're looking for, which is appropriate for any constant directivity horn that doesn't have excessive ripple:
|
|
|
|
| Re: 4Pi crossover study [message #77212 is a reply to message #77210] |
Fri, 19 July 2013 13:55 |
|
Wayne Parham
Messages: 18787
Registered: January 2001
|
Illuminati (33rd Degree) |
|
|
One more thing:
You and Bruce are focused on the measurement setup, and that's good because you should really start there. I am focused on the crossover and the horns and drivers, because that's the subject we're trying to understand. But really, if you're seeing more rolloff up high than you expect, it is just as likely the microphone or measurement setup as it is the device under test.
Do you have any way to baseline the measurements? Do you have a known source that you can test and see if there is rolloff up high? I expect you would have done that already if available, but it doesn't hurt to ask.
The compression driver itself may rolloff more than you are expecting too. The JBL drivers certainly do, which is why they need more compensation. That's not a bad thing, as it indicates reduced or highly damped breakup. But ideally, we have the output up high and it's smooth. We want to have our cake and to eat it too.
Anyway, try and find a way to baseline your measurements, so you can know whether or not to trust them where rolloff is concerned, or if maybe they're showing more rolloff than is really there. If you can validate the measurement setup is accurate, then try to find the cause of the rolloff in the DUT by swapping it - Change things one at a time until you find the cause.
I'd start with the compresion driver first, if possible. Borrow a different driver if you can, and see how that tests. If that's not possible, try some really basic filters and see how they act. First, try no filter - straight wire to the driver. Then try a basic first order, just a capacitor in series. Then try putting in a resistor shunted by a capacitor in series, the R1/C1 part of the network. This should give peaks where there are impedance spikes, but should also give rising response. Then put a shunt resistor in there to damp the impedance spikes. Build up one step at a time to see what each part does to the system response. Let the documents below be your guides:
|
|
|
|
|
|
|
|
|
|
|
|
Goto Forum:
Current Time: Sun Nov 24 00:17:29 CST 2024
|
Posting Rules
Members
Search
Help
Register
Login
Get a T-Shirt!
