Crossover experts - what's the difference between these two?

[DrWho]

Thanks - that makes a ton more sense too [Y]

[Deang]

Your plot is interesting. Not surprising really. Keep in mind that most speakers are designed with a somewhat downward tilt. Speakers with flat responses tend to sound bass shy and bright. I know you know this already.

Parts upstream (between the autoformer and driver) are scaled based on the impedance of the driver. Parts downstream (between the autoformer and amplifier) are scaled based on the reflected impedance through the autoformer. So, you can move the tweeter cap off any tap and not effect the crossover point.

Fritz, you're thinking what I thought you were thinking. You're making a differentiation where none exists. What you are saying about the second design applies equally to the first. Both networks preclude the ability to move around the taps without shifting the crossover point. Both designs accomplish the same thing using different methods. In the first, the resistor swamps the squawker down to half its rated impedance, so when you drop down on the taps the reflected impedance back through the autoformer is the same as it was when you were on the original taps -- hence the reason you don't have to change the cap value. In that network, the impedance will still fluctuate between 6 and 70 ohms just like the Type E (reactive). In the second, the goal is to create a somewhat constant impedance, with low pass and high pass sections somewhat equal in impedance. The design assumes the K-22-E is 11 ohms at the crossover frequency (700Hz). To get the matching high pass section, and the proper resistor value for the autoformer -- we need to know the reflected impedance with the tap we are going to use. In this case, it's tap 1, and we'll go with 256 ohms (even though it's actually a little lower than that). At this point we bang on our resistor calculator until we find the value that gives us our closest match to the impedance of our low pass section at the crossover point. Using the calculator, we discover that if we parallel an 11 ohm resistor with 256 ohms -- we come up with 10.54 ohms. 11 ohms is close enough here, and we strap it across taps 5 and 0. Now we need the correct cap value for 11 ohms at 700Hz. That value is roughly 21uF. This design presents a benign, stable load to the amp -- and since impedance isn't swinging around all over the place, the transition points are actually pretty much nailed down. Not the case with the first one.

When I said both designs used "fixed" taps, I meant that neither could be used to make attenuation changes on the fly. With either, if you move to different taps -- you are going to have to change some things, and unless you float the ground, you are only going to get 3dB increments.

[DrWho]

I can see the tweeter cap not changing....but won't the frequency response of the squaker change when you move the tweeter around? The 2uF capacitor between the amp and the autoformer is going to see a changing impedance on the load side. Any attempts I make to model the tap change results in a 2dB rise in the squaker response between 2kHz and 4kHz. Tweeter response stays the same, but is reduced in output.

Or am I missing something painfully obvious here? An autoformer is just a transformer right?

[Deang]

Just for clarification -- your comments are in the context of simply dropping the tweeter onto the same tap your squawker is without making any changes to your E2 network?

I've talked to Bob and Al both before about doing that kind of thing. We all agreed that it's probably O.K. to do it, but also don't know exactly happens with the response when it's done. I originally put the question to them because I wanted to try it as an inexpensive way of attenuating my Beymas a bit more when I got them. It seemed to work O.K. Your modeled response is something I don't think anyone here has done before, so your results are a new thing we can add to the knowledge database. Nice work!

[Guest " "]

"Or am I missing something painfully obvious here? An autoformer is just a transformer right?"

It's a single winding transformer.

Works on the principle of impedance multiplication or division, depending on which way you wire it, based on the turns ratio, see sample diagram.

Factors in evaluating autoformers are it's bandwidth (20 hz - 20khz), deviation in db (+- 1db), insertion loss in db (.6db), power rating ( 50 watts), etc

The turns ratio determine the impedance change, which defines the power pass thru profile.

Also keep in mind, if impedance increases, distortion decreases, dampening factor increases, and power decreases.

[Guest " "]

Dean,

I think your saying there is no difference in the two diagrams, thats fine, your views are noted.  Or maybe your saying they are different, but for other reasons, it's not clear to me what your really saying. 

Floating the ground is not the only consideration in determining the effect an autoformer has.  It's simply a matter of determining the required amount of "turns ratios" needed, then identifing the possible combination of taps that will provide the required total turns ration.  This means that there are probally more than 30 possible combinations of tap configurations that can be used to provide predictable results.  

[DrWho]

Just for clarification -- your comments are in the context of simply dropping the tweeter onto the same tap your squawker is without making any changes to your E2 network?

Yes, but I've also just been using 8, 16, 8 ohms for the woofer, squawker, and tweeter respectively so there are some rather large innacuracies in the model. I'm thinking this would be a good excuse to set up a system to measure the complex impedance....and then the filter predictions will be more accurate. And if I'm going that far, I might as well measure the independant acoustical and phase responses too and then just start a design from scratch, limiting myself to the same components.

Right now I'm thinking the easiest mod would be to remove the existing 2uF capacitor and then put a new cap inbetween the autoformer and the squawker. While I'm at it I think I should at least lowpass the squawker, which should also reduce the comb-filtering in the 7-9kHz region, and bring the 4kHz output down. But now you're talking a complete redesign []

I think I will measure the output of the squawker alone first (by just dumping the mic in the mouth). If it's running full range, then it's possible that a lot of the extra output is the result of a lack of lowpass on the squawker. It is certainly apparent that there is some destructive interference in the 7-9kHz range.

Anyways, here's the my model of the E2 network:
(red = woofer, green = squawker, blue = tweeter)

[DrWho]

And then the tweeter on tap2 of the E2 network:
(red = woofer, green = squawker, blue = tweeter)

[DrWho]

Ok, so I updated the plots with actual transformers and slightly better predictions for the impedances of the drivers. It looks like the change on the squawker output is much much less now so I think I'm going to try just moving the tweeter to the other tap.

And if my output in the 4kHz region doesn't improve, then I'm going to try a .25mH inductor in series with the squawker inbetween the autoformer and the driver. Attached is what that plot looks like.

[JohnA]

Reading some old posts on how to tame the
tweeter and squawker on older heresies... and I came across these two
networks. What's the difference?

I
designed them both. The net effect of either change is
essentially identical, though the Type EModified should drop the
squawker and tweeter a hair more due to the resistor. I'm running
all 3 of mine in my HT with the Type EModified crossover because they
sound much closer to my La Scalas that way.

Unbeknownst
to me the Type C crossover is very similar to the Type EExpr , with the
squawker cap adjusted for the change in Impedance. My early Type
H speaker with an EV woofer sounds better than the Heresies do with the
horns knocked down. I'm not sure what I'm hearing, but there is a
brightness to voices, La Scalas don't have.

[JohnA]

Here is the comparative response of one Heresy with an unmodified Type

E crossover (dark Line) and another one with the Type EModified

crossover in it (light line). Since there are 2 different

speakers you are also seeing manufacturing tolerences in the

drivers. i had them measured at a local small speaker

manufacturer, Digital Phase.

[DrWho]

Are those anechoic John?

They look fairly similar to my measurement which is encouraging. Mine look smoother as you go down in frequency because I gated the response (which reduces resolution).

I find it very surprising the that squawker is left to run wide open until it peters itself out. I haven't had the chance to mod my speakers yet, but I wonder if there aren't "cone-breakup" effects happening in the squawker...something that can easily be remedied with a low-pass (but not fixed by reducing the squawker output). Heck, I wonder if that wouldn't fix the entire brightness issue altogether. It's the 4k region that I'm most concerned about...a little extra from 8k and up doesn't usually sound bright.

[tigerwoodKhorns]

Compared to the stock Type E configuration, the squawker's lead was moved from from tap 2 to tap 1 and the tweeter from tap 3 to tap 2 in both cases. I assume this will decrease the output of both by 3dB, but I'm mostly just curious if one configuration is preferred over the other since I'm using a tube amp (Scott LK-48B).

I had these Heresys hooked up to a Yamaha CR1020 and on stands about 6" off of the floor. In the middle of a pretty large room they had great bass response. It really surprised me. Maybe the receiver had something to do with it, but then again, the Scott is a great amp (I used to have one).

One thing to note, the crossover rebuild has about 100 hours top on the new caps. Maybe they need to break in. I rebuilt three pairs of crossovers at once and none were in the main system so they did not get much time.

[JohnA]

Mike,

It wasn't in an anechoic environment, but
the speakers were swept with a sine wave pretty quickly. I doubt
there are significant break-up issues with the K-55-V. The early
-Vs and the -X are rated down to 110 Hz by Atlas and at a pretty good
power rating, so a 700 Hz x-over point should not pose a problem.

[stereohermit]

Also, if I recall, the C network kept all drivers in phase.