Heresy Crossover Mod - Less Bright

[skywave-rider]

I know this thread is old, but it seems the right one to add to.

My K-77M tweeters are too hot, but not the mid. The woofer seems balanced in what it provides. So my question is, what would you suggest I do to attenuate the tweeter alone, by what looks like, potentially, about 8dB? If I read this thread correctly, the mod developed here attenuates both mid and hi.

I measured my Heresy 1.5 (k22k woofer/k55mid/Crites and K-77
tweeter/E-2 crossover) at 1 meter and 3" nearfield. The speakers have
the angle back stands, so they are close to the floor at the standard
tilt up.

"Normal" connection follows the HF connection in the schematic posted below.

If you can't seem the full RTA image, you could go here: Click to go to Audio Karma thread

[Invidiosulus]

The different taps on the autoformer each offer a different level of padding each in 3dB increments IIRC.

I really like mine with both the midrange and the highs attenuated.

It sounds very balanced to me but I don't have any RTA stuff to prove to my eyes what my ears enjoy.

[skywave-rider]

The different taps on the autoformer each offer a different level of padding each in 3dB increments IIRC.

I really like mine with both the midrange and the highs attenuated.

It sounds very balanced to me but I don't have any RTA stuff to prove to my eyes what my ears enjoy.

Thanks. Yes, I would like to try using the taps on the autotransformer, and was wondering if there is a good way to do it for the tweeter only. I can figure a fixed L-pad, but I note that some here believe there are superior methods to doing that.

It's easy to see and hear that the K77 is much too hot. That's why I bought the Crites tweeter well before I had RTA capabilities.

[]

[Invidiosulus]

Unless I'm mistaken(happens now and then) you could just move the tweeter negative and leave the squawker alone.

That should do as you want, although just dropping it one tap lower may not give you the attenuation you want.

[skywave-rider]

Unless I'm mistaken(happens now and then) you could just move the tweeter negative and leave the squawker alone.

That should do as you want, although just dropping it one tap lower may not give you the attenuation you want.

Referring to the Albright mod, keep the 11 Ohm, 20 Watt resistor from first cap to ground? And I notice he has changed that cap from 2uF to 21 uF. What of that? I see he says to keep the tweeter cap at 2 uF.

Thanks!

[Deang]

You kind of have to know how autoformers work for it to make any sense. Since I'm pretty burned out on explaining it you should try the archives where I think no less than six of us have explained it a dozen times or more. Trust me, it's almost better if you don't know!

The 11 ohm resistor is actually strapped between taps 0 and 5 on the autoformer. Tap 5 is the input tap. The resistor changes the reflected impedance through the autoformer back to the amplifier, so you have to raise the capacitor value from 2uF to 21uF to keep the crossover point the same. There is no impact to the tweeter leg which comes off the output side of the autoformer, so the cap value stays the same.

[skywave-rider]

You kind of have to know how autoformers work for it to make any sense. Since I'm pretty burned out on explaining it you should try the archives where I think no less than six of us have explained it a dozen times or more. Trust me, it's almost better if you don't know!

The 11 ohm resistor is actually strapped between taps 0 and 5 on the autoformer. Tap 5 is the input tap. The resistor changes the reflected impedance through the autoformer back to the amplifier, so you have to raise the capacitor value from 2uF to 21uF to keep the crossover point the same. There is no impact to the tweeter leg which comes off the output side of the autoformer, so the cap value stays the same.

Thanks, Dean. I'll dig into it. Can't promise myself it will sink in though.

To re-iterate, I only want to bring the tweeter down, not the mid. I assume others have tried to do what I want.

I have a Crites tweeter which is what I normally use. I'm just playing with the pulled K-77 to see what I can learn. But if the mod makes it a burden to switch the Crites back in, I might just make an L-Pad I can cut out. Dunno. We'll see.

Thanks for the info.

[WMcD]

There is a lot going on as far as "impedance."

You have to consider the entire system, which is itself made up of components which have sub-components.

Amp <--> crossover <--> voice coil <--> mechanical system <--> acoustic system.

I've typed that with arrows pointing both ways because energy flows in both directions.

Actually we can break down the acoustic system into:

back cavity <--> (diaphragm) <--> front cavity <--> throat of horn <--> horn <--> load at mouth.

The mechanical system has the compliance of the suspension and the mass of the diaphragm. It is also the link between the force from the voice coil and it causes, because of motion, the acoustic pressure and gas flow in the acoustic system.

The voice coil has the length of wire and the magnet. This is the link between the electrical current and the movement or force on the diaphragm.

The crossover it what we are most familiar with but it has at least and L or C.

= = = =

The links to Olson shows the impedance looking into the throat of the horn. He shows that lenght of horn and mouth size will cause it to vary. The wiggles are because of the mouth reflection.

The link to the JBL sheet shows the electrical impedance looking into the electrical system. Note that it is the performance with a plane wave tube. That is a pure acoustic load without wiggles in real impedance and no reactive components (as shown in Olsen). But note there is a peak. That is caused by the mechanical resonance of the mechanical system. We see the same sort of thing in a sealed box system or a raw driver.

It not uncommon to see the electrical impedance when a horn is attached. There are more wiggles.

Generally the lowest peak is the effect of the mechanical system. When it moves there is little mechanical resistance and that translates to a light electrical load, which is to say, a peak in electrical impedance.

The ones farther up the graph are the effects of the horn. Consider that the Olsen peaks (acoustic) are affecting the electical load just like the mechanical system does.

= = = =

The finite length acoustic horn is difficult descibe. However, it is related to electical transmission line theory. Even that is difficult to describe.

Basically: (is this hopeless?)

We must realize that a pure resistance follows Ohms law. R = V/I. Suppose you have a 50 ohm resistor. If you apply a 50 volt pulse, the current is 1 amp. The shape of the voltage pulse and the current pulse are identical. Nothing to do about it. The mindless resistor simply makes it so within the resistor.

- - - - -

Source ===(long wire or acoustic tube) -===== load.

The next question is what happens with a very long wire pair. Suppose we put a 50 volt pulse from a function generator into it which is so brief in duration that it can't reach the other end of the wire before the pulse stops. Is there any current drawn? Scratch head.

One temptation is to say there is no current drawn but the voltage pulse (without current) goes down the line where there is a resistor on the end, which draws current from the voltage pulse which arrives. Of course Ohms law applies and there is current equal to 1/50th of the voltage pulse.

But wait!! How the heck can we have current going into the load resistor if there was none at the source? Something else must be going on.

Actually there is. Some current is drawn from the generator. If we measure that, we wind up with an I. This is the charactristic impedance of the wire pair. It arises from the L and C in the wire. We have a V and I and thus an R equivalent, even though (gasp) there is no R like that in the wire. This also called "surge impedance." By analogy, this is the acoustic impedance looking into an infinite horn, or a finite length horn for a short time. We have the voltage pulse and its fellow current pulse.

Suppose the V/I ratio of the pulse in the wire is 50. When the pulse gets to the resistor, the resistor is perfect load for that pulse. It is happy and absorbs it, entirely. We have to look at the node. There is voltage and current coming into the resistor, and the resistor abosorbing those in the correct ratio.

But suppose we have a 100 ohm load? Or a 25 ohm load. In each case, there is a mismatch of the ability to absorb power. I get a little hazy in the armchair, here, about book-keeping. Nonetheless, this is a real train wreck of Kirchoff's rules. The voltage at the node is what it is. Some amount of power comes in from the line, the resistor absorbs some but not all. How do we make up for the imbalance in the book keeping.

The answer is that some current going toward the load (into the node at the load) is negative, meaning that it is going backwards, towards the source, or reflected. Whether the voltage is positive or negative in this reverse direction depends on whether the load is too big or too small. Different load resistances.

Yikes, now we've got a bigger riddle. Remember that we put that pulse into the line and shut it off. We wait a while and now, at the source, we have the reflected energy coming back at us. It will have positive or negative voltage depending on the load resistor.

So, you say, what does this have to do with the impedance looking into the transmission line? We have to look at it long term.

Suppose we put another 50 volt pulse into the line, just as the reflected pulse of voltage is coming in. Now, for the purposes of this thought experiment, let us imagine that the reflected pulse is 50 volts, plus. (Not exactly our condition in the example.)

But this is like hooking up a 50 volt battery in parallel with other 50 volt battery in parallel with the polarities matched. How much current flows? Answer, none. We have no current. This is like trying to charge up two 12 volt auto batteries with jumpers when both are already charged.

So let's look at "impedance" to current flow. Again, it is necessary to note that impedance at a node is not a" thing," it is just a V/I ratio. We have 50 volts with no current and the R=V/I ratio is infinite. Let's call that very high impedance or R. No current flows. This is like a peak in throat impedance.

Suppose the reflection comes back at a negative voltage. Now we have, by analogy, batteries connected in parallel but backwards. (Mis-connected jumpers) Another way of looking at this is the two batteries being in series (note backwards parallel is a series circuit), and therefore with a shorted out load. All of sudden, we have a lot of current. In the V/I ratio, we have a low number because I is very big.

I know that is a long row to hoe. But it shows that mismatches at the end of a transmission line (horn) can cause variation in impedance at origin because of the reflection of simple pulses.. A much more complicated subject is timing. When we have sine waves of many wavelengths and finite lenght transmission lines (like a horn), the phase (delay) issues gang up in odd ways.

Olson's graphs are showing two things. One is that a bad mouth causes reflections and variation in throat (sending end) impedance.. The other is that the length of the horn (delay of reflection) also affects the throat impedance.

It was easy for us to say that the reverse travelling pulse gets there just in time for the outgoing pulse. But in sines they are going plus and minus and very complicated.

This is the real story but may be less than illuminating.

Wm McD.

[DrWho]

Here's the results I got when just dropping the tweeter tap on the autoformer:
http://forums.klipsch.com/forums/t/85571.aspx

Sadly I never got around to finishing the project, but in a couple weeks I should have some more time on my hands (too busy right now trying to find a job)

The quick summary in case you don't want to dig through that other thread is that lowering the tweeter didn't fix all the brightness issues. The squawker might need to be attenuated at the top of its response too.

[Erik Mandaville]

A more common and widely accepted (e.g. Altec, JBL, other well-known companies -- including Klipsch) way of going about the business of driver attenuation and balancing is to use variable L-pads. If nothing else, it reduces the potential for confusion in dealing with the impedance variations and ramifications associated with the autoformer.

The tweeter (and squawker for that matter) can be removed from the autoformer and connected directly to the input of the amplifier by way of values of capacitance chosen for the impedance of the driver itself and the desired crossover point. Since the behavior of the drivers in question is well-known, the calculations are straight forward. I found an improvement in performance of the squawker by designing a true band-pass with resistive attenuation.

Something interesting to consider, though: What is the order of tweeter filter? a 6dB/octave roll-off has been indicated many times. What, though, is the value of required capacitance for 1st order 6kHz network with an 8ohm tweeter? The answer to this question isn't found in the original schematic, and that has certain implications. What are those?

An easy way to reduce tweeter output is to disconnect it from the autoformer; slightly increase the capacitance in series with it; and install an 8ohm variable L-pad just ahead of the driver. That will provide a constant 8ohm load regardless of the level of attenuation, as well as maintain the 6kHz crossover point (or lower if using a tweeter that can handle it). Bob's tweeter setup works very well with a 4kHz third order network. Just adjust the output to the desired balance in relation to the other drivers. Much finer adjustment and tuning is available through the use of a pot here.

Erik

[Erik Mandaville]

And if the squawker output is too great once the tweeter has been attenuated, a variable pot can be used there, too. I'm doing this very thing with the Heresy center channel I've been gathering parts for. I just have to wind the coils and find a woofer somewhere....

Erik

[DrWho]

Since the behavior of the drivers in question is well-known, the calculations are straight forward.

Are you kidding? The varying impedance and varying phase of drivers makes it extremely difficult to "just calculate" a filter that will work as intended. Every professional loudspeaker engineer I've run into usually goes through at least a few iterations of crossovers before they start listening to it...and even then they're gonna make changes to alter the voicing.

Btw, keep in mind that your L-Pad is going to dissipate a crap load of power at the high impedance points of the driver.

Also keep in mind that resistive attenuation can have the effect of changing the frequency response of the driver - very similar to what happens with high output impedance amplifiers.

I'm not saying it can't be done, but resistive attenuation in general is nowhere near as simple as adjusting taps on an autoformer.

[Erik Mandaville]

I'm not kidding in the least. What do you see in the way of impedance equalization circuits or resonant peak and notch filters in, for example, a type A Heritage network? It's entirely possible to create a network that not only works as well as the 'A,' but arguably better. What 'special provisions' were made in the type A? As far as I can tell, not any in terms of corrective circuits. What was used to design it was based on the same parameters I used to make my own -- driver impedance and desired crossover frequency.

Right. Resistors do dissipate heat, and we know that the late Paul Klipsch was opposed to them and chose to attenuate by way, at least in the example of the type 'A' I'm using, of intentional impedance mis-matching.

It's not just MY L-pad, BTW. The majority of well-known speaker makers that provide a means of user-ability and preference in terms of balancing driver outputs use adjustable L-pads. It's not only slightly easier than altering outputs on a multi-tapped choke, but much, much easier. In your view, is it easier to turn a knob to adjust gain, or shut down the system, take off the back panel, unsolder-reposition-and-resolder autoformer connects, and then alter values of necessary capacitance to compensate for the resulting reflected impedance, and then put the back panel back on, power up the system, and then listen again -- perhaps only to find the the change made on the autoformer was too course, resulting in too much attenuation or gain.

I've done both, and find that turning a knob much more straightforward. In other words, a 3 dB drop in output will be a 3dB reduction regardless of whether an autoformer or resistor (or pair of resistors or a variable L-pad) is used as the tool. I would just rather do it while listening to the result, and prefer the far finer control of a variable L-pad.

Erik

[Erik Mandaville]

"I'm not saying it can't be done.."

That's good, because it can be done and isn't very difficult to do.

Erik

[skywave-rider]

Thanks, Gil.

What you summarize is very interesting, it connects acoustic/electrical systems for me very well and is much appreciated.

Please note that the Olson and JBL links you refer to do not appear on your post.

Let me test my assumptions now to see if I understand...

If I attenuate more than stock at the autoformer, I am cutting voltage, increasing current; equivalent to reducing impedance. Couple that with changing the inductance, which requires a change in cap value to prevent the crossover point from shifting.

The resistor across autoformer 0 and 5 gives attenuation of both mid and hi.

That resistance in parallel plus the autoformer's impedance in series act like a fixed, though perhaps frequency dependent, L pad for both mid and hi.

I am still collating reflected impedance....

I have little knowledge, that makes me dangerous.

[skywave-rider]

Mike and Eric:

Thank you both for your thoughts.

Mike, I looked at your measurements.

Observations:

I have to agree, I hear comb filtering which is probably the worst aspect of the Heresys, or any speakers, for me.

My K-77M tweeters are much hotter than yours.

Eric, I will look at your bandpass design, which would seem to solve some of the issues.

If you'd suggest a particular thread with your most recent work, I'd appreciate it.

In the mid term I hope to get Soundeasy software, which will help me design networks and increase my measurement capabilities.

What to do right now? Well I'm learning....

[Deang]

Apparently Gil is still up to the task of offering detailed explanations! Great post.

Skywave-rider: I don't think the tweeter is as hot as you think. Try moving the 2uF cap for the tweeter from the tap that it's on to the same tap the midrange is attached to (from 3 to 2).

btw, it's nice to know how and why something works, but it's just as nice to know that it just does. Don't get yourself all tied up trying to get your head wrapped around this stuff. Some people have done that, and still ended up right where they started -- and most find the simple thing usually works best.

[skywave-rider]

Apparently Gil is still up to the task of offering detailed explanations! Great post.

Skywave-rider: I don't think the tweeter is as hot as you think. Try moving the 2uF cap for the tweeter from the tap that it's on to the same tap the midrange is attached to (from 3 to 2).

btw, it's nice to know how and why something works, but it's just as nice to know that it just does. Don't get yourself all tied up trying to get your head wrapped around this stuff. Some people have done that, and still ended up right where they started -- and most find the simple thing usually works best.

Hey Dean, thanks again.

Well I think this is your subtle way of saying "you just don't understand..." LOL

Perhaps.

I'll try moving the tap. The K-77 looks (as posted) and sounds, real hot. 3dB would sure help though. I'll post back with that.

In the long run, a new crossover designed with a bandpass mid with crossover points selected to compliment horn directivity in the vertical and horizontal dimensions, (something else I'm learning about....), if possible, would be fun to try after I get better measuring abilities. And there are the other projects which never get done. The 2 way Econowave over at AK has to get done first.

Econowave

[kg4guy]

"I'm not saying it can't be done.."

That's good, because it can be done and isn't very difficult to do.

Erik

I agree Erik[Y]

[Erik Mandaville]

"ll try moving the tap. The K-77 looks (as posted) and sounds, real hot. 3dB would sure help though. I'll post back with that.

In the long run, a new crossover designed with a bandpass mid with crossover points selected to compliment horn directivity in the vertical and horizontal dimensions, (something else I'm learning about....), if possible, would be fun to try after I get better measuring abilities. And there are the other projects which never get done. The 2 way Econowave over at AK has to get done first. "

The Econowave project looks interesting, Skywave R. A true bandpass with this driver is not difficult to do, as you seem to understand. Just as there are minimal passive parts in conjunction with the mid and HF drivers in the Klipsch E and A networks (again, for example), so does a first order bandpass for the midrange require only two: an inductor and capacitor in series, the values of which are determined by the upper and lower crossover points. The autoformer is just one way of attenuating a speaker; the more common approach as is even mentioned over at the AK site in relation to the Econowave project, is to use a resistive L-pad. I hope someone will correct me if I'm wrong, but it's been my understanding that the current Heritage networks no longer use the autoformer. If so, that's neither a negative nor a positive thing. A Jubilee passive design that's shown up here in the past also uses a single fixed resistor of a few ohms. They work, and do not in my experience, degrade the sound. One should of course use whatever one feels most comfortable using, though.

For your Heresy, if you didn't mind making a slight modifcation to the back panel (which you may not want to do), it would be very easy to install an 8 ohm variable L-pad for the tweeter as I mentioned above. You seem comfortable with the balance between the midrange and woofer, and thus it would work fine to take the tweeter off it's connection to the autoformer and instead connect directly to the input. It does not need to have to be associated with the input capacitor between the amp input and autoformer in order to work properly. However, if I remember right, the capacitance needed for 8ohms and 6000 Hz is just over 3 uf, so to maintain the crossover point, you would need to bypass the 2uf now being used with another cap. If you wanted to do this, I have two nearly new 3uf caps (Solen) you can have. I probably have a couple of .22 or .33 uf caps I can solder in parallel with them. The L-pad, which is technically not really part of the crossover, is inserted between this crossover and driver terminals (which I'm sure you already know).

You would need to drill a hole of about 1" in the center of the cabinet back, probably close to the top. The L-pad can in turn be mounted on a small piece of masonite with a hole drilled in it for the shaft. The mounted L-pad shaft is inserted through the hole in the back of the cabinet until the masonite or acrylic plastic or plywood (which you chose to use) is flush againt the inside back of the panel. It can be glued in place with silicone rubber and set to dry overnight. A knob is attached to the L-pad shaft, and you can then adjust on the fly until you get the balance the way you want it. If you have a good idea of the actual drop in dBs you want, you can even experiment with a fixed L-pad.

...Or, of course stick with the autoformer and experiment with that! I only mention the above as something you can try. I played around with a 12dB/octave/reverse polarity filter with my Klipchorns for awhile, and in that case just installed the variable pad on the crossover board itself -- easy to adjust since the back of the Klipschorns is easy to get to. See below. With a 1.5 watt OTL amp the results were not pleasing, but with higher horsepower it worked fine.

Good Luck!