New hardware

[John Warren]

Response plots of the top section.

The setup is 96kHz sampling rate of white noise. Plots are FFT. Microphone is about 7 ft from the front face of the Klipschorn bass unit and on the midrange axis.

First plot shows background noise referenced against system signals.

Second plot is system signals, blue is composite, purple is Faital and green is Selenium + B&C.

There is a response dip that starts at 2kHz to a little over 3kHz.

[John Warren]

The response above is compared to a plot on the ALK eng website showing a network ESN 5800.

I scaled my plot so the two are about on the same references.

Look similar.

[John Warren]

Gated MLS of Faital tweeter setup. Couple of mid levels shown.

[Tarheel]

So Al is sort of back

[Schu]

Al's here!!!

[Arkytype]

I've posted a schematic of your "autotransformer based crossover network" based upon images on your web site.   an hour of your life you'll never get back. 

 

Naw, it only took a minute or two to recognize a Klipsch knock-off and sketch it. Just so I didn't feel like I was wasting my time, I rotated the tires on my car at the same time.

 

Lee

[Marvel]

It's not a full moon, is it?

[Thaddeus Smith]

I love to see so many people with strong technical smarts having a coherent and civil discussion indicative of their age and maturity.

[babadono]

It's not a full moon, is it?

The dogs (and coyotes) are howling in my neighborhood

[Arkytype]

John,

 

First, I wrote: I won't start by criticizing your numerous grammar or spelling errors in your forum post and web site; instead, let's start with how your "top-section baffle insert" is going to sound based upon your published data.

 

Then you wrote: ...from the get-go you made it clear that you were really in this for the fight, nothing more. So from my perspective why bother trying to engage in a dialogue with you.

 

When someone with (or without) a PhD posts misspellings and butchers the King's English, I immediately lower my expectations and hold their credibility suspect. As your company purports to offer "TECHNICAL WRITING AND DOCUMENTATION SERVICES", perhaps you should consider how others view your credibility based upon the above criteria.

 

Up to this point, I'm not "in this for the fight"; just offering helpful advice.  

 

The "autotransformer based crossover network" you are hawking is flawed. It is flawed because the Klipsch crossover network from which it was copied is also flawed. The chief culprit is the autotransformer (I'll use the preferred term autoformer from here on); or rather how the autoformer is used in the circuit.

 

The use of an autoformer is an elegant, low-loss way to attenuate the level to, say a squawker driver without the attendant issues of using a fixed resistive pad. For those who may not know, an autoformer is a type of electrical transformer in which the primary and secondary share a single winding. The source (amplifier side) is connected across the ends of the single winding and the load (squawker driver) is connected to one of the winding's end points and a tap somewhere in between.

 

An autoformer can be misused by the unwary or inexperienced crossover network designer who thinks that all you need to do is connect the input of the autoformer to the amplifier side and the squawker driver to the appropriate tap on the load side and, voila, levels are matched! What's not to like? What's not to like is that the amplifier sees a load impedance change with each tap setting.

 

The voltage change (up or down) through an autoformer is proportional to the ratio of the number of turns the source is connected across, to the number of turns the load is connected across. That is, the voltage ratio equals the turns ratio. However, the reflected impedance seen by the amplifier changes as the square of the turns ratio!

 

Here's an example of exactly how the load impedance can swing wildly just by changing taps:

 

Let's take the venerable 3619 autoformer used in the Klipsch AA network and connected to the K55V 16 ohm driver. If you set the 3619 taps to -6 dB (3 and 0), the amplifier doesn't see the K55V as a 16 ohm load; it sees a 64 ohm load! Set the taps to -9 dB (2 and 0) and the load impedance jumps to 127 ohms! If you are running a SET amplifier, good luck driving that load!

 

While I used the Type AA network in my example, your "autoformer based crossover network" suffers from the same impedance swings.

 

The solution? Hint: It's a passive device with axial leads and has a nominal value of 10 ohms. Yep, you just put that "swamping resistor" across the input to your autoformer and those pesky impedance swings are reduced to a load any amplifier can deal with.

 

So, John, for less than a dollar, you can turn your flawed copy, er design, into a more competitive product. But then there's that frequency response anomaly to deal with. Sorry, can't help you there!

 

Lee

[Arkytype]

 

Hello All,

 

Al Klappenberger has written me an e-mail which does a better job explaining how to design crossover networks better than I could ever do. So, for those of you who have missed Al K., here (unedited) is what he has to say about John Warren's "autoformer based crossover network":

 

Lee,

 

I'm sure you have noticed that I have been following this thread. I am writing you to help explain what is going on here. JW's web site states:

"The autotransformer provides for the ability to change the midrange output. The so-called "swamping" resistor (or shunt resistor) across the input side of the autotransformer winding is not needed in the design."

 

The acoustic plots he provides clearly show what has been pointed out by many people on the forums, and that is, you can NOT move the autotransformer taps without the electrical crossover moving on the Klipsch networks. JW has his autotransformer wired exactly as it is in the Klipsch AA network: before the squawker filter (which the AA does not have) and after the single capacitor of a 1st order 400 Hz filter (13 uF in the AA).

 

What is not often noted is that moving the tap also causes the passband to tilt. This is shown on JW's plots by the fact that the curves of different tap settings move apart as frequency increases. It also shows the attenuation moves in the opposite direction below 400 Hz. It actually goes UP slightly with increased attenuation settings as the high frequencies move DOWN!! The reason is that the electrical crossover moves down allowing more energy at 400 Hz to get through to the driver. There is NO innovation here but rather a clear example of why you can NOT move the taps without adding a parallel "swamping" resistor to restore the impedance reflected through the transform back to where it was.

 

I have attached a computer simulation study to illustrate the effect. The green curves show the frequency response and impedance seen by the amplifier in JW's network as closely as I can duplicate it. This is with the autotransformer set at -9 dB and the single capacitor selected to yield a 400 Hz electrical crossover. That is the small white "+" marker showing 400 Hz at 3 dB down. t The lower traces are the impedances seen by the amplifier.

 

The yellow plot is after moving the autotransformer tap to -12 dB. No other changes. Note that the level above 400 Hz goes DOWN and the tilt becomes obvious. The level BELOW 400 Hz actually goes UP!

 

The cyan trace is with a "swamping" resistor added across the autotransformer to reduce the impedance back to where it was. The passband frequency response is now flat and "shelved" like it should be.

 

JW's acoustic plots show that the part values he used were picked to allow the 400 Hz ACOUSTIC crossover to stay in one place. This requires a HUGE amount of effort picking part values or, more likely, just plain luck just to avoid using a 50 cent 10 ohm resistor!

 

Another problem here is that the lowpass filter connected to the DCM50 driver is designed for 8 ohms. A 16 ohm driver could NOT be used without totally upsetting the response. With the autotransformer wired between the filter and the driver and a 10 ohm resistor across the autotransformer, any driver impedance or sensitivity can be used simply by moving the tap. The 8 ohm filter will never see a load greater than 10 ohms. There will be no passband tilt and you will have a correct "shelving" adjustment.

 

I suspect the real reason a special autotransformer was designed is because the original 3619 I used for years is no longer made. The new 3619-ET is available from Universal Transformer only to myself and Dean Wescott (Deang). Likewise, the 3636 is only available to Bob Crites. All of these autotransformers were specified to be bifilar wound.

 

BTW: You have my permission to use anything in this letter, in part or in its entirety, in any way you please.

 

Respectfully,

 

Al Klappenberger

 

So, there you have it from the master. Now let's have some reasoned discussion, shall we?

 

Lee

[John Warren]

<SNIP>

Edited October 14, 2015 by John Warren

[djk]

"I figured that the shunt resistor was the ALK approach and felt to use it would be somewhat inappropriate to use a design feature that was developed by another."

 

Just for the record: It was my idea, and ALK gave me a bit of grief when I first suggested it.

 

Of course it was not an original idea, JBL used it eons ago, and I have posted links to those old networks many times.

 

http://www.jblproservice.com/pdf/Theatre%20Series/4674.pdf

 

Note the level taps on the inductor and the different resistor for each tap (so you can use the same cap value).

[John Warren]

Thanks for setting the record straight.

There's really just two settings that I seem to go back and forth on and there within 1dB. For the small changes like this it seemed to make no significant difference. What can I say? Perhaps life is better with the shunt resistor. Dean provided me with his autotransformer a year or so ago and we measured the behavior with and without it. The electrical performance of the net, measured at the driver terminals, is better "behaved" than without it, audibility questions aside.

With three suppliers sourcing autotransformers from the same shop it seemed a bit odd that no one else was in the market. I needed a one with 0.5dB increments and the supplier I sourced them from had good pricing. Bifilar wound too. These are bobbin wound, very low cost. They were wound for another job, I've got a bunch to play with so need to make use of them.

[John Warren]

I've got a bunch to play with so need to make use of them.

Also, I might make the autotransformers available for sale at a "smidge" over my costs.

Edited October 14, 2015 by John Warren

[Tarheel]

Mr. Klappenberger.....you have now used 90% of your allotted emojis for the new forum software.  You may want to switch to a language of your choice.  Sorry but sign language and braille must go through the special decoder and will show as a delayed response. 

 

BTW....still using your xover design for my LS/Altec system.....now 9 years old and doing fine

[spezjag]

I love to see so many people with strong technical smarts having a coherent and civil discussion indicative of their age and maturity.

Hear, Hear!

[CECAA850]

Mr. Klappenberger.....you have now used 90% of your allotted emojis for the new forum software.

 

It's like Harpo posting

[Deang]

"If you set the 3619 taps to -6 dB (3 and 0), the amplifier doesn't see the K55V as a 16 ohm load; it sees a 64 ohm load!"

This would be for the Cornwall. For the Klipschorn, the correct settings are 4 and 0 (-3), with a reflected impedance of 32 ohms. However, I believe this to be an oversimplification. For example, the Heresy squawker sits on tap 2, which is 128 ohms, but an impedance plot of the loudspeaker shows the impedance climbing to around 70 ohms.

In real life, we have two things; 1) we have a loudspeaker with impedance fluctuating with frequency, and 2) amplifiers that are designed to deliver power at all frequencies, regardless of impedance (with the exception of a dead short). This idea that amplifiers are designed to "work best" at 8 ohms doesn't make much sense in light of what the impedance is doing as a loudspeaker is playing. A loudspeaker is not an 8 ohm resistor, and amplifier designers know this. They do not design for an 8 ohm load resistor -- they design for a complex system. We can push the autoformer off to the side for a moment and consider that many modern loudspeaker designs move between 3 and 20 ohms (or more) using only resistors in their networks.

I prefer the sound without the swamping resistor. I'm guessing most everyone here started with a Klipsch loudspeaker with one of PK's crossovers in it, and liked the sound just fine. I happened to find that if I rebuilt the crossovers with vintage type parts, I was able to get the Klipsch house sound that everyone kept yapping about when I joined the forum. It sounds good.

I agree with John that the DCM50 can easily handle the power likely to be tossed at it with his network.

If John is moving the attenuation in .5dB increments, it's not doing much in the way of moving the crossover point. It's a first order network for crying out loud. I do the same thing, but limit the change in attenuation to 1dB total.

The swamping resistor is an intergral part of Al's designs. However, the ability to adjust the attenuation is more of a convenience feature. I personally don't think it's a requirement unless you like trying a lot of different horns and drivers.

If you have something you know you want to stick with, know how much to attenuate the driver, and know what the impedance is at the crossover point -- you can use impedance scaling to find the primary capacitor value. I would imagine this is what John did, and it works perfectly fine.

I don't like the proxy posting while Al plays the cheerleader. Seriously?

Edited October 14, 2015 by Deang

[rbrown0678]

Dang appreciate your last post, separating the fluff.the flawed design by pk which most here are seems to like. Kudos for your real world approach, remembering if it tests good and it's performance is good this is only to expected.

The anomaly attracts attention, perhaps a set sent on a trip will determine it's capabilities.