Bob, Dean, and Al... x overs

[Deang]

'UT' is Universal Transformer. I buy mine from BEC, as well as the wood I use. I think that's one of the reasons Bob can build for less than me -- I'm sure he's marking that stuff up %300. He's a good guy, just PM or email him at bobcrites@centurytel.net -- he'll get you all set up.

Colin, Bob doesn't jerk around with the circuits, in fact, even his inductors are exact recreations of the orignals, all done through Universal Transformer, which used to have the contract with Klipsch to supply the parts.

I, OTOH, like to play around a little. I prefer the steel laminate Erse Super 'Q' from Partsexpress for the woofer, and really like the Janzten inductors for the tweeter (in the AA). I also build John Albright's DHAxover, which is a constant impedance design, incorporates a bandpass filter for the squawker (electrical cutoff as opposed to acoustical), and a nice fixed L-pad for the tweeter. I like this design not only because it sounds better than the Type A (cleaner), but also because it's friendlier to tube amps, and I can tweak it to suit a person's listening habits by adjusting the resistance of the l-pad. It's a sweet design.

[Marvel]

Dean,

Where are you putting the L-pad in? Instead of the 65 ohm resistor he has in his schematic, or has he made another one?

Marvel

[BEC]

Well, I am back in town after an almost 600 mile trip today. I had to quickly leave this morning and forgot to add one thing to my post above on crossover building and rebuildiing.

When I first messed with crossovers in an attempt to make a set of 1979 Lascalas sound more like my 1990 Lascalas, I found that just replacing the caps in the old AAs with good ones got them most of the way there. I did, after that, do a lot of experimenting with different brands of caps and different types of inductors. For the caps, I just went with a good grade of industrial caps which also seemed to be the most consistent cap to cap of any I have tested. For the inductors, I tried the I-core and the air core and the original iron core type PWK used. I could see no advantage to any really and one of the replacements (the air core) seemed to give me somewhat strange results on the spectrum analyzer. I could not really see (on test equipment) or hear any difference between the I-core and the original inductor, so I went with having UT resurrect the original for my crossovers.

I have a slight preference for the type AA crossover in my normal (pretty loud) listening level. I would slightly prefer the Type A crossover if listening at low levels. That is why I decided to make all my AA crossovers that I build and also the ones I rebuild easily convertible to Type A to give customers the opportunity to hear each version.

Bob Crites

[Dflip]

Dean, thank you for the info on where to buy the Universal transformer, but what is the advantage of it over the T-2A? I continued to use the existing part when I redid the rest of my crossovers. Better just because they are newer and more accurate, or just upgrading while you can?

Don

[BEC]

Don,

There are really two autotransfromers now that will replace the T2A. The "original" replacement is the 3619 built for Al Klappenberger by Universal transformer. It is a technically better unit than the original T2A and has additional functions related to Al's crossovers he builds. The other is the 3636 which is built for me by Universal Transformer. It is also technically better than the T2A and has additional taps which allow it to be used to replace other autotransformers that Klipsch used and that are not available any more. I don't think either are enough better technically to make an audible difference in a crossover, but each expands upon the T2A into other areas of usage.

Bob Crites

[colterphoto1]

edit

[JohnA]

cjgeraci,

If Dean did what I think he did to the DHA design, common sense and a 200 watt amp should not hurt the tweeters, but remember the common sense.

[Olorin]

Dean, thanks for the pointer on the autoformer.

This is just an open question and expands on what was already asked -- is there a good reason to replace the autoformer on a thirty year old speaker? I mean, does it wear out and drift out of spec the way a capacitor does/can? Or is it more of a "shiny" thing?

I have nothing against shiny, mind you, but if it's only going to shine on the inside of the speaker I'm cheap enough to just leave it alone.

[Al Klappenberger]

Guys,

On the subject of transformers, there's a new one coming along. This one is being designed for me for use as a tweeter attenuator in place of the L-Pad I am using in my ES5800 network. It will have taps for 2,4,6 and 8 dB but will have other ratios by going between taps. I haven't worked that out yet. It was to handle 10W of power above 6Khz but it has 10 times the winding inductance of the 3619 and a flat response over the full audio spectrum. I could measure no harmonic or intermod distortion at 10W with an 8 Ohm load. I think it's capable of much more. It too is from Universal Transformer. I am also having a PC board made to use it as a tweeter attenuator.

Al K.

[Audible Nectar]

OK, the search function is failing me now, but could you please elaborate on this (Al, Dean, Colin, any and all?).

"Please note that in ALK's technical answer, he achieves a constant impedance; something all amplifiers love, but which flea-powered or tube or both amps need..."

I understand there are less distortions, etc. theoretically if a tube amp is presented with a constant impedance. Obviously this would not apply to the AA (even if recapped). My question is would there be a reason to be concerned with running AA crossovers in terms of reliability, stressing the amp unnecessarily, and so forth? IOW, if it sounds good enough using rebuilt AA's (until proven otherwise), should I be concerned about my MC-30 tube amps in terms of reliability/stressing the amps from running these crossovers, or is this primarily a concern of fidelity?

[Al Klappenberger]

There is more to consider here about "constant impedance" than just what the

amp likes to see and what it performs best into. There is a direct

relationship between the passband response of any filter and the impedance it

is terminated with and what it presents to the source. This is a factor I

haven't point out directly. Did you even wonder what happens to energy that

is lost in a filters passband when it has a 1, 2 or even bigger "glitch" in

it's passband (like the AA has in it's tweeter filter)? This happens even

with "perfect" lossless parts in a computer simulation. The answer is that it

is reflected back to the source. This IS THE IMPEDANCE MISMATCH! It's like

the old question of what comes first, the chicken or the egg. They are the

same thing. The term usually used in the filter world to define it is "return

loss". You want as much "loss of returns" as you can get. There is a direct

relationship between "return loss", "passband ripple", VSWR (Voltage Standing

Wave Ratio) and impedance. The usual standard is: Less than 1.5:1 VSWR, which

is greater than 13.98 dB return loss, which is less than 0.15 dB passband

ripple, which means a complex load / source impedance match in a 8 Ohms

system between limits of 5.33 and 12 Ohms.

What all this boils down to is that if the filter is not constant impedance

it can not have a smooth passband response! The AA network is simply a very

rudimentary design. Its impedance goes as high as 30 Ohms for a VSWR of over

3.75:1! If you build it with the best parts in the world it still has this

major defect! For a manufacturer like klipsch, this is gold because it saves

the money on parts. It's a perfectly justifiable engineering trade-off. For

people who are looking for the ultimate upgrade it makes no sense to me for

him to build a design like that with high priced parts when a better design

could be built with only a few more parts. You are only building two. The

manufacturer is building THOUSANDS! The parts count means a lot to them.

Al K.

[Deang]

EDIT/FYI: I typed this out and submitted it before seeing Al's post.

I understand there are less distortions, etc. theoretically if a tube amp is presented with a constant impedance.

I don't believe the issue impacts distortion levels, only the behavior of the amplifier as it relates to it's frequency response. Tube amps are voltage driven, and the voltage fluctuates with the impedance of the speaker. In other words, some tube amps deliver non-flat response when given large swings in impedance.

Obviously this would not apply to the AA (even if recapped).

Actually, it does apply to the Type AA. The AA starts at about 8 ohms, climbs rapidly to 30 ohms, and then drops back down to about 10 ohms. One should expect just a bit of elevated output from the amp in the area of the highest impedance.

My question is would there be a reason to be concerned with running AA crossovers in terms of reliability, stressing the amp unnecessarily, and so forth? IOW, if it sounds good enough using rebuilt AA's (until proven otherwise), should I be concerned about my MC-30 tube amps in terms of reliability/stressing the amps from running these crossovers, or is this primarily a concern of fidelity?

For the most part, it's a stictly "fidelity" related issue. Well designed tube amps like the MC-30 could care less. If you were running a 3 watt SET amp with no feedback -- I'd tell you otherwise.

[BEC]

We do have quite a long history of original PWK designed crossovers being driven by virtually every type and wattage of amp imaginable. I don't ever recall any comments of amp damage caused by his crossover designs. Constant impedance would be a good a thing, but in electronics, we have to give up something to get something else. I don't think you can make the crossover a constant impedance device without giving up some efficiency. That may be most important to those who use the lower powered amps.

Bob Crites

[Al Klappenberger]

Bob,

"Constant impedance would be a good a thing, but in electronics, we have to

give up something to get something else. I don't think you can make the

crossover a constant impedance device without giving up some efficiency.

That may be most important to those who use the lower powered amps."

NO! This is totally opposite to reality! The very best way to transfer power

is to match impedance. When you have impedance mis-match, you have reflected

power (bad or low return loss). This is DIRECTLY related to efficiency. A

constant impedance network will have LESS LOSS and GREATER efficiency than a

poorly matched one. This is EXACTLY what I just explained in my last post.

The mismatch and the fact that the tweeter and squawker are both working at

the same time (both getting power from them, that is, in parallel) in the AA

and A (and all Klipsch networks) is the reason. This divides the power and

causes mismatch with causes power to be reflected. It is a compromise that

you can get away with but it is certainly no "correct"!

Look at the attached filter plot. It is of a 7 th order, .15 dB ripple

Chebyshev lowpass filter having PERFECT parts. It's in an 8 Ohms system. Note

where the vertical dotted-line marker is located (about 164 Hz). At that

point the loss is zero, the impedance is exactly 8 Ohms, and the return loss

is 56.7 dB (no reflected energy). All the power is going to the load BECAUSE

the impedance is matched. Look at 240 Hz on the plot. This is the opposite

condition. Here the loss is .15 dB and the impedance is about 9.8 Ohms. The

return loss (brown line) shows about 15 dB. Clearly the lowest loss is where

the impedance is matched!

The ONLY excuse for the impedance mismatch of the Klipsch networks is

efficiency of PARTS COST, not power transmission efficiency. This is an

engineering trade-off favoring a manufacturer, not the customer.

Al K.

[Deang]

There is a direct relationship between the passband response of any filter and the impedance it is terminated with and what it presents to the source.

Would you believe I'm still struggling to get my hands around this issue? I have books out the wazoo now, and none of them talk about this.

Did you even wonder what happens to energy that is lost in a filters passband when it has a 1, 2 or even bigger "glitch" in it's passband (like the AA has in it's tweeter filter)?

I don't know what you mean by "1, 2, or even bigger glitch". What do the numbers represent? At any rate, I always figured "lost" energy is lost in the form of heat.

This happens even with "perfect" lossless parts in a computer simulation. The answer is that it is reflected back to the source. This IS THE IMPEDANCE MISMATCH!

I relate impedance mismatches to something that goes on between source components, like an amp and preamp, or preamp and CD player. The thing that gets me with this subject is I don't think any current loudspeaker manufacturer in the industry builds their networks to show constant impedance. However, I did read that KEF did it with their speakers in 80's.

A bit off topic here, but the way I understand it, rising impedance is caused by the inductance of the voice coil, and everything I read says to use a Zobel filter for this. Notch filters are used when the crossover frequency is near the driver's resonance point, and swamping resistors are used when you don't know what in the hell is going on, and you just want to flatten the whole thing out. Seriously though, some might say a swamping resistor is a "crude" way of dealing with anomolies in a loudspeaker's impedance response, since there are methods that can be used to specifically target the problem areas. I think swamping resistors lower the overall sensitivity of the speaker too.

The term usually used in the filter world to define it is "return loss". You want as much "loss of returns" as you can get. There is a direct relationship between "return loss", "passband ripple", VSWR (Voltage Standing Wave Ratio) and impedance. The usual standard is: Less than 1.5:1 VSWR, which is greater than 13.98 dB return loss, which is less than 0.15 dB passband ripple, which means a complex load / source impedance match in a 8 Ohms system between limits of 5.33 and 12 Ohms.

How I wish I understood everything you just wrote, but I only understand enough of it to give me a headache. Worse yet, I need to relate it all to below:

"The source impedance is the voltage "stiffness" of the source. In this case, that's the amp driving the speaker. It's equivalent of the "damping factor". Most amps have a source Zo of about 0.1 Ohms. The filters look very bad from that source Zo. The analysis shows it looking much better from a 6 or 8 ohm source (high damping factor of about 1.0). This is a fudamental error made by amateur filter designers."

EDIT: I can't keep up with your dag gone posts. BTW, if you put an ALK in one speaker, and a Klipsch first order network in the other speaker -- the speaker with the Klipsch network sounds "louder".

[Al Klappenberger]

Dean,

There is a direct relationship between the passband response of any filter

and the impedance it is terminated with and what it presents to the source.

Would you believe I'm still struggling to get my hands around this

issue? I have books out the wazoo now, and none of them talk about this.

*** This is no surprise. It is something I realized by running computer

simulations for years. I later found equations showing the relationships in

filter books. I doubt if you have may of those.

Did you even wonder what happens to energy that is lost in a filters

passband when it has a 1, 2 or even bigger "glitch" in it's passband (like

the AA has in it's tweeter filter)?

I don't know what you mean by "1, 2, or even bigger glitch".

What do the numbers represent? At any rate, I always figured "lost" energy is

lost in the form of heat.

*** Look at the lower corner of the tweeter filter of the AA network. Instead

of a smooth roll off at the corner, it has a "hook" in the response. This is

casued by impedance mismatch. I calle it a "glitch" (a very technical term!).

A bit off topic here, but he way I understand it, rising impedance is

caused by the inductance of the voice coil, and everything I read says to use

a Zobel filter for this. Notch filters are used when the crossover frequency

is near the driver's resonance point, and swamping resistors are used when

you don't know what in the hell is going on, and you just want to flatten the

whole thing out. Seriously though, some might say a swamping resistor is a

"crude" way of dealing with anomolies in a loudspeaker's impedance response,

since there are methods that can be used to specifically target the problem

areas. I think swamping resistors lower the overall sensitivity of the

speaker too.

*** Yes, to a degree. The swamping resistor, as I use it in my networks, does

allow for variations in load by being much lower than the variation and

therefore a huge part of it. The voice coil mainly makes the complex

impedance go inductive rather than just up! Impedance will go up or down or

become inductive or capacitivy reactive due to the total effect of all the

parts in the entire netwrok. It's not just one part, like the voice coil.

Zobel netwrok and notch filters are an entirely seperate subject, and a long-

winded one as well!

The term usually used in the filter world to define it is "return

loss". You want as much "loss of returns" as you can get. There is a direct

relationship between "return loss", "passband ripple", VSWR (Voltage Standing

Wave Ratio) and impedance. The usual standard is: Less than 1.5:1 VSWR, which

is greater than 13.98 dB return loss, which is less than 0.15 dB passband

ripple, which means a complex load / source impedance match in a 8 Ohms

system between limits of 5.33 and 12 Ohms.

How I wish I understood everything you just wrote, but I only

understand enough of it to give me a headache. Worse yet, I need to relate it

all to below:

"The source impedance is the voltage "stiffness" of the source. In

this case, that's the amp driving the speaker. It's equivalent of the

"damping factor". Most amps have a source Zo of about 0.1 Ohms. The filters

look very bad from that source Zo. The analysis shows it looking much better

from a 6 or 8 ohm source (high damping factor of about 1.0). This is a

fudamental error made by amateur filter designers."

**** NOPE! Only a doubly-terminate (or doubly loaded as the M,Y and J book

calls it) will it look better from an 8 ohm source. This is part of what's

wrong with the AA tweeter filter. It is symmetrical ("constant-K" and DOES

work best from and 8 Ohms source, BUT it's being fed from a .1 Ohms source! A

singly-terminated filter is what's needed. Even it needs a mirror-image

highpass filter connected parallel with it to make the impedance a constant 8

Ohms. Otherwise its impedance goes reactive, inductive or capacitive

depending on if it is a highpass or lowpass filter.

EDIT: I can't keep up with your dag gone posts. BTW, if you put an

ALK in one speaker, and a Klipsch first order network in the other speaker --

the speaker with the Klipsch network sounds "louder".

*** OF course! I set the squawker taps LOWER on mine! The loss in my squawker

filter is only about 0.5 dB. Compare the loudness of the tweeters once

though. Mine will be about 3 dB louder than the AA.

[John Warren]

----------------

On 3/19/2005 11:01:34 AM Al Klappenberger wrote:

NO! This is totally opposite to reality! The very best way to transfer power

is to match impedance. ----------------

AL-

You must explain to me. To obtain a linear spl as a function of frequency a loudspeaker(s) cone

must have a constant acceleration, independent of frequency, over the entire bandwidth of operation. That criteria is fact and is a topic that has been discussed for many years in the literature (Small, Locanthi as examples). Small, in fact, has a paper that explains how to preserve a constant voltage transfer in a passive network, the necessary rqmnt for constant cone acceleration. He states (and quote):

"consideration of the electroacoustic behavior of the common loudspeaker drivers leads to the general crossover network design rqmnt of constant total voltage transfer" (Proc. IREE Australia 31, 1970)

That said, how can a constant power transfer satisfy this criteria?

jw

[BEC]

Any resistor in any circuit that performs any function in the circuit converts electrical power to heat.

I think the apparent louder sound when comparing the ALKs to original Klipsch crossovers is likely because of AL's attenuation of the midrange more than PWK did at least in Al's recommended tap settings. To me, ALK type A crossovers definitely produce lower volume in a Lascala than a PWK designed network with the same input. PWK once said "the midrange is where we live" and that "I have spent more man hours of R and D on the midrange than on the bass". After all that, though, some would have us believe PWK got it all wrong.

Bob Crites

[Al Klappenberger]

Bob,

The setting of the midrange (squawker) is really a matter of may factors. Look at the Fletcher Munsen curves for example. These curves say that the setting even depends on how loud you like to listen to your music. PWK didn't allow for the users to adjuste it. I think I remember him referring to it as a "confusion factor" somewhere. He had to pick a level. At the time the AA was being produced Klipsch had no anechoic chamber, so it was probably done in a room or outdoors. So, tap #4 on the T2A got it! I used to set the taps at 2-5. That's my particular taste. I asked everybody to comment in a thread here on the forums a while ago and the consensus seems like x-4 was what most people wanted, so that where I set them now. It's entirely up to you and your situation!

Al K.

[Al Klappenberger]

John,

I have seen you post this entire quote before. I believe it is out of context

jargon that does not apply here. I need YOU to explain how it DOES apply! Who

cares about "linear SPL". Define it! Do you mean a flat frequency response?

"Total voltage transfer". Voltage is just half the story. Power involves

voltage and current in phase. Remember "wattles power" like in a capacitor.

Voltage transfer to where, the air in front of the speaker? I don't think so!

"Constant acceleration", another SO WHAT? All we care about is frequency

response and distortion out and what complex impedance it presents to the

filter driving it. "Acceleration" means an increase or decrease of speed. A

rocket engine burning on a space ship going in a straight like would probably

have constant acceleration until the fuel runs out too! Let the loudspeaker

driver designers worry about stuff like that.

Al K.