Question on Super AA Crossover

[Deang]

Ignorant response deleted.

Edited July 1, 2016 by Deang

[Al Klappenberger]

Erik,

Reading Dean's post got me to realize what you were asking, MAYBE! If you mean simply moving the 2.2 uf cap (input to the tweeter filter) to the other side of the 40 uF cap, the answer is YES, you can do that. It is NOT correct, but in works, but not quite as well. You can't always do it. You can only get away with it on this and similar designs. The actual effect is about a .2 dB increase in the tweeter filter passband ripple. Assuming that you can always do this will get you in trouble later.

Al K

[ajsons]

LOL for reasons I won't discuss!!

Ours is not to question why or else get whacked by the Klappenstick.

Dean,

IKMAO! (I'm Klapping My *** Off!)

[ajsons]

Ours is not to question why or else, you can Klap your *** good-bye!

I can't help it , I'm KOL (Klapping out loud!)

[ajsons]

(I know, Al, one of these days, I'm gonna pay for these!)

[Al Klappenberger]

Guys,

My name is John Smith.

aka Al K.

[Al Klappenberger]

Erik,

I owe you a debt of gratitude! It seems that on some occasions a person with
virtually no experience in a given subject can open up the mind of a so-
called "expert" with 30 years experience in that field! It happens when that
so-called "expert" gets set in his ways and assumes any variation in his
procedures must be wrong!

The idea of moving the highpass (tweeter) filter in the Super AA network to
the input side of the 40 uF cap struck me a totally wrong because of my long-
standing practice of chaining 2-port "diplexers" together to make 3 (or more)
port multiplexer (crossover) networks. I had noticed on at least two other
occasions where moving the tweeter filter input to the common input worked
ok. Both times I shrugged it off a coincidence. It finally sunk in that doing
this is simply using the time-honored technique of connecting a highpass
filter in sereis with a lowpass filter to make a bandpass filter. The books
all show this method for making 3-port crossover networks. My experience at
microwave frequencys showed me many times that this works but requires that
the junction of the two filters be optimized to work properly. This is the
case with doubly terminated filters, that is, filter intended to work between
equal impedances. That would be 8 ohms input AND output. In the case of
crossover networks, the filters need to be singly-terminated. That is, they
must operate from a voltage source (0 Ohms) to 8 ohms. Combining a
complementary pair of these together makes a constant impedance 2-port
crossover or diplexer. Putting two of these together to form a 3-port
crossover is how all of my networks are done. As it turns out, the lowpass of
the high crossover can be paired with the highpass of the low crossover to
form a bandpass (for the squawker). This works! It has one big drawback
though. The two filters that make up the high crossover (squawker lowpass and
the tweeter highpass that from the bandpass) need to look through the low
crossover highpass to see each other. This impedes the nice reactance
cancelling that gives the nice constant impedance. In other words, it doesn't
work quite as well as my usual dual diplexer scheme. I tried this in the
computer using the 4 filters of my ES700 and ES5800 netwroks. The combination
showed an impedance error at the squawker / tweeter crossover (plot
attached). It also has the drawback of not allowing the two diplexers to be
built on seperate boards like I like to do (ES700 and ES5800 seperate). The
third drawback is that it does not allow the 8 Ohms high frequency diplexer
to be swamped down the the 6 Ohm woofer diplexer. That's a very minor one
though. A transition between 6 and 8 ohms is nothing compared to the 6 to 30
Ohm transition you get with the Klipsch AA network! For the attached computer
experiment I scaled the 6 Ohm low diplexer to 8 Ohms to match the other one.

If all this rambeling doesn't make sense, don't worry about it. Consider it
me talking to myself! The bottom line is that if Erik Mandaville had not
asked what seem to me like a dumb question, I would not have realized all
this! Who says you can't teach on old dog new tricks!

Thanks Erik!

John smith. []

P.S. I just realized that I used the ES8200 squawker / tweeter netwrok for the test rather than the ES5800 network. That makes no difference to the test, it just explains why the crossover is up at around 8000 Hz rather then 6000 Hz.

[Al Klappenberger]

Guys,

I decided to take the experiment one step further. Rather than use a bandpass
filter formed of a highpass and lowpass grafted together, I scaled a "true"
bandpass filter using the 8200 Hz highpass filter as a reference. This type
of bandpass filter has the filtering function of both the upper and lower
passband edges distributed through the filter rather then lumped at the ends
like with a highpass-lowpass type bandpass. This means the filter forming the
tweeter port doesn't have to look through the squawker lower limit highpass
half of the bandpass to see its mirror image. As the computer plot shows you
get a nice constant impedance and a much flatter squawker passband! This
network could probably be built but I think it would be harder to do than the
cascaded 2-port scheme I have been using.

Guess who..

[Deang]

Keep it Simple Smith.[]

Thanks, it'll take me 3 hours to decipher those posts.

[Al Klappenberger]

Guys,

Here we go again! This plot is identical to the first one with the highpass -
lowpass type of bandpass. This time however the lowpass and highpass have
been reversed. Now the woofer lowpass filter must look the squawker upper
frequency limit lowpass to see its mirror image highpass. Now the impedance
goes to pot at 700 Hz rather then 8200 Hz.

I think I have proven (to myself at least) that highpass - lowpass type
bandpass filters are not very good for squawker filters compared to cascaded
2-port diplexers.

Soooooo.... On the super AA, connect the tweeter filter to the .3 mHy
inductor side of the 40 uFd cap, NOT to the input side!

THANKS AGAIN ERIK!

Dean,
It's when I try to simplify that I get myself confused! [:$]

Me again.

[Erik Mandaville]

Al:

I've tried to post here half a dozen times, but have constantly been getting error messages.......so I just gave up and went on to other things.

Maybe it will work for me this time around. At one point above you indicated that, " This means the filter forming the tweeter port doesn't have to look through the squawker lower limit highpass

half of the bandpass..."

I know you have done some more work on this since the above conclusion, but this is what I was suggesting -- for this particular network, which as you said (Dean, too) can't work in all instances. I in fact just built a new pair of type 'As' with some parts I got earlier in the year to try with the autoformer (see pic). and left the tweeter filter in it's original place. I also think this network could work in the same fashion, but would require some changes in value of capacitance.

Truly, Al, I can't teach you new tricks, but I thought this might be one different way of approaching this -- that might have validity. And you're right. I have asked millions of probably pretty dumb questions, and continue to do so, not to mention make dumb mistakes with my own projects. Luckily, I have learned to check and recheck myself, and most often catch where I messed up!

edit: just to clarify -- the mylar cap across the 2uf GE oil is not a bypass capacitor for the sake of bypassing (which never seems to make much difference to me). It is a paralleled capacitor for the overall value I needed.

Erik

[Erik Mandaville]

oops. I meant to show both finished type 'A's (one of which is on the left using an old autoformer I had). The one on the right is another network I made last year with the variable L-pad to attenuate the midrange. I'd love to be able to test the one on the right, which is a higher order network, but I'm sure I would be embarrassed by it's plotted performance! I made it with what I had on hand, using values that were in the ballpark for what I needed. It actually didn't sound too bad, but I it sounds a little 'lossy' to me compared to a straight ahead and very simple 'A.' With the bridged SE OTLs I'm using right now, the new crossover seems well-matched.

Erik

edit: A final note, here: The network to the right was one I designed, and the tweeter branch is on its own, connected to the input to the crossover -- in the same manner in which I was suggesting might possibly done with the Super 'AA'. I enjoy the challenge of sometimes making things work with what I might happen to have at the time, which can't and won't always work out. In this case, the values I had for 'L' and 'C' were close enough to do the trick reasonably well.

[Al Klappenberger]

Erik,

The place to connect the tweeter filter depend on how you desinged it. If it was designed as a true "diplexer" having two mirror-image filters, the heed to be tied directly together. My designs are done this way. Desings calculated from formulas found in textbooks that have specific equations both a highpass and lowpass also need to be connected directly togher to work best. On the other hand, with the stand-alone filters grafted together, like ALL the "official" Klipsch netwrok designs, it doesn't matter ho you connect them. They will never form a "correct" reactance nulling configuration like a "true" diplexer will. What the excercise I posted a week or so ago tought me is the difference between connecting singly-terminated versus doubly-terminated filters together to form bandpass filters appropriately.

If you are courious how the specific netwrok you built will perform, I will be happy to put it into the computer and analyze it. The computer simulation is a very good indicator of how it will perform. All you need to do is send me the schematic with all the part values labeled.

BTW: I am impressed by you build job. It's quite nicely done.

Al K.

[Erik Mandaville]

Al:

"If you are courious how the specific netwrok you built will perform, I will be happy to put it into the computer and analyze it. The computer simulation is a very good indicator of how it will perform. All you need to do is send me the schematic with all the part values labeled."

This is very generous of you, I appreciate the offer! If I get brave enough, I might like to take you up on that, but as I said, the results might really be nothing to be very proud of on my part.

Thanks for the compliment, too. The markings I made on the network to the right are kind of rough, but I didn't have a pencil on hand, so I just burned them into the soft pine with the tip of my soldering iron. Speaking of high quality work, Al, I was very truly impressed with the new networks a friend of mine received from you. Very, very nice soldering and layout, and they would have looked just as good displayed outside the cabinet as within.

I understood your descriptions of the test you did concerning all of this; thanks for sharing the findings.

Erik

[Al Klappenberger]

Erik,

Figuring a good way to mark the connections to the barrier block is a problem for me too. I just print them on heavy paper with my ink-jet printer from Generic CADD and glue them on the edge. That's a but hoky too!

AL K.

[Deang]

I've been using crayon -- it looks nice.

[Erik Mandaville]

Al:

That sounds like a good idea, and especially so since you are in the position of needing to make sure customers are clear on the connections. Dean and Bob are of course in this same position.

On this last little 'A' network, I just left the terminals unmarked, since I know which terminals are devoted to specific drivers and their polarity.

You have got me thinking about some things in this thread that might be interesting to try, which -- either fortunantely or unfortunately -- is sort of my M.O. I thoroughly enjoy the experimentation, but constantly yanking components out of the system, which I seem to tend to do on a fairly regular basis, makes it difficult for others to just relax and enjoy some music.

It's a curse, I tell ya.....a curse!

Erik

[BEC]

I just use these things.

Bob Crites

[Erik Mandaville]

Bob:

Those look nice, but not quite as professional as the wobbly Flintsone-lines I created with the heat of my trusty Weller!

Actually, I really like this new iron that my dad gave me for Christmas. The other one I had was a little more expensive, but I like being able to adjust the temp. with a dial rather than by tip changes. It's has the convenience of a variable L-pad!

Erik

[Al Klappenberger]

Those edge plate labels Bob uses are the right thing, but I hesitate to invest in them becasue they won't make the network sound any better! They just look pretty and more importantly, they will hold up for years. My heavy paper ones are a bit fragile.

Al K.