Changing values in AA network

[Randy Bey]

Hi,

djk posted a variation on AA networks where the last 2uF cap in the tweeter side gets replaced with a 6uF cap. This is supposed to "smooth out" the HF.

Generally speaking, what does this do? Does it raise or lower the crossover frequency?

I tried this, by running three 2uF caps in parallel, and was not pleased with the result. It lost some liquidity over the 2uF version. Odd since it was supposed to "smooth".

[pzannucci]

Raising the size of the capacitance lowers the crossover frequency, depending where you are in the network.

Now if you look at the thread on Phase Coherent, you will see that there is destructive/constructive driver overlap. Sounds like you got destructive because constructive will typically cause a frequency hump.

Peter Z.

[Al Klappenberger]

Hi Crossover fans,

Here's some computer analysis studies on changing the 2.0 uF output cap on

the AA network tweeter filter.

Here's the stock AA (2.0 uF):

With an intermediate capacitor value (3.3 uF):

Last but not lease, with 6.0 uF:

For comparison sake, I have moved the marker to 6000 Hz in each plot.

Amplitude is in Green, simple impedance is in red, and group delay is in

light blue (Cyan). Flat group delay is equivalent to linear phase. The

analysis assumes a 0.1 Ohm source impedance, so the top reference line of

the amplitued plot is of no interest.

I will let the world decided if this is an improvement or not.

Al K.

[Randy Bey]

Well I've listened to it more since my last post and my mind is changing. There was a flatter sound to some things, which might be the lowering of the 5-6K peak that I had previously.

Otherwise detail seems OK, and it will take a bit more listening to determine if I take it out or not.

I also got this configuration by putting the original PIO caps from the crossover along with the Hovland Musicaps that I had replaced them with, essentially "bypassing" (although not really) the last 2uF cap with the two Hovlands. I might (MIGHT) be hearing difference in materials (Hovland v. PIO).

[KenB]

Hi folks,

Thanks djk, Randy and Al with all the helpful info & input. Hope to try this

AA mod soon and post.

My two cents: Keep the paralleled caps of equal or better quality when trying mods involving capacitors. Even a correct value can be rejected because you can readily hear quality differences in them with Klipsch.

Try to use caps of the non-metalized type for this if you can, using film/foil dielectrics like polypropylene or polystyrene opposed to say mylars.

Alligator clipping a 4uF in parallel with the 2uF is fine to do, and one can lift a lead and drive yourself crazy with a back

and forth comparison. Just don't short the loose clip end.

Cheers / Ken

[Al Klappenberger]

Hi guys,

I couldn't resist abusing the BB a little bit with a plug! Here's how the

tweeter filter looks in my "ALK" network:

Al K.

[johnnyp]

Al,

Thanks for the graphs.

I am still trying to interpret the results but I think you are misleading (and short changing your design) by not using a consistant scale. For example, in the A-A graphs you indicate "Delay 0.100 mSec/div and "Z 5.0 ohms/div." For your ALK graph you use "Delay 0.050 mSec/div" and "Z 0.2 ohms/div". The comparison is A bit misleading to a casual viewer.

Looking at your schematic and the AA schematic, both designs for the tweeter circuit include the big capacitor used in the squawker circuit (in your design, it's the 39uF + 1uF caps). The formula I use for summing caps connected in series tells me the additional cap from the squawker changes the total capacitance very little (1/40 + 1/2.2 = 1/Total = 2.08 vs 2.2 without the 40uF in series) So why do you use it?

[Al Klappenberger]

Johnny,

You are quite right about the scale factors on the plots. The reason is that

the scale factors are all automatic. I can't force the scale factor on the

impedance channel. I can on the others though. Actually, if I did force the

same scale, the impedance plot would be virtually a straight line. The

program (PCFILT) was designed to display the results of analysis to display

all the plots big enough to see.

You are interpreting the circuit improperly when you try to series the 40 UF

and the 2.2 uF. These caps are in different sections of the network. The 2.2

uFd (and the rest of the tweeter filter) works opposite the .3 mHy inductor

(and the rest of the squawker filter) to form a 6000 Hz tweeter / squawker

division having constant impedance. The 40 uF combination operates with the

2.4 mHy inductor to form a 400 Hz division, also of constant impedance. The

entire tweeter / squawker division is way above the woofer / squawker

division so the two do not interact. These are simply two different

"diplexers" connected in series. Each is a separate entity.

Al K.

[Elso Kwak]

Hi, Let me pour some oil on the fire!

When you design the filter as a fourth order Bessel highpass filter you get a constant group delay and a much nicer picture for the response, but with a "soft" edge because that IS the Bessel characteristic.

Also I think Bessel is the least sensitive of all (Butterworth, Chybesev, Bessel) to driver impedance variations that you can NOT avoid in a practical design. And the Bessel has no overshoot on a pulsresponse. And the most important I listenedto the filter. The Bessel was the best. I did the same for the 400Hz highpass for the squawker.

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

ELSO

[johnnyp]

Al,

If I understand you correctly, the .3 mH squawker inductor shunts some of the high frequencies to the squawker but also "balances" the tweeter circuit. Likewise, the 2.4mH woofer inductor shunts highs to the woofer and is essential to "balance" the mid/tweeter circuit. I think there is a dim light beginning to glow between my ears!

I have read some posts here that suggest running the woofer "full-range" without the 2.4 mH inductor (letting the natural rolloff of the woofer/cabinet shunt the highs). But you would not advocate this for your design or the AA, would you?

[Al Klappenberger]

Johnny,

Right, but there is a little more to it than shunting frequencies from one to the other. The "reactance slope" of each filter is mirror image to the other casing them to cancel leaving prue resistive impedance. If the values are not correct, or the opposing filter is not there, the reactance does not cancel. The number of element in the two filters in each diplexer must be nearly equal too. If not, you need computer optimization to make them work together. This is how I make an N=2 squawker filter work with an N=3 tweeter filter.

Al K.

[Al Klappenberger]

Elso,

Let me shoot down a rather common myth here. The group delay of a Bessel

highpass filter does NOT have flat group dealy! Only a Bessel lowpass

filter will show flat delay. I have even tried to use computer optimization

to flatten the dealy of a highpass filter. I have never been able to do it. The

cutoff frequency simply moves down! If you scale normalized element values

for Bessel response to a highpass filter you end up with an amplitude and

impedance response (return loss) exactly as you would expect, but the delay

is NOT flat! If you don't beleive it, I can post plots of any element values

you suggest!

Al K.

[djk]

>>I tried this, by running three 2uF caps in parallel, and was not pleased with the result. It lost some liquidity over the 2uF version. Odd since it was supposed to "smooth".<< Thanks to Al for the graphs.I recommend you try the 230µH + 4µF on the K55V along with the 6µF change on the tweeter.PWK designed this to lift the falling response of the K55V before slamming it off hard at 5Khz.Because of the non-coincidence of the mid-range and the tweeter there are a series of dips and peaks on either side of the crossover point.Cutting the midrange off with a steep filter like this will reduce both the frequency spread and amplitude of these peaks.Just changing the second capacitor in the tweeter circuit does improve the sound but the full effect won't be there until you help the mid-range clean up its act too.As far as the 128µF cap in parallel with the woofer goes I have seen values from 100~140µF used.I have also used 68µF in parallel with the woofer and a Zobel of 6 ohms in series with 27µF and this combination also wired in parallel with the woofer.This last combination is the most benign for an amp to drive.The 140µF is the worst.Anybody having a real POS amp should use the 68µF with the 6 ohm + 27µF Zobel.Good quality electrolitic caps are ok for the woofer circuit although Klipsch uses high quality film types here.The 4µF in parallel with the midrange must be a very high quality film type.It should be pointed out that it is very easy to try the different caps on the woofer.Put a card table in front of one speaker at a good listening distance.Run a suitable length of zip cord from the woofer terminals out to the table.Wire one of the 27µF caps in series with the 6 ohm resistor.Set the other 6 ohm resistor to the side.Buss together one end of both 68µF caps, the second 27µF cap, and one end of the Zobel.With a bunch of clip leads from Radio Shack you can try both 68µF in parallel (136µF),a 68µF and a 27µF in parallel(95µF),and a 68µF by itself and with the Zobel.Keep the volume low but loud enough to hear what is going on.Whatever sounds best is best.This is a $10 experiment and the worst thing that could happen would be to find you liked the 136µF value the best and found you had to order another pair of 68µF caps.Otherwise you have enough parts to make two of every combo that you tried.

[Elso Kwak]

Hi Al.

Oops you are right. They delay is only flat until the crossover frequency(and only for zero Ohm coils); than it is a straight line falling down on loglog scale for dB/frequency. I have been doing some thinking why my filter sounds so good and I came to the conclusion that the rounded corner or knee of the Bessel characteristic comes in handy to tame the peak at 5500-6500 Hz.

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

ELSO