K55V Traps

[JonM]

Apologies if I should add this to an older thread (BobG - feel free to do so if appropriate). I had to go search the archives for info on the P-trap for the K-horn squawker (Al K mentioned it and it got my curiosity up), and got good ideas I'd never find anywhere else (Thanks all!). For those who haven't seen this before, you can tame a lot of the excessive brightness of the K55V squawker by putting a 3 uF capacitor in parallel with a 0.1 mH inductor, and putting that combination in series with the squawker.

A demo speaker design package I downloaded had a circuit for taming resonances using the same values (for a 9 kHz peak), only they were in series, and the series combination was in parallel with the driver. A book on speaker design I checked had both approaches, though it suggested putting a resistor in parallel with the C and L in the P-trap to avoid ringing, and a resistor in series with the C/L of the other approach.

Anyway, I thought I'd give them both a try, since they used the same value parts, and if I didn't like the results one way, I could try the other. I tried the series R/L/C (with R~=8 ohms, to keep the load across the crossover similar to that of the squawker's typical impedance below resonance), in parallel with the driver, first, because it required no soldering of the leads to the squawker. Test leads clipped the parts in place.

Results? Amazing! That hard brightness seems tamed, and the sound is smoother and much more listenable. My missus called the K-horns' sound "primitive", I think in response to the bright upper mids. (It's a great word for the sound, actually.) The sound now is much smoother and, yes, less "primitive". (Yes, I caulked the K400 horn - that had very little effect compared to this change!)

Now, do I bother with the soldering iron to try the P-trap? Not today - I like the results of the filter I tried so much, I'm not seeing much reason to. Maybe in the future when I'm bored and itching to play around.

I'd be curious to hear others' (especially Al K's - our resident filter expert!) opinion of the approach I used compared to the P trap. I'm on a steep learning curve here - this is a great forum. Thanks to Klipsch for hosting it. (What other vendor does the same? None that I can think of!)

[Al Klappenberger]

JonM

I did some computer analysis on the AA network model and discovered that the

"P trap" of 3 uFd and .1 mHy worked equally well parallel connected in

series as it does series connected in parallel. The 8 ohm resistor does not

accomplish anything useful though. Connected across the driver, it simply

reduces the squawker level.

Al K.

[djk]

Parallel traps that have the LCR in series and then in parallel with the driver work fine.AS LONG AS THERE IS SOME RESISTANCE IN SERIES WITH THE SPEAKER.The series resistor portion of the LCR modifies the Q of the impedance dip.If you eliminate this resistor the LC will be a dead short at its resonant frequency,hence the need for some resistance in series with the speaker.Normally most speakers have some sort of resisive pad but Klipsch uses an autoformer and thus has no resistance in series with the speaker.With the 8 ohm resistor the impedance at 9Khz will be tollerable.Without the resistor the impedance at 9Khz would be 0 ohms plus the DC resistance of the .1mH inductor and the autoformer.A better trap would be to use 3.9µF and .080mH with an adjustable resistor to modify the Q of the circuit.Use a 100 ohm pot with a 4 ohm resistor in series with it and adjust for the best sound.Measure the value and replace with a fixed resistor.Remember the pot goes in series with the inductor and capacitor when you wire the trap in parallel with the driver or in parallel with the inductor and capacitor when in series with the driver.

[Al Klappenberger]

JonM

The analysis shows that an 8 ohms in series with the trap is high enough

that the trap does virtually nothing. 1 Ohm yields about a 10 dB rejection

at 9000 Hz. Using no resistor at all in series with the trap reduces the

impedance to that of the component losses. This will be nearly that of the

"Q" (X/R) of the .1 mHy inductor used at 9KHz. With a Q factor of 20

(assumed) the R value will be about 0.28 Ohms. This is too low.

An 8 Ohm resistor connected directly across the squawker driver (outside the

trap) does nothing except lower the impedance a little. I mention this

because I am not sure how you proposed to connect the 8 Ohm resistor.

Connecting the L and C as a parallel trap connected in series avoids the

resistor question entirely.

Al K.

[JonM]

Right now, I'm running 3 uf, .1 mH, and 5 ohms, all in series, and that series combination in parallel across the squawker's terminals. The speaker design book I checked recommended that the resistor be the same as the driver's nominal resistance. The idea is that at the driver's resonance (ideally, the same as the frequency of the trap) the driver's impedance rises and has a higher voltage applied to it "at the output" of the crossover. By making the resistor the same as the speaker's nominal impedance, you get the same voltage across this trap paralleled with the driver at the resonant frequency as you do at other frequencies, taming the acoustic output at resonance.

I tried 10 ohms, and didn't hear much of an effect. 5 ohms seems so far to sound better, but it's too soon to tell (i.e., too much else is going on for me to have listened much yet).

I like the idea of the trap in parallel across the squawker - it leaves the driver connected directly to the crossover w/out intervening losses, it tames the resonance by keeping the load on the xover more constant, and it involves no cutting. It just seems more elegant. (For whatever that's worth :-)

This message has been edited by JonM on 05-20-2001 at 08:04 PM

[djk]

>>I like the idea of the trap in parallel across the squawker - it leaves the driver connected directly to the crossover w/out intervening losses, it tames the resonance by keeping the load on the xover more constant<< Actually the series connected trap across the squawker causes the impedance to drop at the resonant frequency.The load on the crossover is less constant.The way you have it now the impedance will dip to around 4 ohms in the 9Khz region(5 ohms || 16 ohms = 3.81 ohms reflected up to 7.62 ohms by the autoformer.This in || with the 8 ohm tweeter will drop to 4 ohms).If you are happy with this,fine.For most people I think the parallel connected trap in series with the driver would be a better bet.If you want more rejection of the 9Khz use 3.9µF and .080mH.

[johnnyp]

Al or djk: Instead of the P-trap, what about a simple inductor in series with the K55V on the A-A network to tame all high frequencies above ~6K? Has anybody tried this? To calculate the inductor value, what value of R should I assume for the K55V -- 12 ohms? Is the R for the inductor calculation affected by the autoformer R?

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

John Packard

[Al Klappenberger]

John,

A single inductor would simply be a first order lowpass and would not yield enough attenuation at 9 KHz. The network I offer uses a second order lowpass and yields about 8 dB. That is just about the minimum to do the job. A trap (in sereis) will do a far better job.

Al K.

[JonM]

Al -

The RLC series (in parallel with the squawker) should work by lowering the impedance of the paralleled combination at the resonant frequency of the trap (hopefully the same frequency as the driver's resonance). That results in lower drive voltage applied to te squawker at that frequency, and less output by the squawker. (I know you know this.)

My earlier description assumed that the driver's impedance rises at its resonant frequency, and the trap would keep the overall impedance presented to the xover relatively constant. I think your calculation assumes that the K55V's impedance does not rise at its 9 kHz resonance. Frankly I don't know if it does or doesn't, but I assume it does. Has anyone measured the impedance of a resonant K55V to see what happens at 9 kHz?

Either way, the combined squawker||RLC will have a lower impedance at the resonant frequency of the RLC, and hence lower output at that frequency, than the squawker alone would have.

Isn't this also what the LC "elliptical filter" hanging in parallel with the K77 tweeter does? Seems to me that the resonant frequency of that combination is somewhere up around 12-15 kHz (I don't have the numbers in front of me). I have read that the ceramic magnet K77M is more efficient than the K77V was around 12 kHz - is this filter just aimed to reduce any resulting HF peak that might be there? (I don't see how that filter protects the tweeter from much of anything, especially considering the tweeter is fused (in my AK-xovered Khorns) with a 1.25 amp fast blow fuse.) Just thinkin' here...

[Al Klappenberger]

JonM

The K55V and M drivers have their resonance well BELOW the operating

frequency range. The 9 KHz "return" is not a resonance. Rather, it just

looks like an extension of the frequency response. You might just call it a

very rough roll off.

The classic series connected "P trap" operates by effectively opening the

circuit to the driver at 9 Khz. This does no harm. Connecting the L-C as a

shunt trap is trying to short out the source at 9 KHz. This is why the

resistor is needed. The single series cap in the squawker does not provide a

suitable series impedance (reactance) to allow this. It's too large.

The "elliptic" tweeter filter has the notch at 5000 Hz which is below the

6000 Hz cutoff of the filter. This sharpens the lower skirt shape protecting

the tweeter by removing high power low frequencies. It makes the filter

skirt MUCH sharper. The trap is located on the output side of the filter

allowing the rest of the filter to operate as the series impedance the trap

is operating from. It does not load down the source since the trap is

actually part of a complete filter designed for 6000 Hz. The single cap in

the squawker, on the other hand, is a filter designed for about 400 Hz which

is well below the 9 Khz trap frequency. That's why the value is so large

compared to those in the tweeter filter.

Al K

[JonM]

Al -

Thanks for the post. My mistake - for a 15 kHz trap, the program I used specified C=2 uF and L=0.05 mH (not the 0.5 mH that is in the AK xover). Oops!

And now that I think about it (gee, I wish these newbies would think _before_ they post :-), I see why the xover prior to the squawker wouldn't be a high enough impedance to let the RLC trap in parallel with the driver act as an adequate voltage divider to drop the voltage across the squawker's terminals. Classic case of not starting brain before engaging mouth.

So maybe what I heard was the effect of my vivid imagination? The placebo effect? Alien mind control?

Time to try the "real" P-trap.

(Oh, I would still expect to see some impedance change at the K55's 9 kHz peak. Take a look at the impedance charts in Stereophile of whole speakers, and you'll see impedance effects from the resonances of the enclosures, let alone the drivers. I assume that the 9 kHz peak is a resonance of some sort, which has almost got to be reflected back as a bump in the impedance. I'd love to know if anyone has actually done an impedance plot of a resonant K55 to see what the magnitude of this bump might be.)

This message has been edited by JonM on 05-24-2001 at 01:10 PM