Taming harshness in horn tweeter

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I hope I can get some help from the group. Thank you in advance...

I have a pair of Klipsch KG 4.5 I've been, er... tweaking.

So far, I've applied Duct Seal to the woofer frame and magnet, and to the backside of the tweeter horn and around its magnet. That made a noticeable improvement in midrange clarity.

After that, I bought the upgrade titanium tweeter diaphragms from Bob Crites. That improved the highs, noticeably smoother.

The most recent tweak has been lining the cabinet walls with polyester felt, and putting a layer of felt around the tweeter horn on the front of the baffle, which helped the imaging and clarity a bit more.

The final frontier is how to get that last bit of horn harshness out of the tweeters. There is a hardness and sizzle to vocal sibilants and cymbals.

I don't want to mess up the midrange around the crossover region, which is pretty good right now.

I thought I'd try a Zobel network on the tweeter, to see if that would make the trebles sound "softer." I don't have specs on the Klipsch K-85-K tweeter, and can't find them anywhere, so I don't know what the voice coil inductance is (Le). I can tell you the voice coil resistance (Re) is an unexpectedly high 8.1 ohms.

Can I assume the tweeter impedance is 8 ohms and use an 8 ohm resistor? Or should I go with 1.25*Le and use a 10 ohm resistor?

What would be a typical Le value for a cheap-ish horn tweeter like this? 0.05mH?

I have available 7.5 to 10 ohms for R and capacitor values from 0.22uF to 2.2uF for C, at least to start with. I put a zobel using 0.47uF and 10 ohms in last night, but haven't listened to the speakers at full listening levels yet.

I just read that a resistive attentuation circuit actually damps horn resonances better than a zobel. So I could calculate an L-pad and use that instead. But once again, I'm confounded by the Re of 8.1 ohms for these tweeters. Should I calculate the L-pad values based on an 8 ohm impedance, or something like 10 ohms instead?

I could try a 1.5 or 2 ohm series R with a 50 ohm parallel R. That should just pad down the tweeter by about 1.5dB.

I've already tried paralleling the tweeter with 100 ohms, but removed it before I upgraded the tweeter diaphragms and added the felt layers to the cabinet. I will try this again.

--
 
There is a hardness and sizzle to vocal sibilants and cymbals....

I thought I'd try a Zobel network on the tweeter, to see if that would make the trebles sound "softer."...
--

That "Glare" may not be from the tweeter. Could well be from the cone breakup of the 10" Carbon Graphite Cone woofer. If there's cone breakup, a zobel on the woofer can tame that sizzle. The other way is to apply a notch filter.

Another cause could be the tweeter is resonating. This happens when it's crossed near it's resonance frequency. For this, you'll need a conjugate network (LCR).
 
Thanks for the replies.

I got home and took a quick listen to the speakers with the zobel networks on the tweeters. The zobels took all the life out of them. Definitely rolled off the highs above 10kHz, but left them sounding too polite and reserved.

For test listening, I use a Denon CD of Ravel's Ma Mere L'Oye (Orch National de France, Eliahu Inbal, cond.) that has nice triangles, col legno violins and cellos, spooky muted violins and some nice blatty trombone blasts. Lots of good high frequency stuff.

Right now I've got 100 ohm 5W wirewounds in parallel with each tweeter. No series resistor. It smooths out the trebles just a hair. I guess they sound pretty OK.


bwaslo said:
Try some reticulated foam in the tweeter horns.

That's an intriguing idea, might just be the ticket. I took a quick look to find some for sale, but couldn't. Where to buy in the USA or NYC? Canal Rubber?


tomtt said:
hardness and sizzle? electrolytic caps would do something like that.

No electrolytics in the tweeter circuit (all are small metallized mylar), but there is a 33uF 100V non-polarized electrolytic in the woofer circuit. The crossover board is really small, so fitting a 33uF film cap would be difficult to impossible. Maybe a much better quality Panasonic NP 'lytic would improve things? I'll take a look at DigiKey, Mouser...


Michael Chua said:
That "Glare" may not be from the tweeter. Could well be from the cone breakup of the 10" Carbon Graphite Cone woofer. If there's cone breakup, a zobel on the woofer can tame that sizzle. The other way is to apply a notch filter.

Another cause could be the tweeter is resonating. This happens when it's crossed near it's resonance frequency. For this, you'll need a conjugate network (LCR).

Food for thought! Yes, that makes sense. I think I'll try a zobel on the woofers. I can't seem to find specs for these Klipsch drivers. It's easy to get the Re of the woofer, but I doubt I can find the Le (I have no L meter). I could ballpark guess, it I suppose. It's Re * 1.25 for the R, and something like 22uF to 33uF for the C, correct?

I wonder if anybody knows what drivers Klipsch used in these. They didn't make their own, correct?

Thanks again for the suggestions.

--
 
diyAudio Moderator
Joined 2008
Paid Member
Before you go adding this RC circuit (or any zobel compensation), you should consider the potentially double-edged effects...

An RC circuit like that usually doesn't help when it affects the top end where the crossover is just passing the sound through, because you need components between the amp and speaker which are effective at those frequencies, and unless you measure....hence the fact it is rarely used this way.

However since you said it did make a difference, it has obviously altered a part of the impedance that was under the influence of the crossover and this is the point...when you add a zobel for zobel's sake, you should do it before adding the crossover. When you add it after, you not only change the expected behaviour of the crossover but you add a new level of complexity to it with results that can be hard to predict.

If you are going to add components to the crossover you will want to consider the entire circuit in the bigger picture. Some here might be able to make this easier if you wanted to post your crossover (if you can).

lining the cabinet walls with polyester felt
If this is the thin felt, then it may not have a noticeable effect on the walls but might be useful around the rough edges behind the woofer.

Using felt on the baffle may be a good idea, especially if it is a little more thick. It wouldn't hurt to try covering the entire baffle.

I could try a 1.5 or 2 ohm series R with a 50 ohm parallel R. That should just pad down the tweeter by about 1.5dB.
This may be worth trying as it's intended effect is likely to be greater than its side effects.

You need to be sure you've considered all reasonable options with damping materials first. Then if you need to change the tone of the speaker, it should be done as a whole. In other words, if you want to bring down the upper treble, you should prevent changes in the tweeter down near the crossover. Maybe all you need is an RC with different values, or a shelving network.

Hope this helps.
 
Ex-Moderator R.I.P.
Joined 2005
when you add a zobel for zobel's sake, you should do it before adding the crossover.

When you add it after, you not only change the expected behaviour of the crossover but you add a new level of complexity to it with results that can be hard to predict.

ahh, now I know why it takes me months to adjust those bloddy zobels :D

I also spend months to get tweeter filter/attenuation combination right :eek:

dull or harsh sound is easily caused by crossover malfunction
and the differences where this change happens can be ridicoulusly small
 
...when you add a zobel for zobel's sake, you should do it before adding the crossover. When you add it after, you not only change the expected behaviour of the crossover but you add a new level of complexity to it with results that can be hard to predict.

If you are going to add components to the crossover you will want to consider the entire circuit in the bigger picture. Some here might be able to make this easier if you wanted to post your crossover (if you can).

Easy enough, here it is:

Kilpsch_kg-45_crossover.jpg



If this is the thin felt, then it may not have a noticeable effect on the walls but might be useful around the rough edges behind the woofer.

Yes, it's the thin felt. What is "around the rough edges behind the woofer"? Do you mean the baffle and mounting hole? Or the back wall, directly behind the woofer (i.e. soften the back wave reflection)?


Using felt on the baffle may be a good idea, especially if it is a little more thick. It wouldn't hurt to try covering the entire baffle.

Yup, thicker felt. Not wool felt, but it's thicker polyester felt with adhesive backing, from Home Depot.


...(L-pad) may be worth trying as it's intended effect is likely to be greater than its side effects.

I have a 100 ohm resistor parallel to the tweeter now, and was listening earlier. It makes a very slight difference, sort of softens the highs in the 8kHz and up area.

I wonder if Michael Chua's suggestion is accurate, that the sibilance might be caused by resonances closer to the crossover region. Maybe 4kHz or so? That could be... If that's the case, a zobel across the woofer might help. But this one's tricky. There's already that 33uF cap in series with 2 ohms to ground. Sort of looks like a zobel already, right? But that 2 ohms is way lower value than the voice coil resistance (which is 7.2 ohms on these woofers). If I do go with a zobel on the woofer, I calculate its values as R = 9 ohms, C = 22uF. Even just electrically, that would interact with the 33uF and 2 ohms in series from woofer + to ground, would it not?


You need to be sure you've considered all reasonable options with damping materials first. Then if you need to change the tone of the speaker, it should be done as a whole. In other words, if you want to bring down the upper treble, you should prevent changes in the tweeter down near the crossover. Maybe all you need is an RC with different values, or a shelving network.

I wish I knew exactly what the problem is. I thought I did, but now I have doubts. It could be the lower end of the tweeter response that has a resonance causing harsh sibilants. Or it could be cone breakup at the top end of the woofer's response. Or it could be an overly elevated treble above 10kHz. Measuring tools and more knowledge and expertise than I have are required, I think.

I'll try an L-pad over this weekend, to see what that sounds like. Easy to experiment with that.

I could also try a simple series resistor and nothing else. I gather that reduces output down at the crossover end of the tweeter's response, leaving the highest frequencies relatively untouched. If there's a resonance or peak in the upper mids, then that could mellow it out. Another easy thing to do.

If that doesn't do it, I'll try a zobel on the woofer, on the off chance that cone breakup is causing a problem. Again, easy to do.

If none of that does the trick, well, I suppose I'll have narrowed down the options.

Hope this helps.

It sure has! Thanks for replying.
 
for reticulated foam, try an aquariumshop.
its used for filters.

Ah, thank you.

BTW, with no resistors added to the tweeters, I had stuck a thin layer of sparsely woven polyester batting (decorative Christmas "snow blanket" sheeting) behind the grill cloth, so that it just covered the mouth of the tweeter horn. It did a nice job of mellowing out the sibilance. I liked the sound pretty well. The only reason I didn't just leave it that way, was that a sheet of something over the speaker seems 'wrong' somehow. The reticulated foam seems like a similar idea, but with a more high-tech allure to it (and might work better). Worth a try, for sure.

--
 
Last edited:
Try changing the zobel resistor to a higher value. I'm using 33uF + 10 ohms for some of my speakers. You'll have to make the adjustments by ear. Basically, what you will be doing is altering slightly the crossover slope of the woofer.

Does that mean you see the woofer crossover as a 1st order xover with a zobel after it? But isn't that a 2nd order xover with a resistor stuck under the parallel capacitor, to make the filter action more 'shallow'?

My understanding is that the 2 ohm resistor in series with the 33uF cap is too small in value to affect the impedance curve of the woofer much -- because that 2 ohms is much lower in value than the woofer's Re of 7.2 ohms.

If I try this... How will swapping out that 2 ohm resistor for a 10 ohm resistor affect the slope and/or cutoff of the filter?

I'm afraid I don't understand the action of the various parts of an LC filter well enough to answer these questions for myself. I wonder if there's anything about this kind of thing in Vance Dickinson's book...

--
 
diyAudio Moderator
Joined 2008
Paid Member
I wonder if Michael Chua's suggestion is accurate, that the sibilance might be caused by resonances closer to the crossover region.
Try listening to the woofers on their own.

a zobel across the woofer might help. But this one's tricky.
No. As Michael suggests, increase the resistor value. Don't go too far. The manufacturer made it that way for a reason and it will have a complex and likely undesirable effect if you go too far. If you find out for sure this helps then you can look at more complete solutions.

It could be the lower end of the tweeter response that has a resonance causing harsh sibilants. Or it could be cone breakup at the top end of the woofer's response.
It could also be that you don't have a smooth and balanced response right across the treble...there could be many reasons for that.

I could also try a simple series resistor and nothing else. I gather that reduces output down at the crossover end of the tweeter's response, leaving the highest frequencies relatively untouched.
To do that you could look at reducing the size of one of the capacitors, possibly the first one.

The series resistor is likely to reduce the middle of the tweeters range. The upper and lower tweeter response will probably show less change. This is worth experimenting with.

a sheet of something over the speaker seems 'wrong' somehow
If it's reasonably open and breathable then go ahead.

Does that mean you see the woofer crossover as a 1st order xover with a zobel after it? But isn't that a 2nd order xover with a resistor stuck under the parallel capacitor, to make the filter action more 'shallow'?
Your woofer uses a filter effectively somewhere between first and second order, looking at the components, but even that isn't clear until you figure in other interactions.

A crossover can be made either with or without zobel correction. It means very little that a woofer has has its impedance flattened in this way. The zobel won't make the unit appear flat before (at the amp) or after (at our ears).
 
Does that mean you see the woofer crossover as a 1st order xover with a zobel after it?

Yes, it is a 1st order with a Zobel, not a 2nd order. It is a 2nd order if R is very small. That technique is sometimes used to tweak the crossover. By adjusting R, the slope is altered.

The primary role of a Zobel is to convert the driver from a reactive load to a resistive one. It is possible to get an impedance sweep to be straight line with a Zobel. When the load is resistive, crossover slopes will conform to textbook theory. This, of course, is based on the assumption that the driver has a flat response to begin with.

Perhaps a design of mine ( Starling ) will illustrate better what I mean. The speaker is a 2-way, a Seas ER18RNX with a Seas 27TDFC tweeter. You can see from the manufacturer's frequency response plot that cone breakup starts at 2KHz. At 5KHz, there's a peak of about 7dB.

The middle picture is my LMS gated sweep of the woofer with crossover, tweeter with crossover and the summed response of the woofer+tweeter. The crossover is not just proper summing at the crossover. Often, the problem is the before and after. If you look at the woofer plot (Pink), you can see clearly the cone breakup peaks at 4 and 5KHz are still there. Because they have not been notched out, they boosted the summed response (Red) at 4-5KHz by a few dBs. This is where the brightness/glare/sizzles come in. It's not from the tweeter. It's distortion from the woofer cone breakup. To save cost, I tweaked the zobel values to pull the cone breakup lower. It is still there but under normal playback levels, the distortion doesn't surface. Play the speakers very loud and the glare may reveal itself. A better solution is to use a notch filter to completely kill the breakup. But then cost goes up.

I'm not saying tweaking a notch filter will solve your problem. Right now, without any measurements, it's largely guesswork. But since it's only a resistor and a capacitor, it's worth a try.
 

Attachments

  • ER18_PLOT.jpg
    ER18_PLOT.jpg
    36.6 KB · Views: 419
  • GATED_SUM.jpg
    GATED_SUM.jpg
    107.5 KB · Views: 436
  • XO.jpg
    XO.jpg
    20.5 KB · Views: 455
  • STARLING_PAIR.jpg
    STARLING_PAIR.jpg
    16.7 KB · Views: 334
Ex-Moderator R.I.P.
Joined 2005
Yes, it is a 1st order with a Zobel, not a 2nd order

on the electrical, and theoretical side, yes

on acoustic side it could still be 2. order
but it might change from butterworth to Linkwitz, or even Shebesky, or whatever in between

depends on how the slope looks, and where it 'meets' the summing driver
and on how the slope of the summing driver looks

its a puzzle

if you change your 2ohm to 10ohm, it might work, but would be sheer luck
you will have to try all other values close to it, or in between

and further
any change you make to your woofer or mid could make it necessary to also make changes to tweeter xo, before it works properly
and when you have changed the tweeter xo to suit the new mid changes, you may need to change the mid xo, again

and so it goes back and forth
you need to work exstremely systematically, or it could go on forever, in circles
its tricky

some tricks
the simple one is to hold a cap or resistor in paralel with the component you want to 'investigate'
might tell, if it need be bigger or smaller
or, lift one leg, and solder a new component to the still connected leg

and it typically happens that once a new component is in place, it may sound worse than before
this can go on fore a very long time :eek:
ist only rockandroll, and I like it, like it....and hate it :D

btw, guessing is a part of this, yes...but only a very very small part of it
 
I had stuck a thin layer of sparsely woven polyester batting (decorative Christmas "snow blanket" sheeting) behind the grill cloth, so that it just covered the mouth of the tweeter horn. It did a nice job of mellowing out the sibilance. I liked the sound pretty well. The only reason I didn't just leave it that way, was that a sheet of something over the speaker seems 'wrong' somehow.
--

That's the way all technical breakthroughs are viewed initially. Then people come to say "why didn't we do that all along?"

FYI, the use of foam (actually any attenuating medium) in this manner is patented so don't use it commercially.
 
I have hear it is not possible to up hold a patent on something that have been manufactured commercially before, the patent was made.
Does anybody know about these patent rulles?
I remember it have been up before, because someone belived, that fome or felt material have been used in horn throats many many years ago.

I know a lot of patents is made today on something we have manufactured for decades.
I think some people in big companys get a bonus each time they can make a patent. Even though they are useless....
 
If something has been done before, its called "prior art" and cannot be patented. There is prior art using foam in the throat of the horn, but the pieces were small and did not encompass much of the volume of the waveguide. The more volume that is used the more effective the foam is. Hence, to be most effective a large volume should be used and the patent specified a volume greater than some percentage of the total volume of the waveguide is filled. Used in this way, there is no prior art.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.