Taming harshness in horn tweeter

[system7]

There are simple enough resources available to calculate these effects.

I particularly like this LCR calculator from Alan Yates, even if entering data is clumsy due to choice of units:
Alan Yates' Laboratory - VK2ZAY's Engineering Calculators - LCR Resonance

What you do is treat a second order filter as a parallel LCR circuit. You need to measure the DC resistance of the drive unit Re which you use as shunt resistance or perhaps a nominal impedance about 25% higher than Re, then put in the L and C values. It will tell you the LC frequency and Q of the circuit. A Q of 0.5 is a Linkwitz-Riley with a gentle rolloff and kind impedance. A Q of 0.707 is the butterworth with maximally flat response and a more difficult dip in impedance.

Otherwise, you can play with this crossover designer which is well thought out:
2-Way Crossover Designer / Calculator

I make your proposed XO to be second order butterworth LP @1800Hz and 8 ohm. The HP is third order butterworth @ 4000Hz and about 10 ohm. No bad thing, that gap between the two frequencies, it's a common technique with real loudspeaker drivers.

Bit rough and ready, of course. Real loudspeakers are not 8 or 10 ohms. Rules of thumb, LR is less risky to your amp than butterworth, and keeping crossover points apart keeps load impedance higher too. That Zobel on the bass will get quite hot, I reckon, and reduce impedance a bit too. So be careful.

FWIW, I doubt if adding little parallel caps to electrolytics will make much difference. Polypropylenes would be better, but COST!

[system7]

Originally Posted by rongon

PS - I have a box of chokes left over from my misspent youth. I have pairs of 0.17mH, 0.62mH, 1.2mH, and 1.8mH. Anything useful in there?

Ooh, yes, I've just had a good idea.

Add a baffle step circuit to the bass:
Loudspeaker Diffraction Loss and Baffle Step Compensation Circuits

3.7dB on 12" baffle gives 4 ohms and 1.8mH. You'd want at least 2X 10W W/W 8 ohm in parallel here. No idea if your 1.8mH coils are up to it. But that gives you 4dB to play with on padding the horn!

[rongon]

Ah, some interesting ideas. I might try the proposed BSC circuit, to hear what it sounds like.

Somebody earlier mentioned that ferrite-core chokes in the tweeter circuit would cause distortion. Would this be clearly audible with low power input, say a total of 5 watts rms to the crossover inputs?

I'm thinking I need to swap an air core inductor into the HP circuit in that KG5.2-style xover.

--

[badman]

Originally Posted by rongon

This is interesting. Maybe what was needed all along? Or maybe something to try in tandem with xover changes...

Felt or foam walled waveguide?

Foam around the horn to tame harshness...

Yes.

It's hugely important to properly terminate horns.

[AllenB]

Originally Posted by rongon

I might try the proposed BSC circuit, to hear what it sounds like.

It is likely your speaker has already been compensated. Without measurement you can't be certain it hasn't.

FWIW this kind of circuit is somewhat similar to a 'loudness' control, if that helps to understand what it might sound like.

[Ken Newton]

Originally Posted by AllenB

It is likely your speaker has already been compensated. Without measurement you can't be certain it hasn't.

I'm not sure that is so. We're talking about a broad shelved 3-6dB SPL contour. Many speaker FR measurements are taken near-field and don't clearly reveal whether such a broadband EQ has been applied or not. In-room far-field measurements often aren't much clearer at showing whether such a smooth and broadband correction has been applied.

My own experience is that the greater the midrange and treble clarity of the system, the more audibly obvious is the benefit of baffle-gain/loss correction. All of my own speaker systems to which I've applied BSC (stand mounted dynamic mini-monitors, floor-standing dynamic, tall electrostatic/dynamic hybrid) have benefited.

[rongon]

Originally Posted by AllenB

FWIW this kind of circuit is somewhat similar to a 'loudness' control, if that helps to understand what it might sound like.

It sounds like it has been compensated. It has a sort of recessed midrange, with a definite warmth in the mid-bass and as 'sparkley' a high treble response as you would expect from a large-ish horn tweeter.

Whether the shaping was a conscious attempt at BSC, or was just 'voicing,' I can't say.

I suspect that applying BSC to this system might be adding too much of a good thing. I might try it, though, just to hear what it sounds like.

Thanks for the continued flow of ideas, everybody.

Incidentally, the speakers are sounding pretty good, especially considering its cheap drivers and middle-brow crossover components.

--

[AllenB]

You won't know if you don't try.

The baffle step is not something that necessarily gets separate attention during design which is why I suspected it was OK. It shows up in the measurements and the crossover just flows from the measurements. The lows aren't really missing, just dispersing more widely so you may want to go slowly.

Oh, and I wasn't meaning to make reference to the highs with the loudness control comment.

[strawberry]

Most likely the woofer falls quiet very fast at around 1.6 kHz, though that in it self is probably a nasty peak which needs a Zobel network to flatten out, then a drop to silence, and then there is some nasty loud residue around 5 kHz which no one wants to listen to. To put a crossover point at 4 kHz, yeah?! Failure to launch. Put it at 2 kHz. 1 mH and 6.8 uF on both the woofer and tweeter should go a long way.