Flat response vs. crossover parts count?

[spasticteapot]

I'm trying to better understand crossover design, and was wondering - what's the general thoughts on smooth frequency response vs. parts count in a design? Drivers like the CA15RLY don't appear to perform quite as well in harmonic distortion or several other attributes as several less expensive options, but the frequency response is very smooth.

I've seen some fairly popular designs with -5dB dips, though a lot of the best are razor-flat across the range. What's your thoughts?

[dcibel]

Re: Flat response vs. crossover parts count?

The parts count is whatever it has to be to get the desired result. Certainly some speakers can get away with a lot less parts count than others, so keep that in mind when choosing drivers, as a cheaper driver that requires a more complex crossover may add up to a more costly build. Cost of course is relative to the size of the parts, so if you somehow have a complex crossover but without many large parts, then cost can still be low overall.

There is also often more than one way to skin a cat, so always aim for the simple crossover first, and add parts when its necessary to do so.

[PWR RYD]

Re: Flat response vs. crossover parts count?

I agree and will add that as I design and listen to more and more speakers I've become much less obsessed with a) keeping the parts count as low as possible and b) worrying about trying to fix every little bump.or dip in the FR.

[johnnyrichards]

Re: Flat response vs. crossover parts count?

My experience has been that too many designers use a crossover with a high Q on the bottom end of a tweeter to fill in dips in the lower treble - same mechanism as bass boost to fill in the bottom end of a subwoofer, this can create a lot of issues with tweeter performance. I have found it very satisfying to leave a dip around that area rather than add another capacitor to the tweeter to create a sharper knee to fill things in.

Similarly, many designers will create boost in the knee of the woofer roll off to fill in the same dip. Neither approach is ideal IMHO, but probably better the woofer is tasked with handling filling it in - unless it is exacerbating common rising 3rd order distortion in that region.

[isaeagle4031]

Re: Flat response vs. crossover parts count?

I will mimic what the others have said. Sometimes it takes what it takes. At InDIYana I had 2 2way dedigns. Very different drivers in each. 1, a ribbon based 2 way required 9 parts in a parallel network(6 of those on the woofer) to get the desired result. The other a very simple 2way with arguably lesser competent drivers, a 5part series design. It is what it is.

[rpb]

Re: Flat response vs. crossover parts count?

The parts count is whatever it has to be to get the desired result. Certainly some speakers can get away with a lot less parts count than others, so keep that in mind when choosing drivers, as a cheaper driver that requires a more complex crossover may add up to a more costly build. Cost of course is relative to the size of the parts, so if you somehow have a complex crossover but without many large parts, then cost can still be low overall.

There is also often more than one way to skin a cat, so always aim for the simple crossover first, and add parts when its necessary to do so.

Ditto what dcibel said.

Here's a flat one for ya. It's not mine, but I helped with the x-over. Part count was 14. There was no obsessing really. We started with nice smooth drivers. I suspect that a minor part change could have filled in the slight depression around 2.5k. I never heard these speakers though, and although it might have sounded a little different with the dip filled in, it might not have sounded any better.

[LongHorn64]

Re: Flat response vs. crossover parts count?

Hey rpb, just for grins could you comment on that build, drivers, etc. I realize it's not your build, but you obviously know something about the project. That graph thumbnail is simply amazing.... I may well have to print it out on legal paper, frame it, and call it the Holy Grail of Crossover Design (:->) All the best, Mike

[rpb]

Re: Flat response vs. crossover parts count?

Hey rpb, just for grins could you comment on that build, drivers, etc. I realize it's not your build, but you obviously know something about the project. That graph thumbnail is simply amazing.... I may well have to print it out on legal paper, frame it, and call it the Holy Grail of Crossover Design (:->) All the best, Mike

It's the 3-way that Ani built and took to the Indiana DIY event. The mid was an RS100p. I believe the woofer was an RS150, and tweeter an SB19. Ani took all measurements, and made files for use with PCD. I did the x-over sim.

[dlr]

Re: Flat response vs. crossover parts count?

I think most of us are on the same page. Parts count matters only if you are building for production or just don't have the means or desire. There is a caveat. Flat response has to take polar response into account. Flat on a single design axis can be the wrong way to go. If you've got a good raw polar response (baffle diffraction control or waveguide loading), then flat is a good target. Personal taste can still come into play. Very often I prefer a dip in the 1-3K area for that "recessed" sound.

That said, judicious use of parts can often provide flat response. This is my old 3-way that started with high quality drivers and heavy felt usage. Despite that, it had some problem areas to handle. The midrange required 10 parts, but the woofer and tweeter each only used 3 with one on the woofer to reduce the component values required. Two of those parts played a minor role, the RC at the top of the midrange. You could probably take that off and not hear a difference.



Another consideration is not only complexity, it's system impedance. The raw midrange here (SS 12m) has a DCR of 3.2ohms, yet the resultant system impedance remained at or above 4ohms with the exception of the woofer. One thing I try to avoid is a series trap in parallel if I can accomplish the same thing with a parallel trap in series.



This was with CALSOD. Notice that I set a target with a downward slope, so the result was about +0.5/-1.5 overall. The system impedance is a bit faint, but shows that even low impedance drivers can be used and still have a reasonable system impedance load.

Flat response in this case was best because the polar response was excellent due to the heavy felt application.

Measured result (M/T only in this case). The off-axis was only at 7 degrees, but representative prior to driver rolloff induced changes.



dlr

[PWR RYD]

Re: Flat response vs. crossover parts count?

I think most of us are on the same page. Parts count matters only if you are building for production or just don't have the means or desire.
dlr

Or the Nano Neos...

[JasonP]

Re: Flat response vs. crossover parts count?

Flat response has to take polar response into account. Flat on a single design axis can be the wrong way to go. If you've got a good raw polar response (baffle diffraction control or waveguide loading), then flat is a good target. Personal taste can still come into play. Very often I prefer a dip in the 1-3K area for that "recessed" sound.

I've been anxiously waiting for some design insight gems to come out of this thread, and boom. Here we go. Great comment.

[dustinbagby]

Re: Flat response vs. crossover parts count?

I have decided not to concern myself too much with either, and heres why.

My little Emmalines two way (Dayton DS115, ND20) used ten parts in the XO. My Eileens three way from IN DIY used 15 parts (RS180, RS125, SB19). Both had reasonably flat FR with the Eileens fairing a little better. Both had good phase tracking, and both sound very nice (to me, but I also prefer a more relaxed top end, however thats here nor there).

Now neither had a razor flat response, nor a terribly high or low parts count, but both are very pleasing to listen to. If you have taken good measurements, and applied the right filters in the right ways for the drivers you are using, then there is no reason to concern yourself with either side of this story. Another thing to remember is that you arent going to build a Shelby Cobra kit car and put a little Dodge 2.0L motor in it, in the same way that you arent going to buy a rusted out 1985 Pontiac Fiero and put a brand new small block supercharged Chevy in it (I hope). Do whats right for the job.

[carlspeak]

Re: Flat response vs. crossover parts count?

The elephant in the room no one seems to want to discuss is the use of upstream equilization to flatten response at the listening position. Sure, it's understood there's no guarantee that using an equalizer also assures a smooth power response.

I may be wrong here but there seems to be an aversion to the use of equalizers as perhaps detrimental to the overall sound; whereas adding sub-circuits in the speaker's crossover to flatten this hump here and raise that dip there, and all the associated parts that go with them are okay. I don't own an equalizer but am puzzled by this apparent dichotomy.

[Paul Carmody]

Re: Flat response vs. crossover parts count?

My personal sanity check is that, if the crossover parts start costing more than the drivers themselves, it's time to cut back.

[bullittstang]

Re: Flat response vs. crossover parts count?

My personal sanity check is that, if the crossover parts start costing more than the drivers themselves, it's time to cut back.

I'm glad someone else has their sanity!!! Honestly, that is my little double-check as well (Sure I stole it from you, read your site about 100 times before I ever tried my own design). I especially used it with my first few builds that used $50 in drivers (total), couldn't wrap my head around spending $100 for the XO's, if $40 XO's sounded pretty darn good! Goes back to my internal "cheap-skate" too.

Question - How do you get polars with OM (or can you even)? Is it done by taking multiple angles of the speaker with the mic in the same location and averaging them? I haven't gotten that far into OM or testing yet. Sorry if this is a ignorant question, and don't mean to hijack.

Thanks