• First, component values do not have to be exactly what was modeled. Indeed, it is often impossible to find components that are precisely the computer modeled values. Generally using values within +/- 5% is close enough – and most components are no better than 5% accurate anyway. It will subtly change the performance of the network to not get it “perfect,” of course. If you want to see the effect of the change to decide if it’s acceptable to you, simply jump back onto your model software and enter the exact values you intend to use. That’s the beauty of using software.
• Second, always use non-inductive resisters of sufficient power rating. Generally resisters are used, if at all, as pads for tweeters, so their power rating should be at least that of the tweeter. In a loudspeaker system designed for, say, 100 watts total power handling the tweeter will generally never see more than 10 watts. If the resister is in series with the woofer, however, it can be subjected to the full 100 watts. Wiring 2 resisters in parallel effectively doubles their power handling capability (paying attention to what that also does to the net resistance.) Resisters used in Zobel networks can usually be any power rating that you have available.
• Third, pay attention to the wire gauge of your inductors. Be sure that any inductor used in series with a speaker can handle the power rating of the speaker, and also realize that the smaller wire gauge in an inductor the higher its series resistance will be. I generally recommend using no smaller than 18AWG inductors, and I also recommend generally not using multiple inductors to try and create the effect of a larger one. While this can be done and often is, I find it is less accurate and less predictable than using one inductor that is close to the target value. I also suggest using air-core inductors as much as possible. Sometimes in low-pass networks the inductance value required is so large that the cost of an air core is prohibitive, and iron core must be used. Minimize this, in my opinion, as the hysteresis in iron cores will introduce sometimes audible distortion. Use laminated cores instead of solid cores if this type of inductor must be utilized. If you have the right tools you can hand-wind your own inductors (or unwind commercially available ones) to get the right values. But don’t try this unless you do have the right inductance measurement tools or you can never be sure of the results.
• Fourth, capacitors. This is where most of the controversy and witchcraft exists, and I’ll likely add to it with these statements. I say never, ever use electrolytic caps in a loudspeaker crossover. Yes, they will work, yes many people suggest them, yes there are special non-polarized electrolytic caps for just this purpose, and yes many if not all commercial loudspeakers use them because they are cheap. I’ve used them in commercial designs many times. But since this is a DIY design and you are less restricted by things like target price points and corporate profit margins, don’t use them. Instead, use the best caps you can afford. Save up for them if you have to. Use metalized polypropylene in woofer circuits and film/foil caps in midrange and tweeter circuits. There is no need to buy esoteric gold foil caps that cost $100 each unless you want to, and if you do want to go for it! But don’t cheap out and use electrolytics, you will hear the difference.
As an aside, there is no such thing as polarity on a non-polarized capacitor. It is non-polarized. Duh. That being said, some suppliers of high-end caps – especially those intended as so-called by-pass caps - will mark them with some sort of polarity color code or marking. This usually corresponds to the inside and outside of the cap’s internal construction and is not really polarity at all. I have never been able to measure or hear any difference using these types of caps no matter how they are wired into a circuit, but if you use them then by all means follow the manufacturer’s directions, it can’t hurt.
In order to verify that you are a human and not a spam bot, please enter the answer into the following box below based on the instructions contained in the graphic.