Do you have any datasheets on the drivers you are using? The reason I ask is that very rarely will the crossover parts align with a textbook calculation for the rated impedance of the drivers. Even though it may be an "8 ohm" woofer, the impedance at for example 2500Hz may be 11 ohms or more! With datasheets of the impedance curves and frequency responses, we can at least get close to the crossover values needed. The caveat here is that the frequency response will depend on the enclosure in which the driver is used to some extent, and you will never really know what that is unless you measure the driver in the enclosure yourself.

Let's do a case study of Zaph's popular new SR71 design here:

http://www.zaphaudio.com/SR71.html
The crossover schematic looks pretty straightforward

You have what looks like a standard 2nd order electrical crossover topology. The tweeter circuit has two added components for top end response shaping (C4 and R5), and the green boxed components are optional for impedance flattening.

Here is Zaph's measured response:

Looking at the meat of the crossover: C1, L2, L7 and C9; lets compare what Zaph used to what is calculated from a textbook crossover calculator.

Using the calculator here:

http://www.lalena.com/Audio/Calculator/XOver/
Inputting the rated impedance of the tweeter (6 ohms) and woofer (8 ohms); and inputting the crossover frequency (1750 Hz) and type (2nd order Linkwitz-Riley), we get the following values.

Related to the figure above, C1 would be 7.58uF, L2 would be 1.09mH, L7 would be 1.46mH, and C9 would be 5.69uF.

Here are the results plotted with Speaker Workshop.

This speaker, if built, would have absolutely no midbass or bass, and would have screeching high midrange and treble. So what happened?

The ER18RNX in this case has a rising response; to flatten this out requires more inductance than the textbook formula provides. Using the textbook inductor, the response of the ER18RNX stays at reference level until nearly 5kHz, where it drops off very sharply. There is a bobble in the phase response in this area as well, due to the ER18RNX having a cone mode with a peak at 5kHz. This causes the ER18RNX and the 27TDFC to be quite mismatched in phase and results in a suck out around 5kHz.

The larger than textbook inductor also provides Baffle Step Compensation (BSC) by beginning the electrical rolloff at a lower frequency, effectively boosting the bass and midbass. This is a trick that works well with woofers that have rising response, and is something that a textbook design would have difficulty implementing without much trial and error.

The 27TDFC in Zaph's design has been tuned to compensate for the woofers rising response and doubles to tilt the listening axis upwards to align it with the tweeter. The crossover rolls the tweeter off at a much higher acoustic slope by combining the natural rolloff of the tweeter with the electrical filter rolloff. The resulting inductor value is no where near what the textbook filter would recommend. It is nuances like these that make crossover designing both a science and an art.

I think you will find that textbook filters work best when the crossover frequency is more than an octave (double or half the frequency) away from the usable flat response of the drivers. Even then, it is still going to be hit or miss, with quite frankly more misses than hits.

A far better way to calculate the crossover frequency would be to use the free tools here:

http://www.pvconsultants.com/audio/frdgroup.htm
If you need further help, feel free to ask!

Regards,

David