Verify crossover slopes by:
- Comparing them to a measurement of your DSP.
- Comparing them to my pre-measured templates.
- Calculating bandwidth.
I have often heard the advice: Avoid doubling up your crossover filters.
Most modern speakers come pre-baked with high/low-pass filters. If we insert an additional filter in our output EQ near these native filters the acoustic result will be steeper than either of the filters alone.
The danger is that we might already be planning a very narrow unity crossover between our main and sub and then be surprised when the magnitude looks like a brick wall and the phase like a pinwheel.
I’m not great at spotting these shapes, yet. To help myself practice, I measured a bunch of different crossover filters and stored them in Smaart.
Now I can compare these filter measurements to my acoustic measurements to make sure the slopes actually ended up where I wanted them. For example, here’s an alignment using Linkwitz-Riley 24dB/oct filters at 100Hz.
Comparing them to their corresponding filter measurements, it looks like I got what I paid for.
Going over some of my other previous alignments I did find some asymmetrical filters. In this alignment it looks like I ended up with a combination of 36dB/oct and 48dB/oct filters.
Here it looks like I got a 48dB/oct and a 24dB/oct combination.
Identify crossovers by measuring your DSP
You can measure your output EQ by making it the SUT (system under test). In Smaart, set the REF input of your transfer function to a loopback cable and the MIC input to the output of your DSP.
Store your trace and compare it to your acoustic result. Do they match?
Identify crossovers using these templates
To speed up the process, I already measured 80 common filters for you. Download them here. Drag them into your list of traces in Smaart.
Most of the filters are set to 80Hz so you should be able to accommodate many slopes with a little trace offset. I also measured all of the Linkwitz-Riley filters at 100Hz. If none of those match, you’ll have to measure your own.
Identify crossovers by bandwidth
We don’t normally think of a spectral crossover as having bandwidth, but it does have a clear area of interaction. This can be clearly seen if we load the traces into Phase Invaders.
Notice that the pink Sum trace has a clear starting and ending frequency. Those of you familiar with Bob McCarthy’s summation zones will also be familiar with this concept:
- Isolation: Magnitude relationship >10dB with low risk of cancellation.
- Transition: Magnitude relationship of 4-10dB with medium risk of cancellation.
- Combing: Magnitude relationship of 0-4dB with maximum risk of cancellation.
Let’s look at some example bandwidths.
This leads me to the following conclusion:
- 2nd order 12dB/oct filters ≈ 3.26oct bandwidth
- 4th order 24dB/oct filters ≈ 1.65oct bandwidth
- 8th order 48dB/oct filters ≈ 0.81oct bandwidth
Assuming your filters are symmetrical, you could use the following shortcut:
- If f2 ÷ f1 ≈ 9.5 then you may have two 12dB/oct filters.
- If f2 ÷ f1 ≈ 3.14 then you may have two 24dB/oct filters.
- If f2 ÷ f1 ≈ 1.75 then you may have two 48dB/oct filters.
The steeper the slope…
- The better maintained over distance.
- The more phase shift incurred.
- The faster and possibly more unnatural the transition.
Does it work in the field?
How to measure crossover filters in Smaart without an external DSP
Have you learned to identify crossover slopes in the wild (in Smaart)? How did you do it?