This is true. However, it misses what I think is the point of DSP. An LR4 electrical slope produces an in-phase handoff between drivers. You can easily fix the frequency response with IIR peaking filters.
If you're trying to cross at the extreme ends of the driver's range, I'd say it's time to move to FIR and use something steeper. At least that was my end state.
Yes, this is mainly the tweeter/mid. This is part of why I keep talking about VCad, the dispersion polar "heat map" is easy to see irregularity in the off axis behavior. Otherwise, you have to measure on and off axis, then look at the curves on the same graph, looking to see if the sloping curves fall away smoothly, while maintaining linearity.
Look below, ignore the specifics of this particular design. On the polar map you can see the dispersion widen out after the woofer crosses to the tweeter. It's visible in the curves as well, it's just not as obvious.
Like Busdriver02 shows with the graph from KEF Blade, you see that the tweeter takes over from the midrange smoothly, without changing level at any axis in the horizontal plane. If you cross your midrange/tweeter at - let's say 6kHz - you would have a midrange that has started beaming, and a tweeter that is still quite omni-directional. This means that the midrange started mostly playing more and more straight in front of the speaker but not to the sides - whereas the tweeter is still both playing straight ahead and also to the sides.What is the indicator of matched/mismatched response?
Sides will sound like a shift in total sound energy, because a beaming midrange also sounds like it is sending less energy into the room. And since the tweeter is not beaming at 6kHz, then it sounds like it is louder. The chart from the KEF Blade can be read as the colors representing the intensity of sound energy emanating from the speaker, and speakers who are poor at this, will sound like you can identify the drivers apart, which again is not something we want. The tweeter should be in a smooth and coherent combination with the midrange - which we can only achieve, if they send out the same amount of energy into the room, especially throughout the cross-over area.
Here's an example, where you see that on-axis is actually ok, but the curves below shows what happens off-axis, which tells you that the midrange/mid-bass starts to beam before the tweeter takes over, and therefor sounds lower or "pulled back" in comparison to the tweeter.
We sense a speaker like this as less neutral and with a "character" - something I would say sounds like a typical B&W speaker through the last 25-30 years, where there's bass and tweeter, but everything in between sounds lacking, missing detail, a bit "dull" and less open.
Last edited:
Indeed. But a good power response still start with the cabinet, drivers and the placement of them. So I kinda try to avoid extremes, to make the best of the basics. That way, it definitely feels like you have a better chance of success.
Ok, understand. Basically for indicators I am looking for the off-axis curves to be diverging evenly (slopes, intervals) and be close to ‘straight’, all this especially in the crossover region. The next question is what do I do if I see, in the crossover region:
How these indicators are interpreted and what corrective action to take (specific to the crossover)?
- very inconsistent intervals between curves
- curves that are too wavy
- inconsistent slopes of curves
How these indicators are interpreted and what corrective action to take (specific to the crossover)?
Roughly, by low order slopes you have a wider area to "reach" the next driver. The catch is, that you want to only add the filter slope that, together with your drivers slope, gives you a resulting slope that you measure. This could actually mean, that a low pass filter of 18dB on your midrange, gives you 24dB measured. And with your chosen speaker design and driver layout - you might have to live with some compromises. But try to measure the midrange/woofer first with no LP. Then you see what it does in this cabinet at those different angles I mentioned earlier on. By this exercise, you'll have much more knowledge as to where it would be wise to cross it, to get a smooth transistion to the tweeter.
Next you do the same for the tweeter - protecting it by only measuring with a HP of around 800Hz, so that it will not instantly burn up by over-excursion
and then you'll have the same sets of data, leaving you with two sets of data, which hopefully overlap.
Now, the trick is to use the filters to make those two sets of data smoothly blend. Look at the measured curves and see how steep they drop off, both on and off-axis. It's like an energy dispersion map of each driver, which you now need to sum smoothly.
What a lower order filter might help with, is if the midrange starts to beam at 3kHz, but the tweeter seems to not play well below 3,5kHz. Then you can "extend" one driver, by using a lower order slope, and use a higher order one, on the driver that can more easily reach the desired cross-over point.
When you do this, the off-axis energy goes down, and you might also lose potential max SPL.
Your tweeter has a tiny waveguide and the dome is not visible from the side, which means that you might have little less trouble with all the edges of the non-flush-mounted drivers and cabinet edges. These usually screw up all the off-axis curves.
But again, for ease of example. Take one on-axis and 3 more off-axis - repeat for the midrange too. Then we look at them together and take it from there...
Next you do the same for the tweeter - protecting it by only measuring with a HP of around 800Hz, so that it will not instantly burn up by over-excursion
and then you'll have the same sets of data, leaving you with two sets of data, which hopefully overlap.Now, the trick is to use the filters to make those two sets of data smoothly blend. Look at the measured curves and see how steep they drop off, both on and off-axis. It's like an energy dispersion map of each driver, which you now need to sum smoothly.
What a lower order filter might help with, is if the midrange starts to beam at 3kHz, but the tweeter seems to not play well below 3,5kHz. Then you can "extend" one driver, by using a lower order slope, and use a higher order one, on the driver that can more easily reach the desired cross-over point.
When you do this, the off-axis energy goes down, and you might also lose potential max SPL.
Your tweeter has a tiny waveguide and the dome is not visible from the side, which means that you might have little less trouble with all the edges of the non-flush-mounted drivers and cabinet edges. These usually screw up all the off-axis curves.
But again, for ease of example. Take one on-axis and 3 more off-axis - repeat for the midrange too. Then we look at them together and take it from there...
Excellent way of explaining: blending up the spacial energy dispersion maps, not just on axis magnitude response. Slowly learning the steps.
With painting done, cabinet mods are about 90% complete, so I will be starting DSP setup very soon! 🤞
With painting done, cabinet mods are about 90% complete, so I will be starting DSP setup very soon! 🤞
Last edited:
Yes. You'd find this out when you simulate.
Would be OK assuming they would look good if straightened out.
Agreed.
Draft guide updated. I think sections 1 to 5 are starting to make sense. Certainly will be useful for me when setting up the DSP.
Comments very welcome!
Comments very welcome!
Last edited: