Introduction to designing crossovers without measurement

Thanks GM for the links. Really interesting to learn of all the different types of components.

I'm still trying to wrap my head around piezo behaviors. I now get that putting a resistor parallel to it turns it into a normal tweeter with the impedance of the resistor used - as far as the crossover filter is concerned. So e.g. both 6.8uf+8ohm and 2.2ohm+22ohm gives roughly the same XO at 3khz. Is that correct? The difference here is that the 22ohm version will actually have higher output, because the tweeter will receive higher current? How does one determine the size of this resistor vs the output of the tweeter?

And similarly, if I just want to pad down a tweeter with a resistor, do I just add it in series before the capacitor/XO? And is there an equation for ohms vs db reduction?

Finally, what frequencies would the cap/XO filter if there was no parallel resistor across the piezo at all? with the resistor, I can use the standard XO calculators to work out the XO point by inputting the resistance of the resistor used as the impedance of the tweeter. But with no resistor, how would the XO behave? would it filter anything at all?
 
diyAudio Moderator
Joined 2008
Paid Member
both 6.8uf+8ohm and 2.2ohm+22ohm gives roughly the same XO at 3khz. Is that correct?
Yes, that's how it works (I understand you meant to say 2.2uF+22ohms)
that the 22ohm version will actually have higher output, because the tweeter will receive higher current?
No. By using a resistance which is much smaller than the equivalent impedance that the piezo has, you create a reasonably fixed Voltage across it at that point. Going to a lower resistance doesn't significantly change this. Going too low would only waste power.
if I just want to pad down a tweeter with a resistor, do I just add it in series before the capacitor/XO?
In the tutorial I chose to put it before so it is out of the way of the filter. Actually an L-pad with two resistors placed directly next to the tweeter is also a popular choice.
And is there an equation for ohms vs db reduction?
In simplest terms an 8 ohm resistor cuts an 8 ohm tweeter by 6dB, and you can guess from there for lower values. This resistor is often tweaked to find the best value because listening is more revealing when not measuring.

Even when using full measurements it's difficult to accurately guess where to set the treble level. in any case, I can give you a more exact formula if you're interested.
Finally, what frequencies would the cap/XO filter if there was no parallel resistor across the piezo at all? with the resistor, I can use the standard XO calculators to work out the XO point by inputting the resistance of the resistor used as the impedance of the tweeter. But with no resistor, how would the XO behave? would it filter anything at all?
You can't really use a capacitor to high pass a piezo because it is also capacitive. You'll just lower its level but won't cut the bass. Some run them without a crossover but when you want to add a crossover, it is useful to consider elements in parallel with the driver to modify the nature of the impedance at that point.
 
  • Like
Reactions: 1 user
Thank you for explaining all that. Very clear. I've been trying to read up on tweeter attenuation. I understand that an L-pad is used after the XO which can attenuate the tweeter while keeping constant impedance, so the XO point stays where you want it. What I don't understand is if exactly the same can be achieved by using a potentiometer/variable resistor before the XO. Why is the L-pad preferred?
 
diyAudio Moderator
Joined 2008
Paid Member
An L-pad uses fixed resistors whereas a potentiometer is variable. It is possible to have a variable L-pad and that is different from a potentiometer because it's designed to keep the impedance the same, which a potentiometer does not. It does this by having separate resistances.
 
diyAudio Moderator
Joined 2008
Paid Member
That can work but I wouldn't recommend it. I have to ask what it would solve? It still interacts with the crossover, even though not as much...

However your amp is going to feed bass through it. This not only increases the load on the amp, but potentiometers have varable power handling and little of it in general.
 
diyAudio Moderator
Joined 2008
Paid Member
Oh I see. That's different to what I was thinking.

In cases where the wiper of a potentiometer is connected to one side of it to make a two terminal device, you have a variable resistor. This would make sense as a tweaking device.

With regard to power handling, consider that when the resistance is turned down and the current increases, that higher current flows through a small portion of the device. There are power variable resistors available that will work, but the types that most people know and have seen are low power devices.
 
So I've been further playing with my idea of Dayton Audio reference speaker crossover and this is pretty much the best compromise i could put together to get the highest possible efficiency and impedance not dropping much below 4 Ohms

Screenshot from 2022-10-15 15-23-16.png


Screenshot from 2022-10-15 15-23-23.png


This is just a theoretical model since I don't have the drivers neither have started building the cabinets. Will probably start next year after we renovate the living room and will see how much funds will be left for this.

If anyone has any further ideas about this crossover design I'm open for criticism.

I would do this in Vituixcad but unfortunately can't get it to work on my linux PC so if anyone is interested in further developing this design I'll provide the room dimensions, speaker posittion in the room and rough baffle design.
 
Last edited:
Hi AllenB, I have read your tutorial very carefully. Very interesting and, above all, very useful. Now I would like to ask for your valuable suggestion. I have 3 Philips speakers, quite dated (80s), but new and never used. Unfortunately, for these speakers, the Thiele-Small parameters are not available, but there are the basic parameters. I would like to build a 3-way closed box, with active filter. According to your indications and your advice, the ideal solution would be to make the speakers work at 2 octaves from the resonant frequency, so:
AD0163T8 tweeter fs 1,300 Hz - frequency range 2,000-22,000 Hz
AD0211SQ8 midrange fs 370 Hz - frequency range 500-5,000 Hz
AD12650W8 woofer fs 23 Hz - frequency range 35-2,000 Hz
following your advice, the tweeter should be crossed at 5,200 Hz, but, with the frequency range of the midrange up high, at 5,000 Hz, I thought I'd cut it to 5,000 Hz or even a little less.
For the crossover frequency between woofer and midrange, I thought I'd cut to 1.480 Hz. Would a frequency of 1.480 Hz be too high for the 12 "woofer?
What do you suggest? It might work? Thank you so much in advance for your invaluable support!
 
diyAudio Moderator
Joined 2008
Paid Member
The tweeter can be taken lower if you find the mid is not filling in around the crossover.

A 12" woofer should be happy crossed at 800Hz. They can be used at 1,500Hz but it is close to breakup and close to the directivity limit, but the mid dome will help to fill in so you can experiment.
 
I do not know if it was mentioned in this thread already. There is one elegant possibility of developing a loudspeaker without measurement equipment:

Make a crossover developer board with switchable crossover parts for bass and tweeter (for a two way system for example).

Plus you need a linear reference loudspeaker to compare with.

You just then sit and listen, switch the needed crossover parts until you reach your optimum in sound. By making A/B comparison all the time.

Later you can do some fine tuning if necessary.

The better the developer crossover board is made the better you can work with it.
 
For my part I only build loudspeakers with fullrange drivers and EQ them digitally to linear response with measurement equipment. These fullrange drivers usually have a linear frequency and linear phase response. And you do not need to take care of Thiele Small parameters as you can install whatever target curve.
 
Today you can use EQ on your smartfone with EQ apps like "Extra EQ" a 31 band EQ for free.

With the "Spectroid" app you can measure frequency response. From YouTube you can get white noise test signal.

So you can do it with two smartfones. One as a source with EQ - the other for measuring.

Most important are the mid frequencies, they have to be linear. Like 200 Hertz to 8 khz. Here every microfone is more or less linear, also in smartfones. In the high and low frequencies always people manipulate or adapt the sound to room response and taste.
 
diyAudio Moderator
Joined 2008
Paid Member
I once built a switchable developer board and it's a very desirable thing to have. It will only identify complex issues such as response wiggles if it works on a sufficiently complex circuit itself, so it can be an involved proposition. One way around this is to use it together with an equaliser. When using separate amps as you might with a digital arrangement, you can equalise each separately. This is enough to work through issues but some will still be difficult to identify.
Plus you need a linear reference loudspeaker to compare with.
I just compare to the real thing. Speakers can sound as good as that but not just any speakers, not just through good crossover work.
 
diyAudio Moderator
Joined 2008
Paid Member
While there are some variations (which we could list).. if a speaker isn't good enough to be within the ability to compare it to sounds you hear naturally then there is something wrong with it.

You don't need to compare, you just know when something is real. For example if you were listening to average speakers in a room and I walked in and started talking, you'd just know my voice wasn't coming from the speakers.