"Designing crossovers without measurements" - so how about DSP?

I'm starting to design my first speakers, using DSP for crossovers instead of analog.
I've been reading the "Designing crossovers without measurements" thread and I'm trying to pick up as much as I can from it, but I'm wondering which parts apply when not building analog crossovers but using DSP for crossovers instead.
I'll probably be able to apply most of what I learn, but still - is there an equivalent article that could help me that's aimed at using DSP instead of building analog?
 
In general the rules are still the same. You still need to carefully match your drivers and design a proper enclosure. All that really changes is that you are applying the crossover before the amplifiers instead of after them.

DSP isn't magic. It's biggest advantage is that unlike passive crossovers, it can be reprogrammed at any time.

The thing is to not fall into the trap of "We'll fix it with DSP later"... that's not how good speakers are built.
 
I agree with Eric, it is very unlikely you will get this set up properly without using at least a software measurement tool (REW?) and microphone to determine equalization and crossover points.

Tuning by ear is just too hit and miss.
It's how you end up with smiley faces on the front of equalizers.
 
Last edited:
What measures flat sounds flat to me.

There is some merit to shaping to response to the Fletcher-Munson (equal loudness) curves if you want the speaker to 'sound louder' at lower volumes. Typically you want a smiley-face response for a speaker to have the tonality of a live performance when playing at lower listening volumes. This is what the 'loudness' button some amplifiers have is for - it boosts the bass and treble. However now when you turn it up loud it has exaggerated bass/treble. Better off training your ears to accept a flat system as being flat at various volume levels, imo.

If your speakers are actually playing at live performance volumes, then you want a flat response for it to accurately replicate the performance.

Obviously the voicing therefore becomes highly subjective because everyone has different preferences for listening volumes and because some listeners have been 'trained' into bad habits by non-flat systems. Zaph-audio put it well:
This tweeter sounds dull. Translation: "I've been listening to a speaker without baffle step compensation for 10 years, and this new design sounds different than my personal standard."
This tweeter sounds dull. Translation: "My last system had Dynaudio D21 tweeters, and now anything without a peak at 10kHz sounds recessed and doesn't have sizzle I need."
This tweeter sounds dull. Translation: "My current speaker has a woofer that did not have the breakup node properly filtered, and now I've accepted that type of sound as normal."

Zaph|Audio

There is also some merit to balancing frequency response and non-linear distortion. If a driver has high distortion, it can give the impression of the fundamental being louder too, so a speaker with a clean woofer but not so clean tweeter might ask for a slightly downward sloping response as frequency increases. A woofer with distortion in the midrange might ask for a slightly smiley face response. A speaker with exceptionally clean drivers should sound good being totally flat.

The only time I've felt the need to deviate significantly from a flat response, even with excellent drivers, was in my car, to combat road noise.
 
Last edited:
Guys, relax :) I didn't say you have to design perfectly flat speakers... It was an attempt at a little bit of humour. (Interesting concept, look it up sometime)

The reasons I suggest measurements instead of trusting a tainted ear are exactly those given by @TMM ... because we all have our own models of "correct" sound.

My own preference is for a slightly rolled off high end and just a titch of bass boost... And, yest, it is because that is what I'm used to hearing.

That doesn't mean I should design speakers like that. In fact, it means that I should not. The speaker itself should be as neutral and flat as possible. The "house curve" I prefer and room correction can be added later with EQ or tone controls.

And @TMM ... if my speakers were playing at actual concert levels... A) I'd be deaf as a stone by now and B) the noise bylaw fines would kill me. Common sense in all things, my friend.
 
Last edited:
I'm starting to design my first speakers, using DSP for crossovers instead of analog.
I've been reading the "Designing crossovers without measurements" thread and I'm trying to pick up as much as I can from it, but I'm wondering which parts apply when not building analog crossovers but using DSP for crossovers instead.
I'll probably be able to apply most of what I learn, but still - is there an equivalent article that could help me that's aimed at using DSP instead of building analog?
It depends what kind of measurements you mean, as Eric said you don't have to concern yourself with the driver impedances
 
Pygmy, your problem is connected with performing measurements. I suggest getting a book "Testing Loudspeakers" by Dr. Joseph D'Appolito. It is much much better to possess books than trying to separate common sense from common nonsense (look it up in this thread) that members keep producing here.
 
Last edited:
DSP just makes it faster and easier to design a bad sounding speaker.

Sure, you can tweak and hope that your "ear" is some kind of substitute for a proper set of measurements, but it isn't. Not even close.

When a calibrated mic costs $50 and measurement software is FREE there is absolutely no reason not to make proper measurements.
 
I think it is possible for a seasoned expert to create a good sounding 2-way loudspeaker without resorting to measurements. I do not expect that a novice or even moderately experienced DIYer to be as successful at doing so, even with a kind of "guide" to follow.

In any case, I really do not see why people don't want to involve measurements. If that is your position, then my advice would be to stick to single driver full range speakers.
 
I think it is possible for a seasoned expert to create a good sounding 2-way loudspeaker without resorting to measurements. I do not expect that a novice or even moderately experienced DIYer to be as successful at doing so, even with a kind of "guide" to follow.

To be honest, I rather doubt anyone's ears are that good. He might be adept at creating a speaker that sounds good to him but that is not any guarantee that moving it to a different location isn't going to end you up with terrible sound.

Speakers need to be flat (more or less) so that they can work in a wide variety of rooms and acoustic situations. Nothing says they have to be flat in the room, though. Once positioned and hooked up in their final spots, personal taste takes over and that is best added by EQ or Tone Controls, not in the speaker cabinet itself.
 
Did you guys check the thread referenced in opening post?

There are instructions to do simulations using factory data of T/S parameters, phase curve and response. Reference to do baffle diffraction simulations, reference to do bass reflex tuning simulations etc.

Tuning by ear is another story, not the subject here!

My way is to skip simulations by using common sense and gathered knowledge of how drivers, diffractions and directivity behave. Then I bought a Minidsp 4x10HD, old amps, a UMIK-1 and used REW to make own measurements with my drivers in my test baffles. Then in final baffle/enclosure to test different xos and find appropriate delays.

My way came because I am so poor in mathematics and physics, I just can't think by numbers and find too difficult to learn simulation programs with millions of options. Even ARTA is too difficult for me!
 
Member
Joined 2008
Paid Member
Pygmy -

Here's the secret sauce: you can use your laptop mic or almost any mic to collect all the information you need for your purposes today. I've posted my laptop mic response and its fine for such roles.

First, buy a Behringer DCX2496 and a bunch of Cannon plugs or whatever they are called now.

Then run broad sweeps through each driver - not too loud or too low for the tweeters. This will cause great screams of over-simplification from other posters when I say you can just leave your laptop where it is, providing it is in the same room. Lots of loose talk about XO considerations at this forum; of course you can't control most of these little points anyway but if you really like headaches you can think about them.

Combine your sense of what band each driver can handle well with you judgment about how to divide the music band and pick a crossover point and use 12dB/8ave slope (unless that is insufficiently steep to protect the tweeters from low freq). Very learned mathematicians and physicists have proposed different XO patterns and the Behringer lets you mix and match three pattern and a whole bunch of slope interactions. Any time you waste trying to compare all of these will be less time listening to music.

Adjust shelving volume levels and you're done.*

B.
* OK, you're done with Part One
 
Last edited:
There are instructions to do simulations using factory data of T/S parameters, phase curve and response. Reference to do baffle diffraction simulations, reference to do bass reflex tuning simulations etc.

Simulations will get you close. But ultimately you need to test the whole speaker/crossover system as a whole to ensure things are as they should be. I seriously doubt it's possible to do that "by ear".

My way came because I am so poor in mathematics and physics, I just can't think by numbers and find too difficult to learn simulation programs with millions of options. Even ARTA is too difficult for me!

Download XSim, it's a simple to learn, fee program. You can probably figure it out in 10 minutes or less... but be sure to explore the menus, there's some real nice stuff hidden in there, too... Get it HERE
 
Last edited:
Member
Joined 2008
Paid Member
Simulations will get you close....Download XSim, it's a simple to learn, fee program. You can probably figure it out in 10 minutes or less...

Are you confusing building a sub cab with dialing-in a DSP crossover?

A crossover is nothing but choices of freq, slope, filter pattern, and shelving volume.

At the simplest, you'll be quite fine knowing nothing but the published FRs of the drivers and then adjusting volume by ear.

B.
 
Last edited: