Introduction to designing crossovers without measurement

Distinguished members of this forum!
I have a question for this thread - is it meant to develop a passive crossover circuit without measurements AND without simulation software? If this is the case i wonder if it would be possible to come to a similar result like i found using Boxsim and Visaton chassis (used only as a demo because of the good measurements of Visaton used for Boxsim with the 19 measurements from 0 to 90 degree in 5 degree steps) for a fairly simple demo two way bookshelf speakers with this chassis:
A )
Midwoofer Visaton AL130-8
B )
Tweeter with waveguide
Visaton G25 FFL with round Visaton waveguide
One good thing about this choice is that you get a quite good time alignment with a very simple baffle. Here the screenshots: Demo-diyAudio-2way-bookshelf-FR-Capture.JPG Demo-diyAudio-2way-bookshelf-housing-Capture.JPG Demo-diyAudio-2way-bookshelf-CO-Capture.JPG
added additional screenshot from german version with the theoretical filter curves Fc 2000 Hz LR 4 81 db


    99.7 KB · Views: 81
Last edited:
Yes, with measurements it can be taken further indeed, although I disagree with the value to this extent of phase alignment at this early stage of the game.

During the '80s I wrote some software which I used to precision align phase between drivers, but it didn't bring me satisfaction after several attempts. I learned that while phase aligns at one place, there is a room full of sound that needs to be accounted for at the same time. This will limit the ability of phase alignment on-axis as being able to either make or break a speaker's response. Phase distortion, of itself, has also been shown not to be a major factor.

What matters in the end is the frequency response. This of course is phase dependent in the direct field but this then becomes virtually irrelevant by the reverberant field. At this early stage of the game, it is within the realm of being judged by ear or using a broad frequency response measurement.

All the best with the next step, PRTG, I'll drop by next chance I get.
It is more than 11 years after this post was written I chance upon it and could not stop myself noting a point I have. It is my experience that the best sounding speakers I have ever heard and music out of which moves me, and also every single other listener who has chosen to listen to these, do not measure flat at all. These speakers are actively driven, and even with a wide range of settings on the crossover, it is easy for me to move into meditative state within the first few minutes listening to them. Whether out of sheer luck, but these were built after a couple of years of iteration, without any major software simulated measurements. Perhaps the world obsessed by high precision electronics and feedback and computer simulated anechoic chamber measured speakers, this post from Allen shall stand the test of time, maybe forever.
Thanks for sharing I'd love to hear more on the speakers you mentioned but know that what you are sharing and asserting is anecdotal and isn't in alignment with mountains of hard evidence that has been collected over decades. Of course nothing is ever 100% but can still be true a majority of the time and within a general range. What am I trying to say? The details and evidence matters .
I enjoy a lot fullrange loudspeakers with current drive amplifiers which correct the sound of fullrange drivers "automatically".

Then putting aluminium foil on paper cones up to sandwich construction brings high end sound to these type of drivers which are naturallly time coherent if the frequency response is linear.

See for aluminium foil on paper cones:

Current drive amplifier DIY (thread started by me):

With positioning of fullrange drivers close to a back wall or a corner the baffle step in the frequency response can be cured without any crossover network.

beschichtete membran.jpg


  • beschichtete membran2.jpg
    beschichtete membran2.jpg
    163.9 KB · Views: 54
  • Like
Reactions: 1 user
diyAudio Moderator
Joined 2008
Paid Member
It is my experience that the best sounding speakers I have ever heard and music out of which moves me, and also every single other listener who has chosen to listen to these, do not measure flat at all.
There is more than one way to take this, and I'm not disagreeing in general.

What I failed to explain (but only touched on) is that frequency response can be measured in any direction. Also, the relative responses hold more information than one alone, such as unwanted reflected and diffracted energy as well as the collected room reverberation. The result might end up depending somewhat on the conditions, rather than what simply looks flat.

The other way to see this is less scientific, but a forced non-flat response might be used as a tool to distract from other issues. A very well balanced speaker may not reach it's goals if carelessly placed into a new room. The result may be disappointing. However, forcing the response to some pleasant state of EQ is probably not something an experienced builder is likely to settle on over time.
  • Like
Reactions: 1 users
I once saw visiting a professional for loudspeaker crossovers.

A big developer crossover board where you could make the dimensioning of the crossover simply by switching the parameters for coils and condensers together.

He sat then at listening position just switching the needed parameters.

Had a reference loudspeaker to compare with.

So you can do it without measurements.
  • Like
Reactions: 1 users
Measurements- frequency graphs, do they mean anything?

Philharmonic BMR


Misco JC80PA page 8 Linear X

same speaker

Are these Misco graphs the same?
People like the Boenicke W5 enough to clone it. I look at the frequency response and wonder why.

They must like the sound, that's all that matters.
Last edited:
diyAudio Moderator
Joined 2008
Paid Member
Context matters. A single response measurement with no context may have little to no meaning. Consider for one thing, that with an equaliser you can change that response to anything you like.. however you can't change the acoustic character of a speaker with an equaliser.

A set of polar measurements can begin to reveal the acoustic nature, and is useful for identifying the presence of some issues. On the other hand, it doesn't readily give a clear picture of some of the specifics of those issues.

At the end of the day you may find yourself voicing a speaker by ear. This way you are navigating the acoustic character, including the amount of audibility of acoustic flaws. These are something for which there may be some research data but it doesn't show in measurements unless we put the two together.
I have a three-way design I am looking to simplify from a digital version that uses electronic crossovers to a passive version. I have the desired crossover frequencies, but not really sure if I need measured impedances from the speakers or if I can just calculate the filter values without speaker measurements. Based this thread (I have read all of the posts) it sounds like I should have a pretty good idea of the Z(f) for all of the drivers. But I think I will be out of resonance for everything but the woofer. The crossover frequencies should be around 284 Hz and 2.3 kHz with some shelving. I downloaded XSim but it has no frequency response or Z data for any of the drivers I have. I can measure them, it's just a question of should I, or should I just assume flat response and build the crossovers accordingly?


  • IMG_3153.jpg
    382.8 KB · Views: 53
I used digital eq with measurements for fullrange drivers.

Since I discovered so called CFA (=current feedback amps) amplifiers I feel much less the need for frequency response correction with eq or passive crossovers.

Current drive not only enhances automatically bass and highs but also lowers distortion by 10 to 20 db!

And it does something to impulse response what really shows that frequency response can become less important if all other parameters get better due to appropriate amplification.

Usually with CFA amps the bass and treble attenuation at the preamp does the rest in my ears.

Also I do tricks with paper cone distortion by diy using tinfoil on them:

diyAudio Moderator
Joined 2008
Paid Member
@noahc you could trace the factory impedance measurements to save you making measurements. If you have any problem with that you can post the factory plot and I'll see what I can do.

Presumably you'd then be able to replicate the active filter response to use as a target. You won't need response measurements done with a mic just to clone a filter.
Thanks AllenB I could measure the impedance as I have DATS (V2 I believe) - I was planning to do that anyway to see how well the woofer is matched to the port. But appreciate the help and I will see if I can track down the charts.

Freedom - completely agree that applying current not voltage is the way to go. Unfortunately I only have voltage sources right now. What hardware are you using for that?
Well I think I finagled it - single order on the filters for low/med/high (RLC on woofer, RLCx2 on mid, RC on tweeter). Had a more elaborate arrangement with an additional order on each and Zobel but the cost was out of hand. Minimum impedance about 4 ohms. I could probably go steeper on the woofer but it's not critical - it's a flashlight at the higher frequencies and stops being relevant quickly; it's beaming under the tables at that point and losing response. A change in R can change the slope on the woofer at the cost of higher dissipation if it really is needed. The woofer is 1000W so components end up getting costly and bulky at those levels.

Thread was very helpful on working through the process.

diyAudio Moderator
Joined 2008
Paid Member
I notice you've chosen not to copy the previous electronic filters.

1. Based on that, you will now want responses. Nearfield (I assume near to the cones?) is going to ignore driver directivity and ignore the cabinets. You won't have baffle information and your measurements will be incorrect at higher frequencies.

2. Zobel networks as described in this thread are most useful when you aren't using a simulator. They still work fine, but sometimes you can find a simpler alternative.

3. Xsim will give you better data if you enable phase for each range.