Least Audible Midrange Crossover Type?

[cuibono]

Lately I've been using a computer to develop crossovers and EQ for multi way speakers. It is very powerful, very easy to use, and allows lots of fine detailed control (and the software is very affordable!) - Reaper is an excellent place to start. It even includes the plugins one needs - REAPER | Audio Production Without Limits So I'm learning I have more and more possibilities when it comes to crossovers and able to reach more and more specific acoustic targets.

I'm considering moving my 3-way 'up' to 4-way, but this will require dropping a crossover smack in the middle of the midrange - somewhere around 700Hz. One of my top concerns is power response, second only to the on-axis frequency response.

JohnK has done some nice work on crossovers and their power response: Power Thanks John! What is obvious is that all crossovers introduce dips in the power response, some more, some less.

On the second page of the above link, John states his experience has been that LR2 crossovers have a more neutral presentation (compared to LR4), even though they have a larger dip in the power response. I've heard other people say the same.

Recently though, Floyd Toole published some research that may be comparable - although it may not be! He studied the minimum audibility of resonant peaks in a speaker's on-axis frequency response. The upshot was that lower Q peaks were audible at much lower levels compared to high Q peaks - aka, it took a much louder high-Q peak to be as audible as a low-Q peak. The reason this might be related to dips in power response is the different order crossovers look like they are changing the Q of the dip - with higher orders having much narrower dips - while not increasing their dB level. If the above study can be extrapolated to the power response, it would suggest that higher order crossovers should make the power response dips less audible.

Sure, the best crossover is no crossover, but assuming we have to drop one in at 700Hz, what crossover order/type do you think sounds most neutral?

If possible, lets keep the conditions more 'ideal' - there are no issues with acoustic centers, driver spacing, beaming, limited excursion, group delay nonlinearity, absolute polarity, etc (if possible). Also, 1st order slopes are not being considered. Sorry

[cuibono]

DAMB! I forgot LR2! Can a mod add it? I can't seem to..

[planet10]

There are too many other factors involved to be able to choose a specific topology... IMHO the best midrange XO is one that keeps the XO out of the 300-5k range.

dave

[cuibono]

Thanks! I forgot both those options.

[tinitus]

Originally Posted by planet10

IMHO the best midrange XO is one that keeps the XO out of the 300-5k range.

dave

I would say the best tweeter xo point is 300-500hz

If looking solely at the crossover, a 2way will always be the better one
Only reason to choose a 3way would be if you use much better drivers that doesnt work any other way, or because of other priorities

In a 4way its your complete crossover in relation to driver slopes that determine you xo points an component values
Nothing else matter in this respect
You cant just choose a crossover at will
It depends on ALL your drivers
And the acoustic slopes you can achieve
On bass I would suggest electrically 18db, but modified
The rest would electrically all be a mix of lowpass/6db with paralel RC and highpass 6db with paralel LR
You may need LCR on tweeter ressonance
Use series attenuation
Whether it works only depends on a lot of fidling back and forth

Its beyond doubt that fewer components will always sound better, as long as proper function is retained
So, a certain minimum of components will always be required
And I would never attempt such without correct physical driver phase allignment

[cuibono]

Originally Posted by planet10

There are too many other factors involved to be able to choose a specific topology...

Originally Posted by tinitus

Its all a matter of priorities...

Maybe you guys could give a little more specifics on how you prioritize things? What factors do you look at, assuming you have to crossover somewhere?

Clearly, my priorities are on axis flatness, and off axis regularity (leading to flat power response). The reason for being of this poll is to hear how people feel about the dip in power response caused by different crossover topologies, but also what crossovers people consider 'neutral'. Details, please!

[Geek]

And for the non-Windows crowd, if you want to play with VST/LADSPA EQ plugins, here's a GNU way of doing things: LMMS - Linux MultiMedia Studio



Cheers!

[tinitus]

To be honest, I dont know how or why it works
I only keep trying until it works
But stay clear away from anything that doesnt follow basic rules
I usually try to use a very simple and logical methodical approach to tell me if Im doing right or wrong
My 3ways ended up quite good, after 5-10 years of work
Now sold to a dear friend

I think this xo is up to date

[fredex]

I am no expert but all crossovers seem to do damage. Maybe the crossover type isn't the problem as much as the physical differences and separation of the drivers. As a note rises or falls it sweeps across two different sounding drivers in two different points in space, and across an electrical network with phase changes.

[dwk123]

If Reaper allows flexible enough parametric EQ, you might consider trying a Deuland topology for a 3-way system. This is a topology that uses coupled parametric equations so that the entire 3-way system is defined by a single parameter which basically controls the midrange spread. The really interesting aspect of this is that all drivers are in-phase across the full bandwidth, and the system exhibits a single phase rotation making it similar to a LR2 2-way.
I realize this doesn't get you to your 4-way, but it is suggested from the perspective that it may not be the midrange xover by itself that is responsible for particular audible artifacts, but the overall system response. The Deuland seems to be unique in treating the entire system rather than the low-mid and mid-high xovers seperately. It's tough to realize passively, but with Reaper it may be possible to experiment with reasonably easily.