How do find the voicing you want when you don't know what to tweak?

Are we talking about designing a passive crossover?

I find in crossover design there are usually a few components that have a nice effect over a broad range. Those are the ones to tweek.

If you were doing a simple two-way, for example, you might find that a series inductor for the woofer, varies response from 800 to 3000, that a resistor on the tweeter effects the treble range in total.

These are just examples, you need to investigate your particular network, either by adjusting some components, or if you are using a computer for optimization, by doing theoretical changes. You will find that certain components have broadband effect and others have more minor effects.

In the end it isn't much different than using a graphic equalizer, but the EQ sliders are defined by component effects or crossover point.

Good Luck,
David S.
 

AllenB

Moderator
Paid Member
2008-10-18 11:31 am
Hi speaker dave, yes I am running passive. I've been trying to work out what parts of the spectrum to cut or boost to change what I am hearing. I've tried the tutorials that tell you what changing each slider will do to the sound.

I'd like something that guides me to know what I want to cut, when I guess what to do I can't always get it right. Something like a tutorial on tuning a brightness and contrast control on a monitor for optimal results.
 
Hi speaker dave, yes I am running passive. I've been trying to work out what parts of the spectrum to cut or boost to change what I am hearing. I've tried the tutorials that tell you what changing each slider will do to the sound.

I'd like something that guides me to know what I want to cut, when I guess what to do I can't always get it right. Something like a tutorial on tuning a brightness and contrast control on a monitor for optimal results.

I see. The tutorials that describe what instruments or sonic effect each octave region has, are a good place to start with but you need to hear the effect yourself.

You can train yourself easily if your computer media player has a graphic EQ built in. I think the Windows Mediaplayers do. Start with a file of pink noise and try each slider up and down, up and down, until you get a sense of what the effect is. With noise you will notice that you can full boost each octave in turn and it sounds like moving up a keyboard, an octave at a time. Once you fix the effect in your mind on noise, try the same on music and human voice. Learn the difference between a single slider boosted and the same slider cut. You should be able to hear a significant difference for octave wide adjusts of 2 to 3 dB.

Someone with a lot of experience can listen to a speaker and guess the frequency of aberation, whether it is up or down and how many dB might be needed to fix it. You don't need to be that good, as long as you can put your finger on what you think is wrong and get there with some EQ trial and error. AB comparisons to a few other speakers of known quality might also help you spot the strongest aberations, assuming they are truly flat!

Hope that helps,

David S.
 
I think the key words in Dave's last post are "someone with a lot of experience.....".

The OP wrote "I'd like something that guides me to know what I want to cut..."
What you should use to help solve your problem are frequency response measurements. They can be an invaluable aid during the tweaking process because response curves provide a means to visualize what you are hearing.
 
Last edited:

theresa

Member
2009-08-07 12:06 pm
Without measurement equipment it is trial-and-error. With passive this can take a very long time along with a lot of money. I use an active crossover, the mini-DSP. So far I haven't setup the measurement equipment but changes are so easy that I haven't felt a great need for it.
 
Here's another tool you might consider if you're building a number of different designs and, if you don't want to go the measurement route and rely strictly on your ears. If things work out right, you could voice a speaker before installing the xover in the box. They are sometimes available brand new for <$100 on *bay.

http://www.vidsonix.com/vidsonixnew/boxlit2.pdf
 
Last edited:

tinitus

diyAudio Moderator R.I.P.
2005-11-24 1:47 am
I'd like a method of balancing timbre 'till I get the sound I'm looking for.

as said, its not so easy
and its not just about adjusting level

phase and "timing" are the key words

voicing ?
probably means something different to different people
personally, Im not even sure I like the word
to me it implies something that isnt really true at all

first step is to have a crossover that makes sense
and you need to understand how the component works
individually, and together
then you have to "know" your speaker, and each individual driver
and how every small change in crossover affects it

different crossover topology appears to relate mostly to speaker design, more than to the choise of drivers

so, we need to know your speakers and crossover, to have any idea about what you are facing

you could start with small adjustments to tweeter
it might affect other xo functions
could be a beneficial sideeffect advantage, or not
 
Hi,

Basically see the FAQs at : undefinition

also see : FRD Consortium tools guide

Simulation will let you see what you are changing and by what amount.
Done properly it would very likely indicate what the basic problem is anyway.

FWIW with some good test signals, once you get used to them, its
surprising how ~ accurate the ear is, identifying problems is easy,
fixing them no so easy.

rgds, sreten.

Zaph|Audio
RJB Audio Projects
Speaker Design Works
HTGuide Forum - A Guide to HTguide.com Completed Speaker Designs.
Humble Homemade Hifi
Click below to go to
Quarter Wavelength Loudspeaker Design
The Frugal-Horns Site -- High Performance, Low Cost DIY Horn Designs
Linkwitz Lab - Loudspeaker Design
Music and Design
 
Here's another tool you might consider if you're building a number of different designs and, if you don't want to go the measurement route and rely strictly on your ears. If things work out right, you could voice a speaker before installing the xover in the box. They are sometimes available for <$100 on *bay.

http://www.vidsonix.com/vidsonixnew/boxlit2.pdf

What and forego the pleasure of sitting on the floor with an old cardboard box full of parts?

We use to have a large cart at JBL that had crossover layouts on a plexiglass panel, plus switchable inductors, caps and resistors, plus a bunch of banana plug wires so you could set up a topology and tweek to your hearts content.

"Sitting at the console of the might Wurlitzer", John Eargle would say when he saw us using it. Very handy.

And for tinitus: which crossover component impacts "phase and timing"?

David S.
 
Hi guys,

The funny thing first : The vidsonix device is almost as expensive as a DCX 2496....wow!!!!!!

Anyway, it's very easy (and interesting) to create a passive filter without any calculator, just using a program as REW, Holmimpulse, Arta...(all free) and some electret mike. Just try to stick to the target curves, Arta is the best for this. On the way, you will discover the extraordinaries effects of high Q filters and how to equalise with RLC. You can even play with series filters (not easy to understand !).

For speaker dave, it's not the components that impact "phase and timing" but the design of the filter. Not the electrical topology, but what's really coming out of the speaker. Hence the justification of this method. To know if you are more or less time aligned just look at the step response. Only in some cases it can be nice, take a look at Music and Design first.

tinitus and sreten already gave you all the keys.
 
My bad, you are right, I was just reading this, on the pdf:
 

Attachments

  • vidsonix.jpg
    vidsonix.jpg
    13.1 KB · Views: 126

AllenB

Moderator
Paid Member
2008-10-18 11:31 am
I'd like to clarify the topic, I do have my system simulated, my crossover is not from a book and I can take all the normal measurements.

I feel that even if I had a 1/3 octave variable bandwidth EQ at my fingertips, I'd still be lost. I have played with EQs with various rooms and speakers and sometimes even cutting each and every band, then boosting in turn doesn't produce a good result.

This may be due to power response issues and it may be due to resonances. Sometimes I might guess that I need to boost 1kHz for example and it turns out to make things worse.

I feel that it is such a complex task that it needs to be done methodically. I've been doing this for over 30 years but such a method eludes me.
 

Pano

Administrator
Paid Member
2004-10-07 6:05 am
Panama
It's an interesting problem - "How to train your ear."

An interactive game might be the fastest way, tho I've never seen one. After many years of sitting in front of a console mixing live music, you develop a good ear for it. It often gets to the point where you don't think about frequency per se, but automatically reach for the right slider or button to fix the problem.

A video game where pitch is linked to movement or actions might be a great training tool.
Wish I had one to suggest.
 

AllenB

Moderator
Paid Member
2008-10-18 11:31 am
Hmm, I wonder can a person also train themselves to identify the difference between a level peak, a resonance peak and a reflection based boost? I feel I can begin to tell these apart only after I play with the sound a bit and work out what is and isn't happening.
 
first I'd start with getting the tweeter level to your taste - a variable L pad is handy here, put it in circuit, adjust to taste, measure the arms of the L pad and replace with fixed resistors... but not yet...
2nd, I'd deal with baffle step correction, this will require redoing the tweeter level after the inductor has been chosen.
3rd, then you may wish to play with notch filters, crossover slopes. All these things are to some extent interactive though... and simming with good data first will save you a lot of time...

actually, I'd probably tweak box tuning & stuffing first if I though it needed changing...
 
I feel that it is such a complex task that it needs to be done methodically. I've been doing this for over 30 years but such a method eludes me.

Hi Allen,

I've done it (made my living at it) for over 30 years also. It isn't always easy but here is how I've always done it.

The first part of the task is getting the measurements good. I always take a Linkwitz Riley approach and make sure the drivers are adding in phase at the intended listening axis. The response is smooth and reasonably flat before I start listening. I don't need to listen to itterations that measure badly, they never sound good.

At that point it is into the sound room and set up against a few competitors or references of known quality. About that: the goal is not for a new design to mimic a previous one, but to have some reference points that prevent drifiting off in an odd direction as you voice. As we develop products we would always have favorites that we felt were especially good (neutral). It is easier to EQ by comparing to a target than by trying for an ideal not present.

Listening is done only with material that is extremly well known. It might not be purest fidelity program, but it will be something you have heard a hundred times before. Injecting something new in at this time will have you chasing off in another direction. Program needs to be fairly busy, to excercise all of the spectrum. Solo acoustic guitar, for example, may be wonderful to listen to but the spectrum is sparse so it is hard to voice with. Pink noise is used also to reveal the most obvious problems. Voice is always key because we are very sensitive to the sound of it, even in people we have never met.

Room effects can create a problem. When you think something is getting close, exchange the positions of your new system and the others.

Find the key crossover components that impact spectrum over the broadest area. Pull some wires to your seating area where you can add and subtract to the nominal value of each key component while listening.

As mentioned earlier, your ear is a spectrum analyzer and you must train it to recognize frequency band and level with music as a stimulus.

This has always worked for me.
David S.