Aww Jeez, another moronic post (pls forgive me but)...

[ptwining]

I know this is idiotic but I have honest questions about incremental gains. If I am to advance I need to understant these things - I appreciate your patience - really!

Let's dismiss the obvious - I know I'm not measuring yet (I have SW but the Wallin jig is taking more time to get to than I thought! )

In any case. Y'all know what I'm up to but here's my question. (those who don't can catch up on my previous posts about my project so I don't waste the peoples space and time).

I chose my drivers based on an established design - they're not exactly the same (the tweeters are different) but the manufacturers charts are virtually identical - the woofers (Extremis) are extremely flat regarding impedence - the mids (CSSwr-125) are the same - and the Dayton silky virtually the same as the established design requirement. The cab is to spec.

You already know I'm running them on a PE Dayton XO at 700/5600.

I already hear that the mids are too hot because the design calls for a 3/4000'ish XO point instead of my canned 5600 but I'll fix that later.

Those points are way in the middle of each drivers range, nowhere near the nasty roll-off and/or break-up (according to the manufacturers charts).

I have no Zobel yet because (here's the catch) the Extremis impedence curve is soooo flat - why do I need it?

Next - and here's the point of this post -

When you choose XO points for whatever reason (maybe measurement, maybe charts, or your own preference) Please tell me what else that XO will do for you (aside from LPad and Zobel)....?

Believe it or not I've read piles of XO design stuff - I use Win ISD and XOCalc, etc - I know it seems unlikely - but I fail to see the extra magic (aside from component selection - and all the benefits therein) that these straight-forward calculations yield. And yes I understand the impedence curve.

I am not discounting the art of voicing (something I am years away from) but as to matching commercial loudspeakers - is it really relevant?

Once again, I beg your patience, I have equipment on order and am building my measurement gear but for now - I have to say the following.

If you PAY ATTENTION, choose drivers that match a given canned XO and you build the right-sized cab - why should the result suck?

I have spent hours at Tweeter (a US based HIFI joint) - and a number of esoteric HIFI dealers and have found that my canned, knock-off works okay - not perfect by any means - but worth every dime of the two or three grand the retailers want.

So, here's the meat of the issue. If I get close - just how close do I need to be? Where does the point of diminishing returns come? How much better can I get with a given design if the XO points match the curves and the driver efficiencies are real close (given that cab specs are on the beam).

Am I simply deluded?

I have to admit I have a hard time buying that conclusion.

Who knows, maybe I'm partially deaf. Maybe dense too...

Regards,
Tom

[bjorno]

Next - and here's the point of this post -

When you choose XO points for whatever reason (maybe measurement, maybe charts, or your own preference) Please tell me what else that XO will do for you (aside from LPad and Zobel)....?

Tom,

Only one thought so far about off axis requirements.
I think its important to chose the cross over frequency from the imaging point of view too.

Preferences can vary from one person to another. Is this question in line with what you think is important?

B

[ptwining]

I do.

That is why, at the end of the day I fully intend to measure and do the right thing.

However, these things sound pretty good ( I realize they may measure poorly!).

The niceties (not to take the fine points lightly) are fine tuning.

I guess my real question is as follows :

When you choose XO points (based on the drivers TS specs) how much better than that can you get?

Regards,
Tom

[preiter]

Typical features found in crossovers are the crossover function, baffle step compensation, zobel circuit, l-pads for level matching and notch filters for correcting peaks.

Factors to be concerned with are flat frequency response, relative phase of the different drivers near the crossover frequency and overall impedance.

It is true that the better the characteristics of the drivers, and the better the drivers work together, the simpler the crossover can be.

One reason you can't just design off of the manufacturer's spec, however, is that the behavior of the drivers will change when you put them in the cabinet.

In the end though, it's all about how it sounds to you. If your speakers sound good to you, then don't worry about anything else.

[ptwining]

I understand the necessity for measurement, that isn't my question.

Furthermore I understand (conceptually anyway, I do, after all, have a degree in first order logic systems and have spent twenty years as a quantative analyst in the options market) the quantative aspects of XO design. No question that the engineering aspect is crucial to a successful design.

Jeez, have I got that all out of the way yet?

I suppose what I am trying to ask is for someone to explain to me - apart from the measurement, engineering, and quantative necessities required for a world class design - just what the basic requirements are.

Most XO calculators on the net or in books are simple formalism (that is: equations). I find it hard to believe that it is that simple.

Any second year EE monkey can do that. Hell - I can do that. But it doesn't seem quite enough, does it?

Although there a number of variables involved it ain't like solving the three-body-problem (it's impossible - look it up), is it?

Or is it? I purely don't know. But I'd like to (at least have it explained).

If it is a pure matter of handling a number of dynamic variables then let me know - I used to design real-time options pricing models. No problem. But there is a veil of mystery around the issue of XO design that no-one wants to pierce. That is frustrating for me.

Is it pure maths or not?

I can't simply price options or futures such that I like the answer, favorable gamma-theta ratio or not. There is a right answer (well, within limits). Is XO design like that?

Now, I know - wait - if it sounds okay to me it's good, right?

No disrespect, but that must be nonsense.

[edit] I suppose I sound quite silly- I apologize. I am really curious though. I want to understand...

Regards,

Tom

[preiter]

Head on over to www.zaphaudio.com and read the project write-ups there. Some of them have good descriptions of how he came up with his crossover designs.

[kvholio]

the zaphaudio-link is a good one.. good reading.

imo, the "magic" largely is in doing measurements and interpreting them.

you state:

I already hear that the mids are too hot because the design calls for a 3/4000'ish XO point instead of my canned 5600 but I'll fix that later.

How did you come to that conclusion ?
Just because someone else thinks 4000 Hz is the ideal crossover-point doesn't necessarily mean a crossover at 5600 Hz will do much worse.
Maybe you just need to add a zobel to your midrange-driver to make your canned crossover work correctly.

If you PAY ATTENTION, choose drivers that match a given canned XO and you build the right-sized cab - why should the result suck?

It wouldnt necessarily s*ck, it could work.
But to me, FIRST deciding on what canned crossover you use, sounds like engineering backwards.
Its like you buy a gearbox with given gear-ratios, and then decide on what engine you will use to match the gear-ratios.

There are no easy answers to your questions, to me this is just "learn as you go", and it seems you're already doing that.

With kind regards,

Klaas

[Shaun]

[I started writing this before the other posts appeared, but I've decided not to amend it.]

The inputs to designing the crossover are: driver impedance (curve), frequency response, phase and acoustic offset. Sometimes one needs a zobel or other impedance flattening circuit, but for the most part, no (in my first design I did a thorough job of flattening the impedance curve, but learnt later that this was overkill). Also add to this chosen crossover frequency and roll-off rate (order).

Considerations of what crossover frequency to use are: driver bandwidth, distortion and off-axis response (when mounted). Choices of what topology to adopt can be influenced by the driver useable bandwidth, distortion (relates to useable bandwidth) and personal philosophy.

Getting the loudspeaker to sound good can be simplified by (1) avoiding overdriving the drivers, (2) ensuring smooth power response, (3) applying judicious baffle step correction and (4) getting a generally flat (there are variations on this) frequency response curve.

Your approach so far will get you close, but you’re actually flying blind if you don’t do measurements. Reason being that you’ve used a textbook approach, which assumes a flat impedance curve and flat frequency response. For instance, in practice, if you wanted a 4th order acoustic response, you would have to factor in the driver’s natural roll-off, so your electrical filter might end up being only 2nd or 3rd order…

I don’t know what else to advise you, other than to look up papers by Harman International (on their web site, sorry don’t have more specific info).

Good luck. HTH

[ptwining]

Thank you all for your answers!

Zaph is indeed a good source, I'm diggin in!

You have confirmed a lot for me and made the whole issue clearer. I have much to learn but I've got a better handle on this thing now.

Thank you very much indeed...



Best Regards,
Tom

BTW:

How did you come to that conclusion ?
Just because someone else thinks 4000 Hz is the ideal crossover-point doesn't necessarily mean a crossover at 5600 Hz will do much worse. Maybe you just need to add a zobel to your midrange-driver to make your canned crossover work correctly.

I'll give a shot. I thought zobels were primarily for woofers...

[kvholio]

Tom, the reason i mentioned the zobel for your midrange-driver, is that your canned crossover is most probably a "textbook-crossover",calculated for 8 ohms drivers.
If you look at the midranges' datasheet
http://www.acoustic-visions.com/~aco...r125/wr125.pdf
You'll see that the midranges'impedance at 5.6 khz is already 15 ohms.
This will cause a mismatch between your crossover and your driver.
Shaun is generally right about zobels imo.
If you design your own custom crossover you probably dont need a zobel, because you can shift component-values to adjust the filter for correct behavior.
Sometimes the increasing impedance will even help to get the desired rolloff.

Good luck and let us know how it works out.

Klaas