determining XO "quality"

[augerpro]

So I'm sitting at work tonight getting familiar with Soundeasy Crossover CAD. My first DIY project will use the HiVi M3N and Dayton ND20 in mtm. Now I have not measured the drivers yet, I'm just using the info imported from PE. So I'm not worrying yet about BSC, driver spacing, diffraction, etc. I'm just trying to build a simple XO. I built what I considered a great XO from a FR perspective and phase was nearly perfect (assumes AC is the same for the drivers though). I've included the screenshots for this XO.

After that I decided to use a "textbook" XO using the 3rd Butterworth formula from Vance Dickasons book. This is what the XO values would be giving the Q=.707 and I wanted to compare my first XO values to these and see how bad my amatuer effort was. Well the textbook XO is totally useless. So bad I'm not even gonna post the FR. Now knowing that no XO stays with the textbook formula, how does one determine the quality of "real" XO? The only thing I could think of was plotting group delay, which actually looks pretty good around the XO frequency. I mean if the FR and phase look good what things can I look to determine Q, group delay, ringing, trasient response, etc?

For example my component values for the woofer seem kind of suspect to me compared to the textbook values. Tiny inductors and a big cap?

In the FR graph violet is summed response, brown is woofer, blue is tweeter. In the Group delay graph red is woofer blue is tweeter.

Sorry if I don't make sense, it's 5 in the morning and I've been working for 12 hours.

[augerpro]

Here's group delay. The low frequency results are not valid.

[Marzen]

...how does one determine the quality of "real" XO? The only thing I could think of was plotting group delay, which actually looks pretty good around the XO frequency. I mean if the FR and phase look good what things can I look to determine Q, group delay, ringing, trasient response, etc?

Now that you've done your initial model you can build a test box & measure the actual Qb, impedance & a/c. Then import this data into SE and adjust for BSC & phase tracking. Then use the digital filter to give it a listen. If you have two pc's you can use one for the digital filter & the other to take measurements (easier said than done due to lag compensation).
-Ward

[cjd]

Well the textbook XO is totally useless.

Amazing isn't it? Yet there are people... ahh, nevermind. :lol:

C

[dlneubec]

Hi augerpro,

You need to add some dcr values to represent the resistance of those inductors. Using the default .00100 can throw things off quite a bit.

If you already new this, please ignore this comment! :B

[augerpro]

Thanks for the comments everyone. And sorry for posting messages when my brain is toast from twelve hours of work :roll:
Dan this was my first attempt at XO design, so I just guessed at DCR of the inductors. i looked at PE today and put some actual values in. Dropped the woofer level of course, but a little tweaking and the FR looks good again, just down about 1-2 dB or so.

Does anyone know how to calculate Q of the XO? It's not in V. Dickasons book. I thought I saw it in Speaker Building 201, but alas I left that at home so I can't check until I get off work tomarrow. I think it was Bullock's equations I'm thinking of...hmmm I'll do a google search.

While playing with SE you can change Component values by holding down a button and watch the predicted response change from small to large value changes. Several times I got severe peaking at the knee of rolloff. High Q. But I'd like a way to calculate this instead of looking at the FR and trying to see if the knee is peaking or not.

Anyhow, just looking for some tricks of the trade so to speak. Does anyone think the woofer section's L and C values look normal? They just seem so far from what the equations would suggest. Off to google...

[JonMarsh]

Equations that are based on resistive loads mean nothing when a real speaker has impedance varying significantly as a function of Frequency, due to it's fundamental resonance and voice coil inductance. So, a nominal 8 ohm speaker might peak at Fs in the 40-60 ohm area, drop steadily to an upper midbass value at 200 Hz of 6 ohm, then start climbing to a 20 kHz value of 25 ohms or more. You can make the design problem simpler on paper by using zobel networks to make the speaker load look resistive, but that eats up money in components that most of the folks here don't want to spend. Sometimes you don't have any choice if the fundamental resonance is in an critical region relative to the crossover, or if the driver design has relatively high VC inductance without any Faraday rings to minimize rise with frequency. If you want to use "textbook values", at least use the impedance at the chosen crossover frequency for the filter load.

Also remember Q is inherent in the type of filter and the location of poles or zeros- i.e., the difference between Butterworth, LR, Bessel, etc.

[augerpro]

If you want to use "textbook values", at least use the impedance at the chosen crossover frequency for the filter load.

Also remember Q is inherent in the type of filter and the location of poles or zeros- i.e., the difference between Butterworth, LR, Bessel, etc.

Thanks for weighing in Jon. Yep i used the Z at the XO frequency. I forgot about Q being inherent to the filter though ops: I've done so much reading the last 2 months I'm not sure if 1+1=2 anymore...

I guess what threw me was the peaking I got at the XO knee with some values. I'd imagine group delay cannot be good when that happens, or linear distortion for that matter. Is that a true statement?

I guess what I'm getting at is if the FR on and off axis is good, Z is acceptable, polar response and driver spacing/phase tracking is good, etc., what things should one look for in the component values that could present a problem that is not obvious to the designer from the various other measurements. Pitfalls or complications, that sort of thing.

If all the measurements/predictions look good, then there shouldn't be any major hangup corrrect? No troll lurking in my XO?

[JonMarsh]

Off axis versus on axis behavior of your drivers at the chosen crossover frequenc- typical bugaboo is running midwoofers too high, actually having some dips in off axis behavior which shows up in power response. Many "classic" crossover alignments have some destructive interference on axis, and lobing in a vertical axis, also resulting in weird power response.