Spice modeling of speakers, other newbie questions

[parallax]

Hey guys. A few more questions about speaker design. This is my first speaker and I'm trying to do all the design from scratch, so here goes.

1) I don't have any software to do design with (I am writing some programs to aid me, though), so I'm using SPICE to help me with crossover design. A simple resistive load probably isn't enough to model the drivers. I'm using the P17WJ woofer and the D25AG tweeter from Vifa, and I'd like to have models of these for SPICE.

2) I don't believe I have access to anything fancy in terms of testing. If I do, it's probably an FFT analyzer somewhere around the university. /Maybe/ an anechoic chamber (I know we have one, don't know if I can use it). Can I still do testing? If it's just my computer generating tones and doing frequency/impedance response, is this still better than trusting manufacturer's data?

3) (Off topic) What sort of amplifier do y'all suggest for a system? I don't have one currently and I need to buy one. Speakers I'm building will cost around $300.

Is there a way I can model the drivers using T/S params, or what?

And finally, I'd like to know if I'm missing anything important in the crossover design process. I've picked a crossover frequency of 2370 Hz (chosen by a program I wrote to be in the 2000-2500 range and have component values closest to what's available). I'm going to use a parallel 2nd order L-R network and work on phase alignment. I'm L-padding the tweeter to lower the sensitivity at least 1 dB. I'm using a Zobel network on both the tweeter and woofer to eliminate distortion on highs. I'm using a simple LR circuit in series with my entire crossover to eliminate baffle step, as in the Murphy article. What am I missing?

Thanks.

- R

[Mr Evil]

1) Here's an article on modelling of loudspeaker impedance: http://www.epanorama.net/documents/a...impedance.html
and another: http://sound.westhost.com/tsp.htm

2) An anechoic chamber is ideal, but you should be able to get reasonably accurate results in a normal room (except frequency response measurements) even with ghetto measuring equipment.

As indicated in the two linked articles, you can indeed generate a model from T-S parameters (more accurate). You could also just measure the impedance directly at various frequencies (easier).

You may also want to compensate for the peak in impedance at the tweeter's resonance, if it's close to the crossover frequency.

Don't forget to burn-in the drivers before measuring, as their characteristics can take some time to settle down from new.

[parallax]

Thanks a lot. :-) That was very helpful.

Is there anything that comes close to an anechoic chamber? (like an open field?)

[Mr Evil]

Depending on the landscape, a field could potentially be better than an anechoic chamber due to the larger dimensions, but ambient noise wil surely be a problem.

[f4ier]

1) One would need to enter both measured SPL/SPL phase and Imp./Imp. phase data into SPICE to make meaningful simulations. Models are good for academics, also remember that real-world voice-coils are lossy inductors. Measured data means everything in the real world. Have at look at PSPICE's FTABLE.


2) As was already said, an anechoic chamber would be great. Just remember to take note of the chamber's low frequency cutoff frequency.


Yep, models can be had from T/S parameters, but they're really only useful for low-frequency modeling. And even if we had accurate electrical models (which would include the lossy voice-coil), there's still the issue of obtaining acoustical models. T/S parameters can't give us acoustical information at frequencies where the driver unit's cone doesn't behave as one piston [cone breakup]. We're interested in the acoustical transfer function -- not the electrical transfer function. This thread may help clarify.


BTW, have you tried Speaker Workshop and various other tools from the FRD Group?


Cheers

[foo]

how about using BassBox pro and Xover Pro ?

[f4ier]

how about using BassBox pro and Xover Pro ?

Forget about both, which are made of nothing more than equations taken from some book(s).

Unibox and/or Subwoofer Simulator have more advanced features (for free!).

XoverPro can't even use measured spl/imp. data -- a serious accuracy limitation. Moreover, XoverPro does not let you design filters other than of the textbook parallel 4th-order type. Speaker Workshop / Crossover Simulator can take driver acoustical/electrical measurements and let the user design any circuit size/complexity... And both are free!

[parallax]

Thanks for the advice, f4ier.

My PC is a desktop, unfortunately, and I have a mac laptop that would be ideal to use once I find a testing location.

I have a few days to kill and nothing to do before my drivers come in, so I'd like to code an SPL response measurement program for the mac (AFAIK there isn't a free one out there). Do programs generally output pink noise and then do a high-sample FFT to pick out the response, or sweep a sine wave across a certain number of frequency samples and measure power, or? Does a built-in sound card generally have enough quality for output, or do I need to wire up a noise generator circuit?

And what sort of (cheap!) microphone is best to use for this?

So I guess I should build a box and do the testing in there since that'll give me the best measurements to work with.

And here's a stupid question: Do I need to include any series resistance with my speaker, or is a crossover plus the driver impedance as it is enough not to blow an amp?

Sorry for all the questions, lol. So much to learn!

Thanks a bunch.

[Zaph]

Originally posted by parallax
I'm using the P17WJ woofer and the D25AG tweeter from Vifa, and I'd like to have models of these for SPICE.

I just did a design using these drivers. I crossed over at 2330, pretty close to what you planned for. You'll probably still want to mess around, model and experiment for learning purposes but my design might make a good reference for you. I have exported data available if you want to use it, and I may also have driver equivalent circuits, which would give spice some realistic output rather than just using straight impedences.

Be aware that 2nd order acoustic rolloffs will be impossible to do with these drivers. However, 2nd order electrical components can result in good 4th order LR rolloffs. You'll need much more than 1 db of padding on the tweeter if you use baffle step compensation on the woofer. (and you should)


Vifa Classic Combo - P17WJ-00-08 and D25AG-35-06

[parallax]

"Be aware that 2nd order acoustic rolloffs will be impossible to do with these drivers. However, 2nd order electrical components can result in good 4th order LR rolloffs."

What does this mean? How do the order of electrical components and acoustic order interrelate?

Thanks,

- R