Mixing small 'C' and big 'L' in 2nd order x-over

[SimontY]

Hello one and all,

Firstly, apologies if this has been discussed before, but I had no clue what to search for!

Currently, I'm happily listening to a mixed order x-over - 1st order bass, 2nd order treble, x'ed at about 5khz (a guess). This is where the bass driver starts to roll off and it seems to work well, summing to an audibly flattish response.

Because I've scrapped my stock x-overs, and am experimenting to see what floats my boat best, I've been trying to use the existing parts as much as possible. This has resulted in having a whopping great 0.68mH air-cored inductor in parallel with the tweeter, after a relatively small 1.5uF cap in series. 1st order was too harsh, the poor tweeter couldn't really hack it I guess (also, ~2.5khz [stock] 2nd order was harsh and disorientating, and too forward, if detailed)

I have a small inductor in series with the bass, then zobel and impedance correction.

So... there's my setup, but I'd like to know what, if any, implications there are in mismatching the x-over component values like I've done ie. big L and small C. It seems to work, but would it work even better doing in conventionally?

TIA for any input!

[richie00boy]

Big L and small C is kinda irrelevant; who's to say what is big and what is small?

With any filter higher than 1st-order, the roll-off slope although eventually being 12dB/octave for 2nd-order, 18dB/oct for 3rd and 24dB/oct for 4th, etc. will have a different initial roll-off according to the Q of the filter. Low Q gives more gradual roll-off, high Q can peak up above the nominal level. The Q also affects the transient response of the signal around resonance, with a higher Q being more underdamped.

So, in conclusion, you need to sit down and work out what Q and -3dB frequency you are getting with your combination of capacitor and inductor, taking into account the actual impedance (i.e. not just the generic 8 ohms or whatever) of the tweeter and any level adjustment resistors. If I can be brutally honest, just playing around with values without this kind of knowledge is like flying in the dark with no artificial horizon.

I'm trying to suggest a book for you, but often they are quite heavy on the maths. I don't like maths and used a book in my early years called Practical Electronic Filters in the little Babani range, but I think it may have concentrated on active filters. An easy way out to do it would be to get hold of a speaker design program that does crossovers, or a SPICE simulator with GUI front end.

[SimontY]

Thanks, that's useful information.

I expected to be told I am doing something wrong, but that's fine. I can barely handle basic maths, let alone a complex measurement rig or badly designed software, but my ears are fine, so using some basic logic, then experimenting wildy and blindly for hours on end seems like a reasonable choice, for now. I have had three combinations that were and are eminently more listenable than the one supplied with the kit - so flying in the dark doesn't exactly guarantee disaster does it?

1st order tweeter at ~5-6khz and FR bass sounded great, as did 2nd order tweeter, crossed ~5khz and FR bass. Many things tried sounded weird and wrong, and yes, without measurement or maths, I am just guessing, but I have good sound, believe it or not. I've tried zobel on both drivers, various resistor combos, x-over points, slopes, cap bypassing combos and whatnot. Clearly, this is a crappy way to work, but it's also quite fun, and now the speaker is designed by me!!

I had not considered, nor do I know much about, the Q of each x-over part. I will look this up, as it could really help me understand why it all sounds the way it does. This may explain the slight lack of definition in the x-over region, who knows...

I have a speaker design book, not a very good one though - Advanced Speaker Designs it's called. I want the Loudspeaker Design Cookbook, I'm sure that will help me figure out the remaining mysteries.

So, one thing you suggest then, is that my inductor part of the x-over may have one 'Q' and the capacitor part another, different 'Q', which could cause strange things to happen presumably.

I used WINISD Beta to figure out potential x-over values, but it cannot simulate a non-standard L and C combination, so I need better software. I will look. Can you suggest software that can be given L & C values and will work out the x-over fr. from that?

Thanks for listening!

[sreten]

You"ll get a highly damped second order characteristic.

C/o region will basically depend on the capacitor, at these
frequencies driver impedance will dominate the inductor.

Consequently inductor value is not critical at all,
it will reduce excursion below crossover and
damp an impedance peak if one exists.

Adding a high value inductor to basically a 1st order treble
c/o is a perfectly sound principle for resonance supression
and increased power handling if crossing over well above
Fs. It can also be used for subtle phase manipulation.

A standard circuit simulator can be used to analyse the c/o.

The Tina demo is very easy to use but very limited in
the number of components you can use and you can't
save results, but its quick for looking at simple networks.

http://www.tina.com/demos/60/TinaPro60En.exe


sreten.

[richie00boy]

I used to be like you But I got fed up of faffing around for hours, wasting money on parts and making little progress (OK seems like progress, but until you grasp the theory you realise how little progress it actually was), and I wanted to know WHY things were doing what they did. I hate and am rubbish at maths too, but you really should try and make yourself learn at least a little bit, you are denying yourself masses of benefit if you choose to ignore it.

I was going to suggest Vance Dickason's Loudspeaker Design Cookbook, but it's a very tabular/chart driven approach to design. Like you I wanted to experiment with non-standard stuff, so maybe this book isn't so good for you.

So, one thing you suggest then, is that my inductor part of the x-over may have one 'Q' and the capacitor part another, different 'Q', which could cause strange things to happen presumably.

No! Neither individual component has a Q, the combination of the L and C causes a response that has a Q.

[SimontY]

Hi Streten,

VERY interesting info. you've given. It explains my observations well - such as putting my finger on the tweeter with and without the inductor made a huge difference in vibration/excursion and resulting presence zone unpleasantness, more so I thought than doubling the rolloff rate would do.

Also, it explains why when I used it first order and 2nd order I was using a similar value cap. I am crossing way over FS btw - 5khz to 900hz.

I will look for Tina, thanks!

The search for the near-perfect x-over for my speakers seems to have a long way to go...

[SimontY]

Originally posted by richie00boy
No! Neither individual component has a Q, the combination of the L and C causes a response that has a Q.

Heh, ok

I hear what you are saying, and it's very encouraging. I'd like to say that not long ago I had no clue how x-overs worked, and then gradually I'm understanding them more and more (not series ones yet!) - things like zobels and phase and all that, are starting to make some sense.

I am about to buy an LCR meter, and when I have this I can then really experiment hard! -I'll know what components I'm using at last, then I can simulate what I've done and try to relate it to the sound. Without hands on stuff I get frustrated very quickly...

I appreciate the advice, I realise I must sound a bit daft!

[sreten]

Originally posted by SimontY
Hi Streten,

VERY interesting info. you've given. It explains my observations well - such as putting my finger on the tweeter with and without the inductor made a huge difference in vibration/excursion and resulting presence zone unpleasantness, more so I thought than doubling the rolloff rate would do.

Just note 12dB/ octave will give constant excursion below
c/o down to Fs of the tweeter.

6/dB octave excursion increases 6dB/octave down to Fs.

Below Fs the drivers 12dB/ octave roll-off makes excursion
constant, so with any crossover excursion reduces below Fs.

Hence as I said adding adding an inductor is extremely
useful for any 1st order crossover well above Fs.

It will have a large impact on distortion, especially IM.

And explains why high power tweeters need 18dB/oct c/o.

And why cheap tweeters for use with 6dB/oct have high Fs.

sreten.

[richie00boy]

Tina is very good. I was going to suggest it myself, but thought you would not want to get that deep. Congratulations on taking probably one of your largest steps forward so far

If you really want to get into speakers, save up for a signal generator and very good battery-operated multimeter that can measure true RMS voltage (with wide bandwidth) and frequency, e.g. decent Fluke. You can use a 'scope, but you may need to mess with lifting grounds and looking at <30Hz sinewaves on a non-storage scope (i.e. anything I can afford) is not much fun.

[SimontY]

Interesting points, it explains clearly why you don't often see 1st order tweeter sections...

A couple of relevant questions, if I may:

Do you think the 2-3khz 2nd order could cause too much excursion for a 1" tweeter causing harshness, as in the original design?

Also, you say the inductor value is not critical at all. But, will it still be causing a 90 degree phase shift, or is that text book stuff too simple in real life? I ask, as I fear I may have lost the phase 'coherance' from one driver to the next in this configuration.