Oh helpful people, please help.

[Tenson]

Hi,

I am building a loudspeaker using a Seas CA18RNX http://www.seas.co.uk/acatalog/H1215.pdf woofer and a Neo8 as tweeter http://www.partsexpress.com/pdf/264-713.pdf. I need some help with the notch filter and just a double check of my results please.

First of all, I have been playing with the DCX and know what I want the crossover to do. The settings I found worked well were a 12dB BUT filter @ 1.11KHz on the woofer. 12dB BUT filter @ 1.5KHz on the tweeter and a Notch filter.

Now if someone could check I have the right results for the crossover bits - I believe I need an impedance rise correction network with a 6.1Ohm resistor and a 29.6uF cap. For the LP on the woofer I need a 1.3mH inductor and a 17uF cap. This assumes a flat impedance of 6.1Ohms with the impedance correction. The tweeter then needs a 20.3uF cap and a 0.55mH inductor.

IS that right?!

Now the part I don't know how to do, the notch filter!

On the DEQ I am using settings of 9.87KHz, -5.5dB and Q 2.8. I have no idea how to translate that into a passive filter? What does their 'recommended' one do?

I have also taken a measurement of what the notch filter is doing to the signal. I have 'XoverSim' from FRD which says it can calculate the notch filter given the 'Fm (Hz), -3dB Fh (Hz) and -3dB Fl (Hz)’. Maybe someone knows what these values are?

Here is the measurement of the notch filter.



Thank-you oh so kindly!

Oh and how accurate do the passive components values have to be?

[nunayafb]

their 'recommended' one do?

by who? the pdf file on partsexpress is from B&G and it shows the recommended notch filter and component values. Is this what you are asking?

[Tenson]

I wondered what that notch filter did in terms of frequency and attenuation. I did mine by measurement and tail and error so although I could use their recommended filter, I don't trust that it will be quite what I want. But, someone might say hey, that filter B&G suggest does exactly what you are after so don't worry about working out the values for your one its done already!

[GM]

Greets!

Your zobel resistor selection implies a 4.88 ohms Re Vs the published 6.1 ohms that you used for the XO calcs, so it won't sum correctly. Note too that all these components will interact in complex ways, so the most accurate values will be different from these textbook ones. You will need a filter program to theoretically optimize the system and input actual measurements for best real world results.

That said, accuracy is in the 'eye of the beholder' and the final component selections are usually arrived at by 'voicing' (much tweaking over time), with the understanding that the XO is in our acute hearing BW while the notch filter isn't.

To calc the notch filter, given Fm = 9870 Hz and Q = 2.8:

Fl = ~8,263.484 Hz
Fh = ~11,788.842 Hz

GM

[Tenson]

Hmm XoverSim asked me for the Revc and Levc. I figure that’s Voice coil resistance and voice coil inductance? It returned those results.

Will the passive Xover be that far off? I thought with impedance rise correction and the Neo8 having flat impedance anyway the textbook xover should work very accurately?

I don't want to be buying lots of components to tweak and tweak it if I can. I took measurements and listened then did the Xover with the DCX so as long as the passive components do what they are supposed to it should be alright.

How does the 'Q' on the DCx work anyway? I am used to a Q of 1 being 1octave. 2.8 would be huge.

If XoverSim got the Zobel wrong has it got the other bits right?

Odd… XoverSim gives values of 0.076mH, 3.043uF and -14.840Ohms. How can it use a negative resister? Lol. How else can I calculate this?

[GM]

Greets!

Correct. Don't use it, but all the various formulas I have suggest beginning with a resistor that's 1.25x Re, so as always YMMV.

It's been my experience that it will, though you're right WRT the Neo. Again, it really all comes down to how it sounds to you. A digital EQ doesn't interact with the drivers the way passive components do.

Quality factors (Q) are defined differently for different apps. There's no one 'Q', unfortunately.

They are textbook rounded off 'close enough' IMO.

Hmm, I calc these, so I assume that either you somehow caused it to happen or the program has a problem that you need to discuss with Isaac.

C = 3.04 uF
L = 0.08 mH
R = 14.84 ohms

As for the zobel, given 6.1 ohms/1.1 mH, a common calc would be:

R = 7.625 ohms
C = 18.92 uF

GM

[Tenson]

Thanks very much!

Do you think it would be worth doing the Zobel then taking a measurement of the impedance at Xover point to check or will it be close enough? The impedance will be pretty much flat at 6.1Ohms, right?

Resistors and inductors can be pretty accurate but it is a lot harder (well, more expensive) to get caps the right value. For the Zobel would a 20uF be close enough?

Also, how can I measure the impedance at a certain frequency without building myself one of those jigs for speaker workshop?

Cheers

[lndm]

Do you have a signal generator and a DMM? Do you have an oscilloscope?

[GM]

Greets!

Unless otherwise stated, published Le is at 1 kHz, so in theory it's best to measure it at the XO point and use it to design the zobel. To do this:

Le(mH) = [(Zxo^2-Re^2)^0.5/(2*pi*Fxo)]*1000

Where:

Fxo = XO frequency point
Zxo = impedance at Fxo

Well, you could build/buy an impedance bridge or meter and there's other ways using resistors, sig. gens, etc., but many folks just read it as best they can off a published plot.

Being 'classically' trained (old), I was taught to reactance annull Re, which makes it flat theoretically to infinity, so this is what I've done when I've bothered to do it at all. Find the frequency (f) where Re has doubled, then:

r = 1.25*Re
c = 1,000,000/(2*pi*Re*f)

One more time though, it's what sounds good to you that matters.

GM

[lndm]

Originally posted by GM
Being 'classically' trained (old), I was taught to reactance annull Re,

Not having had this benefit, could you possibly explain the following: Why is R<>Re? and Why does C not complement L at a frequency where Z = Re x 1.414?

Thanks.