Passive audio filters

[vainajala]

Hi,

I have in mind building small 2-way loudspeakers to my stereo system, because the old speakers are in so bad condition. The bass-element's foam surround has come really "weak" and it has many holes even though the surround hasn't been touched by hand practically at all! The loudspeakers are at least 15 years old, so it's really not a wonder

I'm considering Vifa TC14WG49-08 as bass/midrange:
http://www.d-s-t.com/vifa/data/tc14wg49-08a.htm
and (maybe) Vifa D20TD-05-06 as tweeter:
http://www.d-s-t.com/vifa/data/d20td-05-06a.htm

They are most propably to be installed in 14 liter bass-reflex box tuned at 43 Hz.

I knew something about different loudspeaker enclosure types (and how they work) before I was even thinking of starting this project.
But I must admit that i'm quite a new-bie when it comes to crossovers. I haven't designed these before and I thought it might be wise ask you will they work before I start buying the crossover parts. They are quite expensive, you see...

The pictures from analysis program 5-Spice follows. They are (at least they should be ;D) -12dB/octave filters, with Zobel circuits. Any comments are appreciated...

View attachment 35556

 

[audioguru2]

Vainajala,
I have some concern about your choice of a passive 12dB/octave Butterworth crossover filter and low-value Zobel networks:

1) If the woofer and tweeter are wired with the same polarity, then the system acts like a notch filter at the crossover frequency. If one driver has its wires reversed, then the system will have a 3dB peak at the crossover frequency. See my discussion of this matter and a link to a manufacturer's expert's comments, here:
http://www.electronics-lab.com/forum/index.php?board=2;action=display;threadid=636

2) The 12dB/octave filter for the woofer may not have enough attenuation at its 4dB response peak of 5800Hz. This peak gives a "nasal" sound and it is best to attenuate it with a steeper filter.
The same problem may occur with the 1700Hz resonance of your tweeter, since it may reach the limit of its excursion and distort badly.

3) Although zobel networks are good to flatten the impedance of inductive speakers, your low-impedance speakers in parallel with your low-value zobels results in very low impedances. Can your amplifier drive a 2 ohm load?

4) A passive crossover ruins the damping of each driver by the amplifier.

I, and the expert linked in my discussion that is linked above, agree that a passive 18dB/octave Butterworth crossover network will solve most problems, and if you use the woofer's inductance as a component, then you simply must add only 1 cheap capacitor to the tweeter's filter, to obtain 18dB/octave Butterworth filters for both the woofer and tweeter.
Perhaps you can increase the resistances (and correspondingly reduce the values of the capacitors) of the zobel networks, to increase the impedance of the completed speaker as seen by the amplifier.
I have not checked that your filters' crossover frequencies match each other, nor that they match the parallel combination of each driver and zobel, nor that they match your drivers.

Another expert also agrees to avoid a passive 12dB/octave crossover, and strongly recommends bi-amped speakers with active electronic crossover networks. I have heard many professional systems of this type and the sound is truely amazing!
If you are capable and ambitious, then this is the way to go!
This expert's articles are here:
http://sound.westhost.com/lr-passive.htm

What do you think about that?

 

[vainajala]

Hello Audioguru, and thanks for your reply

1) If the woofer and tweeter are wired with the same polarity, then the system acts like a notch filter at the crossover frequency. If one driver has its wires reversed, then the system will have a 3dB peak at the crossover frequency. See my discussion of this matter and a link to a manufacturer's expert's comments, here:
http://www.electronics-lab.com/forum/index.php?board=2;action=display;threadid=636

Actually, I was going to connect the drivers with opposite polarity.

2) The 12dB/octave filter for the woofer may not have enough attenuation at its 4dB response peak of 5800Hz. This peak gives a "nasal" sound and it is best to attenuate it with a steeper filter.
The same problem may occur with the 1700Hz resonance of your tweeter, since it may reach the limit of its excursion and distort badly.

This really got me thinking of using 18dB filters, for their better performance.

3) Although zobel networks are good to flatten the impedance of inductive speakers, your low-impedance speakers in parallel with your low-value zobels results in very low impedances. Can your amplifier drive a 2 ohm load?

My amp is a Sony's model STR-DE585, not a welding machine ;D But seriously, I think it's rated for nominal impedances between 6-16 ohms...
If I use the filter component values determined by the basic calculations (with no Zobels), and simulate that with 7spice, the freq. responese has very large variations, caused by the inductance at the driver... I see no other way than using the Zobels

4) A passive crossover ruins the damping of each driver by the amplifier.

Of course this is a bad thing, especially when one wants only the highest quality and performance. But you see, my goal with this design is to create small 2-way loudspeaker pair with decent soundquality and complexity... So I think a natural choice for me is a passive crossover.

I, and the expert linked in my discussion that is linked above, agree that a passive 18dB/octave Butterworth crossover network will solve most problems, and if you use the woofer's inductance as a component, then you simply must add only 1 cheap capacitor to the tweeter's filter, to obtain 18dB/octave Butterworth filters for both the woofer and tweeter.

That's a great suggestion, but when using woofer's coil inductance as a 0,60mH inductor, I get crossover frequencies about 930 Hz for 18dB crossover... with formula taken from speaker design book.

L=(80*Z)/f --> L=(80*~7ohm)/930Hz = 0,6mH

IMHO, this could work well with 3-way's, but bit too low for 2-way

Perhaps you can increase the resistances (and correspondingly reduce the values of the capacitors) of the zobel networks, to increase the impedance of the completed speaker as seen by the amplifier.

I'll see what I can do!

I have not checked that your filters' crossover frequencies match each other, nor that they match the parallel combination of each driver and zobel, nor that they match your drivers.

These both filters have -6dB points at 3,05 kHz according to 5spice... should there be some "gap" between them, for example, -8dB at 3,05 kHz?

Another expert also agrees to avoid a passive 12dB/octave crossover, and strongly recommends bi-amped speakers with active electronic crossover networks. I have heard many professional systems of this type and the sound is truely amazing!
If you are capable and ambitious, then this is the way to go!
This expert's articles are here:
http://sound.westhost.com/lr-passive.htm
What do you think about that?

I think it's an interesting article...
But for the bi-amping, I understand it is a great technique because there is no need for passive crossovers, but unfortunately I don't have resources and equipment for that kind of system...

 

[audioguru2]

Hi Vaianajala,
That's good that your LPF and HPF have a gap between them so that it will flatten the 3dB peak that is caused when 12dB/octave Butterworth filters have the drivers connected in reverse polarity.
It is too bad that the little woofer has such a high inductance so it can't be part of a simple 18dB/octave filter.
Another concern is that the tweeter is much louder than the woofer. If you add an attenuation resistor in series with the parallel tweeter/zobel, then its crossover parts will need re-calculation for the increased impedance, and it will be easier to drive by your amp.

Now all that you will need is a windows-rattling sub-woofer!

 

[vainajala]

<bump>
I understand this is not really a forum about loudspeakers, but here's some updates...

This might be the final crossover's circuit... I just need to wait until the drivers arrive to me (about 2-3 weeks maybe).

And when I get my "testing gear" (a speaker measurement jig+microphone+computer) ready, I can actually measure how well or bad these drivers perform in this application. And tweak the crossover if needed.
Before that, I must trust my ears ;D

The Jig: http://www.gti.net/wallin/audio/audua/audua.html

View attachment 35608

 

[vainajala]

Response graph....

The extra parts L5 and R6 in schematic are supposed to perform baffle-step compensation, to attenuate the midbass-midrange quite a bit.

View attachment 35609

 

[audioguru2]

Hi Vainajala,
That should be better, but I hope that your step-compensation doesn't mess-up the phase at crossover too much.

How did you manage to have a tweeter rolloff that is much more than 12dB/octave? Did your sim program screw-up?

 

[vainajala]

With some tinkering with 5spice, i added the phase information into the graph. How does it look like...?

How did you manage to have a tweeter rolloff that is much more than 12dB/octave? Did your sim program screw-up?

I just experimented with different ccomponent values. That's all. View attachment 35614

 

[vainajala]

<bump #2>

Here's the frequency response (20Hz - 16KHz) of my old speakers, I just measured it with Speaker Workshop...

The peak at 50Hz is propably mains voltage hum. I think below 35Hz is mostly junk too.

They have 8" woofer and 1" dome tweeter, and have no cross-over. Just bipolar cap on tweeter.

View attachment 35666

 

[audioguru2]

Hi Vainajala,
Your old speakers' response is pretty bad, hopefully only due to room reflections. That is quite a big hole at 1.8KHz. I've seen very good speakers measure even worse in a room.
I take my speaker to the backyard, point it straight up, and hang the microphone above it on a fishing pole. No reflections unless the moon is full. ;D

 

[vainajala]

I don't believe it's because of room reflection, tried in very different positions and curve looks almost perfectly same. My old speakers just simply suck.
Your speaker-measuring arrangement seems cool. Are you using a laptop-computer for measuring? What program you prefer? I use Speaker workshop, because it's free!

I expected that kind of response curve, because high freq. sounds (like cymbals) sound very loud and attacking, and in the lower octaves is almost no energy.
It's quite hard to hear even a normal bass-guitar with these speakers. Placing in room has no/very little effect. >

The Vifa drivers are propably delivered in a couple of days.
In next weekend when I get back home from school, I'll build and measure a prototype-speaker... So we'll see how well/bad they perform :

If they perform reasonably well, that means more work to me: because my brother needs also a pair of compact speakers... ;D

 

[audioguru2]

Hi Vainajala,
I use long cables with my desktop computer, and a speaker manufacturer's program that I copied.
Speaking about speaker manufacturers: I used a portable octave-band LED real-time-analyser (with built-in pink noise generator) to show the local high-end speaker manufacturer that the coils that I bought from them and that they were using in their finished speakers, had the wrong inductance! Their response was much worse than your old speakers and sounded terrible. They probably sold thousands to people with tin ears.

I didn't actually "build" my speakers. I used discarded, cheap but solid and pretty, PA speaker boxes. Then I simply selected drivers that work well in them, and designed the crossover, ports and EQ.

 

[vainajala]

audioguru:
If a speaker manufacturer defines itself High-End grade, and then puts a wrong value inductor to their production speakers, does not seem very "High-End" to me. In fact, it's a cardinal mistake. I wonder if the customers (and you) got their money back, because of a "bad" product?

Do you think their speakers would still be in sale, if you as a Guru of Audio hadn't been concerned about the parts' quality?

From now on I will measure every coil and cap before putting it in work anywhere, especially in speakers!


My 'dream-speakers', I wonder if they got their inductors correctly:
http://www.wilsonaudio.com/products/Alex/index.html

 

[audioguru2]

Hi Vainajala,
That high-end speaker manufacturer who sold me inductors that were marked with the wrong value and used them in their own speakers, make very expensive speakers with graceful curves and a magnificent finish. They made 4 mistakes:
1) One value of inductors that they make was marked with the wrong value. (But their crossover design is excellent)
2) Their quality control for parts manufacturing did not discover the mistake (no testing?).
3) Their quality control for finished speakers also did not discover the bad sound. See below.
4) They sold defective finished speakers.

Apparently, they were so concerned about their magnificent finish that they did a poor quality control listening test (no measurements?).
I can visualise their quality control group for finished speakers:
A bunch of teenagers looking at the finish with a magnifier and listening for squeaks and rattles by listening to "boom boom, chhh chhh, yappity yap" rap.
They are still in business and are doing well.

I'm sorry, but I think that your "dream-speakers" are ugly and weigh more than many cars. But they might sound OK.

 

[vainajala]

Thread resurrection... 8) 8)

Lately I've been busy trying different cross-over types, and found out that 2nd order Bessel filters, crossing at 2 KHz, seems to work quite well. I built a prototype-speaker and measured it, here is the resulted far-field frequency response.

Only bad thing is that the highest frequencies have 6db rise between 10-20 KHz. But when listening (and measuring) the speaker at approximately 15-20 degrees off-axis, response flattens nicely. This measurement was done off-axis.

View attachment 35800

 

[audioguru2]

Hi Vainajala,
What are those curves?
The top one couldn't be for the speaker's phase.
The bottom one couldn't be for frequency response, but maybe for group delay.

 

[vainajala]

Hello audioguru,

The graph is frequency response of the prototype speaker, measures at approximately 1 meter away. Do you think it's totally messed up? How should 2-way speaker's phase response look like?

Your comment made me suspicious to my measurements, so i'll have to re-measure to see if all went right.

Seems like I'll have to increase the levels quite a bit. With as little level as -15dB (as in my graph), even the background noise might affect the results.

And, the bass-area peaks and valleys might be caused by room standing waves.

 

[audioguru2]

Hi Vainajala,
It is no wonder that your curves look so messed-up. You were measuring the room's reflections, not the speaker's response. If you move the speaker or microphone only 10cm, the results will change.
That's why I measure my speakers outdoors, pointing straight up, where there are no reflections.
BTW, what microphone do you use? I use an Audio-Technica electret goose-neck one that is a good price and has good response.