Yes, AGGEMAM, you are perfectly right about the similarity. The nasty dips above 5khz is likely to be BSD, which is why I consider those impossible to get rid of. I'm hopefully wrong here....
The dip at 4khz is also very likely to be BSD. This dip is the worst of them all. What can I do about it?
The dip at 4khz is also very likely to be BSD. This dip is the worst of them all. What can I do about it?
You not gonna like this but tinkering around with the speaker placement on the baffle is one way. Try using Svante's program for that first, even small changes have an effect.
Second, most of the time this tweeter is recommened to be filtered at 3,8 KHz (see this link ) so it's actually not that big of a problem the 4KHz peak because with that filtering the effective x-over frequency is roughly in the 3KHz region (because the peak cancels out the filtering).
Check how that sounds before trying to filter the 9KHz peak with a notch filter.
Concrete recommendations:
12 db/octave filter at 3.3 KHz and 6 ohms using a 5,6 uF cap and a 0,4 mH coil giving and effective x-over at 2,7 KHz.
Second, most of the time this tweeter is recommened to be filtered at 3,8 KHz (see this link ) so it's actually not that big of a problem the 4KHz peak because with that filtering the effective x-over frequency is roughly in the 3KHz region (because the peak cancels out the filtering).
Check how that sounds before trying to filter the 9KHz peak with a notch filter.
Concrete recommendations:
12 db/octave filter at 3.3 KHz and 6 ohms using a 5,6 uF cap and a 0,4 mH coil giving and effective x-over at 2,7 KHz.
sobazz said:
These curves definitely tells me that the 4kHz dip is a diffraction effect. It is more or less gone in the close field measurement, where you are (relatively) far from the edge (compared to the distance to the tweeter. You could try to measure *really* close, like 5 cm or so to confirm this. I see two ways of dealing with this, either you move the tweeter a little on the baffle, or you do "something" in the crossover network. Depending on the crossover frequency this may be hard or (well...) easy. A simulation tool for the crossover network will be needed, and also for the baffle diffraction, but that you've already got
In the simulation for the whole baffle (I posted that before) there is a dip at 4-5 kHz, this makes me happy since that means my simulation is at least rightish
This is good fun!
Svante and AGGEMAM, I'll look into your suggestions. Thank you. Moving the tweeter on the baffle is out of the question, but I'm very happy for the diffraction simulation program.... excellent tool! It should make it possible to avoid such proglems in a future project.
AGGEMAM: I looked at the australian DST website, and found this picture (here). The small speakers showed on the foto actually looks very much what I'm trying to make - although the tweeter is not mouted off centre. As far as I can see they also use a TC11 or PL11 bass/midrange. On the response curve - which is clearly done with the tweeter baffle mounted - the same dip occurs.
AGGEMAM: I looked at the australian DST website, and found this picture (here). The small speakers showed on the foto actually looks very much what I'm trying to make - although the tweeter is not mouted off centre. As far as I can see they also use a TC11 or PL11 bass/midrange. On the response curve - which is clearly done with the tweeter baffle mounted - the same dip occurs.
sobazz said:
Could be that the tweeter is mounted on the same distance from the edge. If you look in my diffract you can get a visual impression of which part of the edge that is most important by counting the "density" of the edge sources. They are all equally strong, and concentrated to places where the driver is near the edge. Having the tweeter on the same distance would produce many sources on the same distance as you have, and a similar response.
Just for the fun of it, you might want to test a few pieces of cardboard and see if you can move or get rid of the dip by altering the baffle shape. If I were you, I could not resist trying it even if it would not be the permanent solution. But then again, I'm not you...
Svante, I did the cardboard test at 60cm. Here's the result:
Slightly different from the others. The 4khz dip is almost gone. But the dips at higher frequencies are still present at an almost unaltered amplitude.
An externally hosted image should be here but it was not working when we last tested it.
Slightly different from the others. The 4khz dip is almost gone. But the dips at higher frequencies are still present at an almost unaltered amplitude.
Great! Isn't it nice when understanding is reached? I guess you are left with the option to compensated for the dip in the crossover filter. The oscillation at higher frequencies is a bit harder to understand, but they seem periodic with a period of ~2kHz. This could be an interference from some kind of reflection with a pathway difference 345/2000=17cm. I dare not guess what that would be.
I think you have showed me the light. I begin to understand.... what a great feeling. I suppose it's about time to start a new project in order to take BSD into account when making the box. This thread has taught me a lot.
The internal depth of the box is 17cm. But more likely, the distance from the centre of the tweeter to the bottom of the baffle is also 17cm. I suppose this is the explanation.
By the way - how can you compensate for the dip at 4KHz in the crossover? I already have to do a line level equalization in order to improve the bass and lower midrange.
The internal depth of the box is 17cm. But more likely, the distance from the centre of the tweeter to the bottom of the baffle is also 17cm. I suppose this is the explanation.
By the way - how can you compensate for the dip at 4KHz in the crossover? I already have to do a line level equalization in order to improve the bass and lower midrange.
I'd go for the bottom edge of the baffle. The tweeter does not generate any sound inside the box, so that will not be it. It *could* also be a secondary edge reflection from the outside back of the box, which should be ~20 cm away from the front? OTOH that would be a 40 cm pathway difference (back and forth), so probably not.sobazz said:
Try the cardboard trick at the bottom edge and see if the oscillation changes. Remember that you will have to change this more than the upper edge since 5 cm is less than 17 cm.
sobazz said:
There are many ways, but one would be a series resonator in parallel with a resistor. Also if you do 2nd order crossover at 4kHz you could let it peak a bit.
But if you already are into active compensation , that would be better IMO.
AGGEMAM, I've decided to do it the proper way in order to save money on filter components. I'm currently measuring driver impedances which will enable me to simulate the crossover reasonably.
5th element: What is your experience with the TC11 electrical parameters? I expected a tuning point of around 73hz for my little 3-5->4L box but the impedance measurement shows that the Fb is around 60hz.
I suspect there may be an error in the measurement. I did some test measurements on passive components. Resistance and capacitance was fine, but inductance was a bit of a surprise. I only have 2 inductors (air core - of the same size) at the moment. Probably made by Dansk Audio Teknik. They are marked 0.5mH - written in hand directly on the inductor - but I measured 0.66mH. I know you shouldn't trust the measurements completely - especially larger values - but this difference is rathar large. It could explain some of the troubles I've had with a previous crossover attempt, though ....
5th element: What is your experience with the TC11 electrical parameters? I expected a tuning point of around 73hz for my little 3-5->4L box but the impedance measurement shows that the Fb is around 60hz.
I suspect there may be an error in the measurement. I did some test measurements on passive components. Resistance and capacitance was fine, but inductance was a bit of a surprise. I only have 2 inductors (air core - of the same size) at the moment. Probably made by Dansk Audio Teknik. They are marked 0.5mH - written in hand directly on the inductor - but I measured 0.66mH. I know you shouldn't trust the measurements completely - especially larger values - but this difference is rathar large. It could explain some of the troubles I've had with a previous crossover attempt, though ....
sobazz said:
That's okay. I think you'll end up at around my suggestion anyways if you really are interested in saving money.
Oh, and regards to the coil problem. If some sort of iron or especially a magnet is within reasonable range (5-20 cm depending on size of magnet) the coil will act differently. Notice that the problem will occur once the filter is in place in the speaker.
sobazz said:
You could check the inductance by connecting a known capacitor in series and find the minimum impedance. If you can measure the driver impedance you should also be able to measure the impedance of this circuit. The inductance is L=1/((2*pi*f0)^2 * C)
where F0 is the resonance frequency.
Svante, I'm not quite sure I understand what you mean. I could do an impedanse vs. frequency vs. phase respone measurement with a known capacitance value in series with the inductor?
Anyway - I tried it.... I probably should have included a resistor...
I have yet to find out if my Audigy has been damaged, but it is very likely. The line out say a load of less than 0.2ohm at 2.74KHz. The response graph shows dips as low as 50mOhm. But what is the inductance?
Anyway - I tried it.... I probably should have included a resistor...
I have yet to find out if my Audigy has been damaged, but it is very likely. The line out say a load of less than 0.2ohm at 2.74KHz. The response graph shows dips as low as 50mOhm. But what is the inductance?
Is the audigy OK? I hope I didn't mess it up for you. I did not know how you intended to measure the impedance of the loudspeaker, so maybe I should have said something about a series resistor. Anyway, consumer products are usually pretty robust, otherwise they would get loads of returns. A guess would be that the Audigy made it, unless you have noticed otherwise. It should handle a shorted output.
sobazz said:
Well... The inductance of a voice coil is far from the ideal inductor. There are several ways to model this. One empirical approach is to add an extra inductor//resistor series with Le, this makes a better fit to real data. There is also another model. I don't know how much you know about complex math, but let's try it:
The ideal coil has the impedance Z= jwL, where j is the imaginary number sqrt(-1). J is used to indicate that current lags voltage by 90 degrees.
The alternative model has the impedance
Z= (cos(n*pi/2)+jsin(n*pi/2)) * w^n *L
n is typically 0.6..0.8 for most drivers, n=1 for the ideal coil.
The first part fixes the phase, 90 degrees for n=0 45 degrees if n=0.5 etc. In the second part we see that the absolute value of the impedance goes as w^n rather than w (as it was for the ideal coil).
This model fits most drivers very well. I don't know where manufacturers get their Le values from, but they can only be valid for a single frequency, given the lossy nature of the real coil.
Bottom line here is that the Le value specified in the data sheets is mostly pretty useless.
I wonder if I answered your question. Hmm... If you can rely on your impedance measurements? Well if the curve matches Peerless datasheets, that would be a good indicator that you impedance measurement method is OK. OTOH I have not seen the equipment, so it is hard to tell.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.