[tvrgeek]

What are all the trade-offs of

#1 putting the tweeter pad at the tweeter or

#2 at the input to the crossover.

#3 I have even seen designs with both

#4 or use of an L-pad ( at either location)

Obvious, if at the tweeter, crossover components are smaller and cheaper, but it is a lot harder to vary the level (redo the crossover)

If at the input, easy to change levels, amp may like it.

I am really looking are sonic issues. Is the difference in damping relevant? Make a zobel not necessary, effect on ringing if any. That kind of stuff. Final design decision.

 

[djarchow]

An L pad assuming you stick to the proper values for the resistors is the best way to avoid other XO changes if you decide to lower the level. For example, a single resistor just before the tweeter works fine for attenuating the level of the tweeter, however if you want to change the level, then the impedance seen by your XO also changes impacting the XO point.


You can do some shaping of the tweeter rersponse by placing the resistor before the rest of the XO or right before the tweeter including putting it in both places.


Regards,


Dennis

 

[tvrgeek]

Not interested in the "easy" adjustments, looking for end points.

Placing in series with the tweeter would make Le less of a factor by increasing Re relative to Le. Maybe saving cost in not needing a zobel. I would also think it effects the Q of any resonance point. In all, making a more predictable load to the network. Assuming one adjusts for equal level and equal f3, what are the sonic effects?


Placing the pad at the input is effectively increasing the source impedance. Again, network values can be recalculated.


An L pad at the tweeter end would seem to lower the impedance as seen by the tweeter. This would be a good thing, but with realistic values, I am not sure it is relevant.


Does anyone know of a good text that discusses passive crossovers in this level of detail?

 

[djarchow]

I believe in the later editions of the loudspeaker design cookbook, Dickason shows the impacts of all three placements on frequency response. There is not much detail beyond some graphs though.


In an L-Pad the parallel resistor is the one that tends to swap out the inductive rise of the driver.


As for the placement of the attenuation, if you are able to optimize the circuit to the exact same transfer function (not just level and f3) then the sound will be the same. This is not always possible though, even with a XO optimizer. This is why I have used all the varients at one time or another to get the proper frequency response. Though I tend to use L-Pads whenever possible for their practical benefits.


Regards,


Dennis

 

[cc00541]

I generally prefer to shunt some value of resistance across the tweeter. This can assist in amelioration of distortion products in the stop band, improving damping below fc, and power handling, especially if cheating' the tweeter crossover down too close to fs. In most cases, I doubt it will have any audible benefit above fc unless a very large value of series resistor is used, but the benefits below fc can easily be audible.


C

 

[tvrgeek]



Quote:
Originally Posted by cc00541 /forum/post/16962516


I generally prefer to shunt some value of resistance across the tweeter. This can assist in amelioration of distortion products in the stop band, improving damping below fc, and power handling, especially if cheating' the tweeter crossover down too close to fs. In most cases, I doubt it will have any audible benefit above fc unless a very large value of series resistor is used, but the benefits below fc can easily be audible.


C

Visualizing here: An L-Pad at the tweeter end would increase the impedance looking back at the network due to the series resistor, but lower the impedance through the lower leg of the pad. It would give a more uniform load to the network. It would seem to me, the rising impedance of the tweeter due to inductance would adversely effect the linearity above fc unless a zobel is applied because it is effectively increasing the parallel impedance of the vc and lower leg with frequency. This may not be a bad thing. I can see how it could be used to advantage if your listening position was on axis. Most tweeters are rising on axis as they assume 30 degrees or so off axis position. Of course, all speakers do this, that is what Rice and Kellogg told us, self compensating the increasing efficiency with frequency offset by increasing impedance. The pad may throw it out of whack.


Off on a tangent. Just putting a shunt at the tweeter would not effect the output level, but it would lower the impedance looking back. On the down side, the network values would have to compensate and current delivered increased. I wonder how low a value would have a sonic effect?


Anyone have a view on the series resistor being at the network input vs network output? It is how the level gets changed and was kind of my original question.


Hmmm. Need to sleep on that. I have been playing with active crossovers too long and not really thought through passive ones.

 

[vasyachkin]

i would put it between tweeter and XO


that way it would flatten tweeter's impedance and make the job easier for the crossover


thats a simplistic answer. in a perfectly optimized design there would be no distinction between, crossover and pad. it would be one circuit.

 

[tvrgeek]

At the tweeter could also help in reducing the delta in Re due to thermal effects. Not sure how much this is audible. What effect it has on damping is my first concern. Guess I need to take a side track here and do some A/B testing. I happen to be re-packaging the Dayton/Vifa's I made earlier, so I could do a couple of alternative crossovers with the same transfer function and see if I can hear a difference.


Yes, it is all a network. But for analysis, one tries to break it up a bit. I don't have MathCad so I have to visualize in my head.


Kind or related, are there any decent network CAD tools out there that one could afford? It would need to look at the complete network including source and load, power, phase, and time domains. if it could do two networks and show their interaction that would be fantastic. A few hundred at the max. The biggest problem I can see is coming up with the source and driver models. It does not sound simple.