Crossover puzzle

[BEC]

Bob,

Well, it's off the input tap and hangs off the back-side of the input cap, and it effects the impedance and FR of the high pass system. When you move the 2uF cap to the input terminal on the barrier strip, the tweeter "gets louder". You've seen this yourself on your AL-3/AK-3 plots. When you move that cap, it becomes a different kind of filter with a different transfer function. At any rate, it doesn't make it more efficient unless it can do it with less power, which I'm suspecting it doesn't.

 Dean,

I don't think so.  After all, the AA tweeter circuit comes off the same point as the A tweeter circuit does and the AA goes to a higher initial tweeter output.  I will have to think about that some more.

 Just putting together a couple of facts, the 13 uF cap is an impedance of about 1 ohm at 10 khz, so it appears to me that it's (the 13 uF cap) effect on the tweeter circuit is essentially none.  

 Bob

 

  

[Deang]

"...the AA tweeter circuit comes off the same point as the A tweeter circuit does and the AA goes to a higher initial tweeter output..."

All my plots are on my machine at work, but I think I remember that -- I good sized bump right after it comes all the way in compared to the Type A.

"...the 13 uF cap is an impedance of about 1 ohm at 10 khz..."

That sounds like a lot to me, maybe it isn't -- you'd know better than me with this kind of thing.

Dennis had a post on this once, but it would take forever to find it.

[Erik Mandaville]

I got into bed, closed my eyes, and then suddenly thought of the ground connection on the autoformer. I think the answer to all of this is actually pretty simple, and it doesn't have so much to do with the impedance of the input cap (which might be a small factor, too) as how the autoformer actually works. I'm tired, but will try to put what I think into words the best I can:

The autoformer introduces a sort of 'shunt element' or low impedance path to ground (since one end of it is connected to ground). It's various taps provide paths to ground of differing impedances, higher to lower, in order to attenuate the signal. That's why going from the #4 tap to the #3 tap reduces the output. #3 provides a lower impedance path to ground (think of ground like a water drain) than does tap #4. Tap #4 will have more turns with respect to ground than #3, so it's impedance is greater. In other words, less of the signal is shunted to ground from tap 4 than from tap 3, so it's louder. The other issue that this brings up has to do with the parallel connection with the driver in question. Each tap will obviously alter the impedance of the driver/s it's used with.

So, even though the input into the autoformer (tap #5) is the highest impedance path to ground available (and thus the loudest), it still provides a certain path to ground where a small percentage of power can be lost. input impedance on an amplifier works very similarly, as does a conventional volume control. If the input impedance on an amp is 100K ohms (that's what the input signal will see when it knocks at the grid of the first stage), and one suddenly changed that input impedance to 5k ohms (greatly reduced impedance), what has been created is a lower impedance path to ground, resulting in signal loss. In light of this, any extra resitance added to the circuit, such as the inductor that comes after the autoformer on the 4.5kHz. type 'A', will be additive in terms of attenuation -- and that's exactly what I found. The 4.5kHz type 'A' was down even lower than the 6kHz. type'A.'

Even though tap #5 (the tweeter connection point on the Klipsch type 'A') would attenuate any driver connected to it less than any of the other taps, it will very possibly, as I seemed to find, still not be quite as loud as a totally direct, no-path-to-ground connection to the input.

Erik

[BEC]

Erik,

 Think AC not DC.  The autotransformer is a very high impedance to AC.  Now hook up something to its taps, and the autotransformer reflects a multiple of that impedance back to its input.  Hook up a 16 ohm driver to tap 4 and suddenly it becomes a 32 ohm load to the signal across taps 0 and 5.

Bob 

 

[Erik Mandaville]

Bob:

You are absolutely right, I was thinking DC. it's a function of stepped changes of impedance loads seen by the amp, and what I just crawled out of bed again to correct myself about. The autoformer is essentially a part of the driver, with the various taps used the change the load seen by the amp, thereby altering the power transfer. I hinted at that in a post or two above in relation to why I thought the tweeter might be down a bit from a direct input. I then thought of what I said about the impedance changes that occur between taps, and......remembered.

What does the value of series capacitance on the tweeter suggest about it's impedance on tap 5?

If an 8ohm tweeter is connected directly to the input for 6kHz crossover, the requried capacitance is not what is obtained by the series connected 13uf and 2uf. What we do have is essentially what's needed for a 16ohm driver, correct? The tweeter is connected to the top of the autoformer, but it's behavior is still, I think, influenced by it. As for the squawker, if it becomes a 32 ohm load on tap 4, is the power transfer loss (determined by output level) greater or less than the L-pad I made for a 16 ohm driver without an autoformer? That's what I wanted to test.

I'm too tired.....

Erik

[Erik Mandaville]

We've also talked about the need to compensate for the change of impedance of the squawker by moving down to tap 3. That compensation is made with a change to the input capacitor. 6.8uf comes to mind, but I can double check that. Is that enough to rebalance the network, though? That change has addressed the changes to the squawker impedance, but the tweeter now has a 6.8uf cap (if I have that figure right) in front of it instead of the 13uf. So, the capacitance is again the sum of the series connection of those two capacitors. How can it not be? The result would be a shift in the crossover point to the tweeter, unless a change is also made in the value of the 2uf capacitor so that the new 6.8uf cap in series with the adjusted value of the tweeter cap in series with it maintains the 6k crossover point. 6.8uf and 2uf in series is about 1.5uf, which is around 6.500Hz for a 16ohm load. That's fine if one's midrange doesn't have a problem with it. Even 6k is already pretty high, however.

Erik

[BEC]

OK, We can put one part of this discussion to bed in my opinion.  Just happened that we were building a new set of type A crossovers this morning and since I always run them on a spectrum analyzer after building them, I tried both configurations for connection the tweeter circuit.  That is the standard way with the tweeter branch connected after the 13 uF cap and then connected the tweeter branch to the input side of the 13 uF cap.

 The only difference I could see in the voltage to the tweeter is in the rise of the slope as the tweeter comes on.  The difference in that slope is considerable less than one db and the ultimate output of the tweeter circuit is identical after about 7 khz.  

To summarize, there might be a possibility of hearing the difference between these two configurations with a steady tone at around 5 to 6 khz if one can hear a bit better than the standard 1 db norm for most folks.  With music or above 7 khz steady tones, my instruments indicate you haven't got a chance.

Now, one more thing.  This test was done with all new and close to perfect parts.  If you do this with old caps with high ESR or other than good polypropylene caps, you might see a much larger effect.  The 13 uF cap I used has an ESR of about .003 ohms.   If that 13 uF cap were a mylar, PIO, or old original Klipsch film and foil, the difference would have been greater and the chance of hearing it would be better.

Bob

 

  

[Erik Mandaville]

Thanks for going to the trouble of testing that. I do appreciate it! It does not translate to what I heard, but as you said the parts in question can very likely make a difference. Oddly enough, as far as the squawker goes, which was the main point of the idea behind this new pair of networks, uses an old motor-run capacitor in the bandpass. Yet it is still obviously more loud with the same input than a type 'A' modified for 4.5kHz. crossover using brand new poly caps.

BTW: If you connected the tweeter directly to the input, what value of capacitor was being used and what was the crossover frequency? It's not going to change your test results, but I'm just curious. The difference of the squawker on a resistor-based L-pad was the main focus of this test. It seems in both cases amplifier power is *lost* -- one through intentional impedance load variations by way of the autoformer; the other through heat using resistors. Given those two factors, the L-pad was more efficient, while the network as a whole sounds not only louder, but better. My opinion.

Thanks again!

Erik

[Erik Mandaville]

Bob:

Is the usual value of 2uf what you used between the input and an 8ohm tweeter (bypassing the autoformer connection)? That puts the crossover frequency near 10,000Hz, no? We are almost above timberline and in the snow at that point! Again, I'm just curious about the values that were used. I'm going to try that 4kHz crossover later this afternoon. I think that frquency is actually a little better suited to the value of the parts I have in the band pass right now.

Erik

[BEC]

Erik, 

Testing was with a standard type A configuration in every respect.  The crossover point did not move noticably.  The two traces are so close that I could still call both of them 6000.

 Bob Crites

 

[Deang]

Thanks for doing the test Bob, interesting.

[BEC]

I sort of hate to post this trace, because I did not take the time to make a pretty trace out of it.  So, as soon as I could see enough of each trace to know how it would end up, I stopped the sampling and plotted it.  If I were making it to be posted, I would have waited longer for the sampling to fill in all the dips.  Anyway, just looking at the black and red trace, you can see that the difference between the output level and the crossover point on the two traces (red and black) is so small as to be insignificant.

Bob

 

[Erik Mandaville]

Thank you.

Pretty enough for me, Bob.

Erik

[Erik Mandaville]

Bob:

I think the tweeter is actually still associated with the autoformer, even though you've connected it on the other side of the 13uf cap. It's impedance characteristics are still being influenced by it, I think. This would be if the 13uf capacitor is also connected to the input (which it has to be for your test for the behavior of the whole network), even though the tweeter branch is connected on the opposite end from where it usually is. With this network, it's hard to get away from that. It's the load impedance issue I'm thinking about, not the impedance of the series connected 2uf and 13uf capacitors. In other words, it may be that the autoformer still influences the tweeter impedance, despite the fact that you've connected on the other end of the input cap. A connection with the autoformer remains. That would explain why the 2uf cap still results in a crossover at 6k, and is probably why there isn't much change from one to the other. The value of the series combined 13uf and 2uf in the normal curcuit is just under 2uf, so using the 2uf from the tweeter branch would have very closely approximated that in your test.

That's why the network I made sounds different (forget the term better, it's subjective). The autoformer is out completely.

Erik

[BEC]

Bob:

I think the tweeter is actually still associated with the autoformer, even though you've connected it on the other side of the 13uf cap.  It's impedance characteristics are still being influenced by it, I think. This would be if the 13uf capacitor is also connected to the input (which it has to be for your test for the behavior of the whole network), even though the tweeter branch is connected on the opposite end from where it usually is.  With this network, it's hard to get away from that.  It's the load impedance issue I'm thinking about, not the impedance of the series connected 2uf and 13uf capacitors.  In other words, it may be that the autoformer still influences the tweeter impedance, despite the fact that you've connected on the other end of the input cap.  A connection with the autoformer remains.  That would explain why the 2uf cap still results in a crossover at 6k, and is probably why there isn't much change from one to the other.  The value of the series combined 13uf and 2uf in the normal curcuit is just under 2uf, so using the 2uf from the tweeter branch would have very closely approximated that in your test.

That's why the network I made sounds different (forget the term better, it's subjective).  The autoformer is out completely.

Erik

 Well, that would, of course, take another test.  But, perhaps we can just reason this one out.  The amp is a source.  Imagine connecting another complete network to the same amp output.  We do not expect anything in one of those networks to effect the response of the other.  We expect both networks to function in accordance with the same specs when they are in parallel that they did when they were separate.

 Bob

 

[Erik Mandaville]

"Well, that would, of course, take another test."

Why don't we just go enjoy some music, instead. I want to add that 3uf cap to the 2 already there for that 4k cross. I decided the oil cans were too expensive, so just bought some new/cheap mylars @ 50 volts (plenty for all of my amps). I'm curious about the change, if any.

Erik

[Erik Mandaville]

Tuning this new network by ear, the 4k crossover sounds just fine. It gets better overall. I've found, though, that I can only get my SPL meter within about 2 dB of where it needs to be in order to balance the system with the Lexicon's internal test tone. I didn't have that problem before using the other network. It was always close with the Moondogs (which have some gain and sensitivity characteristics that are different from the stock version), but I could always get the needle to hit 75dB.

In order to maintain the overall balance I have right now, I'm also going to pad the tweeter a little. So, the squawker will be on a 6dB L-pad, with the tweeter, for this intitial test, down 3dB from a direct input. It means pulling the networks out again, removing the old L-pad and installing two new ones, but It's worth it!

Erik