Let's discuss series and parallel crossovers

[Jim85IROC]

I've been educating myself on crossover design, and in my searches I came across a lot of series crossover information that's very interresting. Parallel crossovers, although still very difficult to design with any sort of accuracy, are at least very easy to understand. The series crossover is a bit more abstract and not as easy for me to analyze by looking at. I found a couple posts by sreten that included some very informative pictures. Those are in the following thread:
series crossover/how to wire ??

It made sense that the component values stayed the same. I started digging to see what else was the same. I was surprised. I redrew a couple circuits, and here's what I came up with. First is a typical first order series circuit:


Next is that same circuit, redrawn:

What's interresting about seeing it this way, is that if you simply remove the electrical connection at the "dot", you wind up with the parallel version of the same circuit. That one electrical contact point is the only difference between the series and parallel circuit.

So, I decided to redraw the Acoustic Reality circuit. Here's the original:


Here it is redrawn:

Now, if you remove that little dot, you wind up with a really interresting parallel circuit.

Now it would seem that with the conventional series crossover, you'd wind up with the same overall filtering properties as with the equivalent parallel crossover. My main source of confusion is just how that one extra connection point in the series circuit effects the behavior of the entire circuit. Series circuits have a reputation for being more forgiving if values aren't 100% spot-on, and they seem to have better impedance characteristics. Why?

Moving on... if it's safe to assume that the series and parallel circuits will provide similar results with similar components, then what about the AR circuit? What happens if you modify that into a parallel config? From the looks of it, the two inductors in the AR circuit are just combining to form one different value. Fair enough. What really boggles me is that the AR circuit is supposed to work without the capacitor, provided you change the R1 value. This is telling me that L1 in parallel with the tweeter is what's performing the high-pass functions. In order for L1 to perform high-pass on the tweeter and low pass on the woofer, the R1 value would need to be such that the tweeter and woofer have the same impedance. This of course ignores the functionality of L2 for now because I'm too lazy to do all that math. Is anybody following me here?

The similarity between the series and parallel circuits exists for higher order systems as well. It's always the same circuit, with the single connection point that separates series and parallel. Can somebody please explain to me exactly what's going on inside the series circuit that makes it so much more forgiving to slight changes in component values? What's really boggling my mind is how the AR design can supposidly work so well for so many drivers without making major component changes.

Anybody care to help me understand this?

[Cal Weldon]

Hi Jim,

Very interesting indeed. I'm glad you redrew the Xovers to make that more plain. If you remove the dot and allow those wires to cross without connecting, then you seem to have a parallel circuit, the only difference being that the choke and cap are on the "negative" side instead of the "positive" which doesn't matter a spit.

When they are connected, it also appears that the choke and cap act as a bypass for the opposite drivers like a 2nd order does.

Am I seeing this correctly? Perhaps others more experienced in this matter might shed some light.

Cal

[Svante]

IMO, please flame me if I am wrong, series crossovers have no advandtages over parallel ones. They seem terribly much harder to design though. You really would need a simulation software and you'd really have to simulate the impedance of the drivers carefully. even more than with parallel filters. To try to understand the behaviuor other than very roughly is deemed to fail. Either build it and measure, or simulate.
Now, in the end you may very well end up with a filter just as good as a parallel one, but hardly better, and with a lot more effort.
BTW, for the first order filter, LP filtering for the woofer is mainly done by the parallel capacitor, and the LP filtering of the tweeter is taken care of by the inductor (which is effectively ruined by any series resistance in the coil).
BTW2 The redraw in your fourth image is wrong, L1 should be in parallel with the tweeter.
Probably I did not answer your question(s), but hopefully I introduced some doubts to you regarding series filters.

OK, flame me!

[Jim85IROC]

BTW2 The redraw in your fourth image is wrong, L1 should be in parallel with the tweeter.

Not really. When drawing out that schematic, I followed the same reconfiguration as I did in the single order filters above. I just redrew it in a more conventional manner.

[Svante]

Originally posted by Jim85IROC

Not really. When drawing out that schematic, I followed the same reconfiguration as I did in the single order filters above. I just redrew it in a more conventional manner.

Are you seriously saying the pictures 3 & 4 in you first post are equivalent? BTW3 isn't the voltage source shorted when the dot is there?

[sreten]

Sorry, but the redrawn AR circuit doesn't look correct to me.

Only first order series crossovers have the interesting property
of changing the c/o frequency if component values are incorrect
such that the responses still sum to unity.
This makes a series 1st order crossover a good "fit it and hope"
crossover for the amateur.

And it is only the first order case where the component values
are the same for both series and parallel connection.


2nd order Butterworth's for 8R :

sreten.

[Svante]

Originally posted by sreten

This makes a series 1st order crossover a good "fit it and hope"
crossover for the amateur.

But what if the coil has a series resistance. Will there be much of a HP effect then?

[sreten]

Originally posted by Svante


But what if the coil has a series resistance. Will there be much of a HP effect then?

if you look at the 1st order c/o vertically the sum of the voltages
across the drivers and the sum of the voltages across the c/o
components, both must sum to unity.

sreten.

[Svante]

Originally posted by sreten



if you look at the 1st order c/o vertically the sum of the voltages
across the drivers and the sum of the voltages across the c/o
components, both must sum to unity.

sreten.

Yes, but I am more concerned about the stopband for the tweeter. I agree that the sum will be 1, but I don't think it is good that say 10% of the voltage at 50Hz goes to the tweeter.

[Svante]

Also, "fit and hope" can lead to a nastyly low impedance as the coil and capacitor form a series resonance circuit. Given that the impedances of the speakers are far from resistive, this is not very unlikely. The first order parallel filter has no such risk.