Why would anyone do this?
I got a pair of crossovers in for rebuilding. They're the strangest thing I've ever seen.
Here is the schematic ...
This is the frequency response...
I'm guessing at the inductor values, backtracking from observed behaviour.
In the boxes, the woofer is very uneven and the tweeter rolls off noticeably at 12 khz.
Is there really some good reason to build something like this?
Should I advise my client to look into replacing them?
Are you sure the schematic is correct? It doesn't make a lot of sense. Right now the woofer is showing a 2nd order with a parallel RLC notch filter. But it would make more sense to be a 3rd order with a zobel. The tweeter is just showing a series RLC notch filter...which I guess can work, but once again doesn't make a lot of sense. It would make more sense if it was a second order with the resistor being used for an Lpad.
It looks like you are testing it with a dummy load, instead of the actual tweeter or woofer. So I couldn't say for sure, since a loudspeaker isn't a flat impedance. That resistor in series is going to really affect the impedance. It would be better to have an actual frequency response measurement to see exactly what is going on. However, it isn't atypical to have a downward sloping tweeter. This is called a house curve and is implemented quite often.
The schematic is correct, I took it directly from the PCB. I must have retraced it about 20 times trying to find my mistake before accepting what I was seeing...
Not a dummy load ... The schematic is a Spice model, trying to figure out what was going on. R3 and R4 represent the speaker loads. The curves in the frequency plot are from that simulation as I tried to figure out what coils they were using. The actual speaker is a Cambridge Audio SX-50, the little brother of the pair I got stuck with because of the glued on tweeters.
It would be interesting to see what happens in your Crossover software, but I have no idea of the parameters for the speakers.
The speaker itself appears (without extensive testing) to work better if I remove L1, L3, C2 and R2 from the boards. In particular L1 appears completely gratuitous.
The only reason I can think of for using series LC circuits --which exhibit a very low impedance at resonance-- in a situation like this is to counter some weird impedance issue... but why bother when most modern amplifiers are not freaked out by it?
I've seen some weird stuff in my days, crossovers that run the woofer on a 1:1 transformer, small coils across tweeters and such... but this one definitely takes the cake.
For now, I've ordered a pair of "normal" crossovers to replace these ugly things.
Cambridge Audio has been nothing but a pain in the ass for me, so I think I'm going to suggest that he take a look at the Elacs and replace those speakers entirely.
Okay. The woofer is a normal 2nd order crossover with a parallel RLC circuit. The RLC is probably taking care of a peak or cone breakup in the woofer. It shouldn't sound better when removing it, unless you are attuned to that frequency. But either way, it will most likely add distortion to it by removing it. The tweeter, I can't say for sure what their goal was for that. It would make more sense if it was a 2nd order with a resistor either in front or behind to attenuate it. But I guess a notch filter will also work. Either way, it definitely is not common.
Thanks. I've been testing for most of the afternoon. I also discovered that my SX-60s have the same crossover in them ... so now I've got 4 of them to mess with.
The woofer does sound better without those parts. All of a sudden I've got better low bass and clarity than ever from the little guys, and it does show on REW with much improved response below 100hz. We'll see if that holds true on my set in the next day or two. The feeling now, when listening is that the woofer no longer sounds constrained, like it's finally free to move and since it handles a significant part of the midrange that sounds better too.
The coil in the tweeter circuit is pretty much acting like a low pass when above it's resonant frequency (about 1800 hz) and appears to be responsible for a pretty serious roll off in the highs, starting around 8khz. So it's coming out too... I'm still debating whether I should re-wire it as a second order high pass or just ditch the coil outright and go with first order high pass instead. I guess that depends on how well matched the woofer and tweeter are... So far they seem ok. Their outputs are almost equal, until that coil starts cutting the tweeter down.
Just to make it more interesting, they're crossing over at 2500hz, right in the middle of the mid/voice range so any mismatch will be plenty apparent.
I really don't understand why this was done. My friend suggested "Maybe this is how they get you to buy a sub-woofer"... and he might be right.
There is something else going on if you are getting more bass when you remove the RLC circuit. That circuit would not affect the bass response at all. It is a narrow band attenuation circuit and those values you gave on an 8ohm woofer would be in the midrange, not bass frequencies. Here is a quick sim of what it would do and something like the 8ohm Esoteric driver. Notice it gets rid of a frequency peak around 750hz and affects impedance from about 100hz to 1Khz. So unless you have traced the schematic wrong or got the wrong values, what you are saying doesn't make any sense. It's pretty easy to check. Just take some impedance graphs before and after along with a frequency response.
Thanks for that ... I at first thought this might be the case. But with the parts removed, there's no peak there, to speak of. At least now we know what it was doing.
Now I have learned they use the same crossover in both the SX-50 and SX-60 ... so maybe it's addressing a problem the SX-50 doesn't have???
Of course. There is a peak for sure. That is what that type of filter is taking care of. The inductor and capacitor are acting like a high and low pass and the resistor is attenuating that narrow area down. However, it might be below your combined response. You would need to take a measurement of the woofer only. If you say you can't see it going down, then it is most likely taking care of intermodulation distortion.
I have done close mike tests of both woofers and tweeters and if there is a peak there it's got to be pretty tiny. As I said it may be in the SX-60s and not in the SX-50s but I won't know that until sometime tomorrow. But you might think that if there was no peak in the 50s, it would show up as a dip because of that filter. I'm not sure... at this point just speculating and running sweeps... so far everything seems go for the stripped down crossovers... (I like simple... Simple is good)
Granted, REW isn't the best software for this kind of testing.
Well, the SX-50s are done. Tomorrow will be my SX-60s.
So here's how to fix a Cambridge Audio Crossover, I think...
Just takin' out the trash!
And here's the result ... Before in Red, After in Black...
Note... I close miked these at about a foot, but there might still be a little room in there as well.
The effect seems to be a more open, expressive sounding speaker. It may be confirmation bias but I think they do sound better.
Nick, thank you so much for your help on this. I've never seen that kind of circuitry in a crossover before and your insight was most helpful.
Thinking more about this.
There is something else going on if you are getting more bass when you remove the RLC circuit. That circuit would not affect the bass response at all.
Here's the thing... any circuit that is not in series with the woofer is wasting energy. When currents travel to ground they don't make sound. 😎
Now we need to ask just how much wasted energy there is. A series resonant circuit is not a switch. At frequencies near resonance, which happens when the impedance of the coil equals that of the capacitor, a considerable amount of current can flow through the filter circuit.
There is a Q (Qualitative) factor for resonant circuits that helps us determine just how much stray current can flow... the lower the Q factor the more leakage there will be and the lower the Q the further from resonance this can happen.
Also a low Q circuit, which is common with large value parts at low frequencies, can exhibit a resonant behaviour that can be very broad, sometimes into octaves of bandwidth.
A full tutorial is HERE ... Note that it confirms there is always some current flowing, even octaves away from resonance.
So, that series RLC circuit in the crossover is not a switch and it's likely to have effects well away from it's resonant frequency (~800 hz)
Now add the final notice that the notch filter is itself in series with another inductor. It follows the low pass filter for the woofer. So it naturally forms a voltage divider with the input coil, acting in some ways like a frequency dependent volume control. As the notch filter approached resonance and the input coil began tapering off, a fair bit of energy intended for the woofer ends up going straight to ground.
Now I only had the smaller SX-50s for two days, so I am less familiar with them than I am with my SX-60s. But I have frequently noticed that they don't scale well with changes in volume. They can sound compressed, kind of weak and lifeless, at times and I'm betting that's the effect of that notch filter circuit shunting woofer power off to ground.
In any case the change in character was immediately apparent when I disconnected the circuit... on ... compressed ... off ... more dynamic. My client/friend also heard the effect and it could be measured with a voltmeter. When I connected the filter the voltage across the woofer was decreased by a small amount. At resonance, given that it has a series resistance of 6 ohms, it was actually stealing half of the woofer's energy.
How broadly this effect played out is unknown, at the moment but I will investigate it more carefully when doing my own speakers in the next couple of days.
Ok, now the real fun begins ... Starting on the SX-60s, which I use every day.
Here is the crossover schematic with corrected part values...
This is a Spice simulation, so here are the predicted curves...
(Note that I marked the crossover frequency)
And this is a one meter sweep from REW ...
If we look at the microphone sweep from REW we can see that it is showing us essentially what Spice is showing. The woofer is up nearly 10db from the tweeter and the tweeter rolls of rather badly above about 8khz.
Please note that I am not yet sure what is going on with that 100hz bulge; it might be the crossover, a room node or maybe an artifact of the port or cabinet.
Of particular interest is the grey trace in the Spice plots. It shows the voltage across R2 in the schematic and it shows that the series LC notch filter is carrying significant current --greater than 12db-- from about 100hz to just over 3khz. That's not a very sharp filter.
So now to start taking out the trash ... I'll write more when I finish and can get some new sweeps...
One speaker is now done. Here are some results...
This is the final crossover schematic...
Lots of stuff taken out...
Here is the final Spice traces for it...
This is a compound of 2 close mic runs of REW, the woofer is brown the tweeter is blue. Keep in mind that some bass got into the tweeter scan and some treble got into the woofer scan, but you can clearly see the crossover region...
This is a sweep of the bass port, which is surprisingly active...
And finally this is a scan from 1 meter, on axis...
Objectively this is a fair bit better. There is no longer a great difference between the levels of the woofer and tweeter, the high frequency roll off is gone and everything is within +-5db of the average level (55db). They are also a bit louder, maybe 3 or 4db. So not bad at all.
Subjectively, listening for a couple of hours suggests that the bass is much improved, more open and less constrained sounding. The midrange is not a lot different, maybe a little clearer which I'm happy for. The treble is surprising; those stupid glue on tweeters actually sound pretty good.
Overall, I think this is an improvement... but there may be some self-deception at work so I'm going to finish the other one and listen for a couple of days and let the bias wear off before deciding what to do... I already have 2 crossovers on order for the SX-50s I worked on yesterday, I may just increase that to 4.
I still have no idea why Cambridge Audio would do this. It doesn't make a lot of sense...
Is there a lesson here? Maybe... KISS... "Keep It Simple Stupid"...
Thanks again to Nick for his advice and help with this problem! Greatly appreciated.
Whew ... everything is done and back together now. Just one last detail...
That big old duck tail at 50hz which is the result of an over-active port. It only took a couple of hours to realize the speakers were now a bit boomy sounding and I needed to tame the bass ports.
Well; foam plastic
Butt Port Plugs to the rescue. This is a REW sweep at 1 meter. Green is no plugs. Red is with the plugs inserted.
By way of general comment, I would advise anyone who asks against buying Cambridge Audio speakers. Their electronics seem OK but their speakers are total crap. I've had nothing but trouble with them since I started with them. First it was glued on tweeters that fell off in my hands, then woofer dress rings that keep falling off, now this crossover thing and... the last straw... an overactive port.
But the job is done now. I will be ordering new speakers in a month or so (when money permits) and they won't be Cambridge Audio!