Which is the best crossover configuration?
After the marathon session dealing with Cambridge Audio's messed up crossovers for 3 days, I find myself wondering what the best crossover configuration is... even if there is a best... As a secondary quandry I also find myself wondering how many so called "high end" speakers are being hobbled by bad crossovers...
Here is a Spice simulation of a super simple first order crossover...
These are the predicted curves...
For those more familiar with XSim, here is the same circuit...
Note that the tweeter is reversed ... this is because the woofer and tweeter are each 90 degrees off true phase for a total of 180 degrees out of phase at the crossover point.
Now the BIG question .... why does it need to be more difficult than that?
(No sarcasm... a real question.)
In a perfect world, this would be all you'd need. Unfortunately we are reminded many times a day that that isn't so. Just looking at the response graphs of any of the drivers tells us that there is no such thing as a smooth graph or response. Couple that with an enclosure and you really have some dancing to do with response.
All these peaks, valleys, plateaus, and cliffs that make up a response graph dictate just how much you "might" need to do to get as close as possible to "perfect" in the end. I think that some companies do go overboard, but that's my opinion. I mean, unless they have a true anechoic chamber to physically test the speaker system itself, and a true lab grade mic, etc., then how can they trust that "the massive dead spot between X and Y" is real or just a byproduct of their test setup? Really, is little Johnny with his XBox microphone running a sweep on a speaker laying on the bed and the mic on a chair using a salvaged portable radio amp going to get good results? Kind of sounds like what Cambridge Audio did. 😉
Trust me on this one ... I know exactly how imperfect this world is. I contend with it everytime I pick up my service kit to head out on a call. In fact, that's how I make my money.
Still I don't see the need to go beyond second order crossovers... and even then I question the value of any part that ends up shunting amplifier output to ground. It never has, and maybe never will, make any sense to me.
I'll bet you three manhole covers (already positioned on your street for convenient pick up) CA hired an audiophile to voice their speakers... and if they didn't do that I'd bet they picked up some pre-packaged chinese
junk stuff and put their name on it.
40 years ago, when we had to work with pencil and paper and didn't even have software or today's specifications to work with speakers were better. In fact they were much better ... Give me a nice set of L-100s or a set of Heresies any day... This new stuff sucks.
There's too many reasons to list on why this isn't typical in real life. It looks good in a perfect situation, but there is no real perfect situation;. Here are some basic issues that aren't being addressed. The first is cone breakup with the woofer. You will always get some form of breakup that will cause distortion and or constructive/deconstructive waves. Without compensating for that you will have issues there. You also have on and off axis performance. In your simulation you are not taking into account for off-axis performance. Depending on where your woofer starts to beam will determine where you tweeter will want to start it's crossover point. But then you have the problem with your tweeters FS. That should typically be at least 20db down from your nominal response. Depending on the crossover point, you might need at least a 3rd order or higher on the tweeter to protect it and or decrease distortion. Once agin, we haven't looked at distortion or even a waterfall graph. All which should be taken into account when creating a crossover. Three more things that we haven't talked about, that should be discussed. The first a tweeter is almost always more sensative than a woofer/mid, so you will at least need an L-pad. Once we take into account of baffle-step, that will also affect the crossover point, without compensating for that. Then the last thing we haven't talked about if you are showing a typical LR 2nd order. which in a perfect world is 6DB down and a flat response with the rest of the graph and in perfect phase. However, that is rarely the case. To get it flat with the rest of your response, you may need to compensate for the phase and or compensate for any humps in the frequency.
Thaks Nick and Charlie .... Isn't what you're describing going to be rather inefficient with lots of currents going to ground?
I've seen and worked with 2nd and 3rd order crossovers, both 2 and 3 way before and it always stuck me there just has to be a better way.
For example: In one 2nd order low pass I rebuilt some time back, the cap to ground was handling 2 and 3 amps of current at 20 watts, which is what caused it's short lifetime.
Last night I spent a few minutes messing with the example crossovers in XSim. Wow, that's some messy circuitry. It honestly looks like someone just kept throwing parts at the problem without trying to address it directly.
This circuit... (Example2Way.dxo)
... is wildly more complex than it needs to be.
When I strip it down to just a simple first order and invert the tweeter I get a slightly different but fully acceptable result...
... it is still within plus or minus 5db.
So why go to all that trouble when he could have just reversed the wires on the tweeter?
For that matter he could have done this....
... using R1 and C1 to compensate for the high frequency roll off, with a much smaller inductor in the low pass, using the woofer's own inductance as part of the filter.
Well my philosophy has always been make the structure as acoustically mated to the driver(s) as possible before even thinking of the electronics, whatever they may be. That way, you have less to interfere with the signal.
Now, as to the better way, that would be using DSP with individual driver amping (however many that may be.) Dealing with the individual frequency problems before the amp does not reduce efficiency and gives greater control to your response curve. It really is that simple.
As for me, I'm more of a curmudgeon when it comes to the hobby. I like the old ways and it's what I know. I may or may not do DSP at some point, but right now it would be way off in the future. At least until I design "my perfect enclosure." 😉
*** I see you were replying when I was too. (had to vacate the office while the landlord took care of a wasp situation outside my window)
DSP is too easy... Seriously, I'm starting to think that DSP and Computer Simulation are taking away from the speaker builder's craft.
The last circuit in my post above is very old school. I would have first designed the cabinet and done a mock up, mounted the woofer with plates over the mid and tweeter openings and seen what the woofer does in my box, I would try various coils to get the best response, then I would have added the mids and worked out a minimum parts bandpass to get it to play nicely with the woofer, and finally a tweeter with the same process, getting it to play nicely with the rest... all done by testing in the cabinet itself.
What the designer in those XSim examples seems to be missing is that (for example) if you have a red hot tweeter you don't always need an LPad; you can just use a smaller series capacitor and let it's reactance attenuate the tweeter for you. Note that the series high pass capacitor (C2) in my example above is only 0.39uf ... it's series reactance replaces the LPad in the original design. The crossover point is determined by the woofer's behaviour, not the values of caps and coils. The tweeter is tuned to rise to the woofer's level once the woofer is done. (I'm sure a bit more tweaking of part values could get a better result, but it is only an example.)
My feeling, especially with that CA crossover fiasco is that in a lot of cases those using these computer sims do not have a solid backing in electronics... they're just trying random stuff until they get what they want, without fully understanding what they are doing. When I can take the examples, using the same driver files, and strip it down from 10 components to 4, I tend to get rather suspicious of the original designer's skills.
I still remember my teacher's admonition when studying electronics ... "First just get it to work at all, then take out everything that isn't needed to keep it working." ... and so far he's been right about that every time.
BTW... if you are looking for a crossover design with constant impedance and no phase shift, here it is....
And here, using the FRD and ZMA files from the Example2Way crossover and only 3 components...
More old-school stuff... 😉
(@tvor-ceasar Yes, it does appear we were writing at the same time.)
Being as Nick is prepping to move and I'm on a whale watching trip, it'll be a while. Plus I need to go more in depth study of the others in order to really understand them well. I have some basic understanding, but that's about it at the moment.
Anyone else out there want to chime in? It takes more than 2 or 3 to make a community.
Get back to you later tonight. Diesels are running up.
No offence, Charlie, but it would be really nice to see more than you and Nick in these conversations...
One example ...
Why aren't we critiquing each other's designs?
Everyone would learn from that....
No offense taken. That's why I asked for others to join in. I, as well, would like to learn more than I know now. There are many more than us few constant posters who could share their experiences for the furthering of the hobby / passion.
On the search for aquatic mammals. 🐬🐋🐳👋
I'll take a stab at critiquing your crossover simplifications above. Your first simplification removing some components is far less flat. You have a strong peak in 1.5-4k, which is where our hearing is most sensitive. This would make the speaker sound harsh. You also have a falling response after 4k. This would make the speaker sound dull with symbols and the like being less audible.
Your first order crossover is better on the frequency response but not as good as the original. That 7db change from 1.5 -2k is going to be very audible.
My last critique is all you have provided is a summed on axis response. As Nick pointed out above, first order crossovers are usually problematic due to the wide operating range required for the drivers. Woofer breakup and tweeter rising distortion near Fs become very audible when using shallow slopes. Add to this the off-axis considerations, and there are very good reasons for using steeper slopes than first order crossovers. Yes, steeper slopes are less efficient. However, the purpose of the crossover is to filter out frequencies outside the operating range of each driver. The concept of a crossover is inherently inefficient, but they are still necessary to remove cone breakup, protect the tweeter, and match two different sized drivers over the small range of frequencies their off axis response curves are similar.