BillFitzmaurice.info

I've got some time this summer and always like making things, so decided to make a crossover, as it'll be somewhat cheaper for a student like me, plus it's interesting, and allows me to put my college electronics knowledge to some use!

So I'm planning on a switch-able frequency 24db/oct crossover, with a high-pass subsonic filter.

2 things which I'm considering at the moment:

Whether to have the high and low pass parts in one, controlled by the same dial (ensuring good spectral and phase coherence within the unit), or having them separately controllable (such as in some plans where it suggests using a lower low pass frequency to compensate for rising mid-bass response, but I was thinking that should be solved with an EQ) - I'm heading toward the former.

The other issue is just the frequencies at which to cross - I'm probably going to use a 5-way switch, and am thinking:
120, 100, 90, 80 and 70 Hz - I'm thinking these are fairly reasonable to be used in a variety of set-ups eg home theatre but also some live sound.

Any thoughts on those points would be muchos appreciated

Use a state variable filter, for sweepable corner frequency.

Bill Fitzmaurice wrote:Use a state variable filter, for sweepable corner frequency.

Yep - that's what I'll be using - but for it to be continuously variable it requires 4-gang pots with accurate tracking tolerance - which are pretty sparse, and expensive, so was going to stick with switch-able - unless there is another way at a relative price point? (not too fond of pricey resistive optocouplers/whipping up LDR-LED combos) - thanks for quick reply too bill

So if i'm sticking with switch-able, unless I'm missing something, do those frequencies sound like they'll cover most eventualities? Also, having it joined rather than a separate high and low pass would be best still?

I would build in low pass as well for sub output...

Grant Bunter wrote:I would build in low pass as well for sub output...

It's only 2 way, so the low pass would be part of the design

Hey, what did you come up with on this?

I know variable Sallen-Key is a pain -- I built a 12dB/oct. variable lowpass for a plate amp project. I wanted to do 4th-order, but had the same problem with trying to find an inexpensive, 4-gang pot with good tracking.

However, it crossed my mind recently to use a microcontroller combined with one of those digitally-controlled resistor ladder potentiometers. The only thing I'm wary of is that they appear to be single-supply devices with limited voltage swing, due to the transistor switching. I'm used to running my audio stages at +/- 15V, so there's a bit of redesign to do...

Perhaps dip switches to select taps on a resistor series string. The frequency resolution you want to achieve will dictate how many dip switches and the resistor values you would need for each filter stage, and then just multiply X4. It's not exactly continuously sweepable, but once you decide on a crossover frequency, why would you need to change it frequently?

Thanks for comments.
I gave up looking for suitable pots, only with a chance of using them if I could pick up an old broken crossover from somewhere, but that wasn't looking likely.
I ended up going with these, and then filling the gaps with resistors. That way tracking issues aren't an issue, they're mighty cheap in comparison to pots (inc resistors), and still give me 5 frequency options.

The only thing I'm still undecided on is whether to do separate high and low pass filter sections, with different frequencies. As so many of the sub designs here would benefit from a roll-off in the mid-bass which could be provided by a 'gap' in the crossover, but it just seems like the wrong way to do it. Also I might be battling with phase issues/have to include an all-pass filter to cover that, but that's just more work(!)

I'm still wanting to keep things simple, so will probably go with the integrated low and high pass, and then build a twin peaking parametric/shelving filter if needs be to sort out the mid-bass/room response.

I've simulated everything how I want it on Spice and it seems fine. If possible, I don't want to built two discrete circuits for Left and Right, I'd rather find a decent way of high passing the L and R, but then have a summed feed to go into a low-pass.

arghhhh decisions!

No reason to duplicate the LF feed. I would assume if you came this far, you know how to sum signals, but your wording suggests you're looking for the right approach. Here's what I did: Take your L and R inputs (after any buffers or whatever's in your input stage), put them through 10-47k resistors, and combine them at the + input of an op-amp. Output from the op-amp is a summed feed that can be low-passed. A typical unity-gain negative feedback circuit would do the trick.

As to your options, if I were you, I would allow for separate XO points -- it seems unnecessary at first, but it's a very handy feature to have. Also level controls are nice, but the necessity of that depends more on what's next in your chain. You might want to consider switchable 2nd/4th order. Easy to do -- just bypass a stage.

The only problem with those rotary wafer switches is the silver contacts; they tend to oxidize and become intermittent as time goes by. Last thing you want to do is pop a driver because the switch contacts became dirty. Some of my old communications receivers use those for band switching. Albeit, some of the radios are 50+ years old, but they also require contact cleaner a couple times a year to maintain reliable operation.

It'll work for awhile though. I would still consider the dip switches; they usually have gold plated contacts. CMOS logic circuits won't tolerate intermittent contacts on a switch so usually they are a quality, low resistance, positive contact.