active crossover board capabilities - I want your input

[CharlieLaub]

It seems that there have been several different active crossover boards for DIY loudspeakers put forth on this forum. Well, I'd like to make yet another permutation of this highly useful and needed piece of hardware.

The capabilities that I would like to include on the "basic board" are:

* Baffle Step with adjustable frequency and gain
* One equalizer band with adjustable center frequency and cut/boost level
* One second order symmetric LP/HP filters with adjustable Q and frequency

The filter and equalizer functionality of the basic board will be expandable using an add on board that will have several second order circuit sections that can implement either an HP/LP filter or a BP. The HP and LP outputs can be used to increase the order of the crossover on the basic board, or to make a 4- or higher way second order crossover. The BP outputs can be routed back to create additional equalizer bands, useful if your drivers have breakup peaks that need to be notched out, or if there is a dip in the response that needs to be lifted to flatten the FR. The add-on board will probably have 4 second order sections.

I want the circuit to be general in the sense that it could be built up fully and then, by adjusting potentiometers, the parameters (e.g. Q, frequency, etc.) can be modified as needed. In this way the board would be re-usable, or adjustable at a later date to accommodate a change or to implement a tweak in the frequency response of the speaker. Each 2nd order section on the add-on board will have its own I/O connectors, so you can configure it in various ways as needed.

All of the filter functions will be implemented using 3-op-amp state-variable filters. The advantage of the SVFs is that all parameters are independently adjustable via potentiometers (trimmer types). I also plan to make it possible to convert all the potentiometers to fixed resistors.

I'd like to get some feedback on this concept, and how much interest there would be in such a board. Would you be interested in bare PCBs or fully assembled and ready to go?

-Charlie

[BobEllis]

Hi Charlie,

I understand your desire for a prototyping board that can also be used for a finished product. However, the board real estate required for the pots and resistors to replace them would mean that you'll end up with a fairly large board. It might make more sense to make a prototyping board and a finished board. I'd like to see a board with SMD parts - even though it would be a challenge to assemble at my age and I have a boatload of through hole OPA2134s.

You should have some bandwidth limiting on the input. You may have thought of this already but didn't mention it. Allow for a decent size cap - you'll see why below.

You might want to breadboard a state variable filter and a Sallen-Key filter to see if you can hear a difference and like the state variable sound. I've heard that state variables don't sound as good, although there are plenty of folks who love the state variable Marchand XOs.

Jens' AF-4 implementation of the Sallen-Key gives adjustable Q while holding the gain to 1 using two opamps per 12 db/octave section. You could use pots for frequency and Q setting.

Have you simulated to see what happens when your pots don't track? Might be an issue, might not. You may be able to get some frequency spread to adjust phase adjusting them individually, but then the math gets more interesting.

If you cannot get phase adjustment in the state variable filter, I strongly suggest that you provide and all pass section or two (another argument for SMD) You can use a pot for the prototype stage to set the shift. If you set up the baffle step input buffer with jumpers to allow the right connections you can use it as an all pass in a three way set up.

An option to make your settings adjustable would be to use sockets on the board to allow replacement of resistors or connection of pots, like you would probably use for your add on board. They can usually be snapped apart to allow single terminals to be placed on 10mm pitch.

Are you planning a single channel or stereo board? If single channel, 4 second order filters on the add on board seems overkill. How about 2 second orders, an EQ and an all pass? I know it sounds like I am an all pass junkie, but take a look at the filter Linkwitz shows in his how to section. You need a bunch of them.

Let's see what you come up with and I'm sure lots of others will chime in with their wish lists, too.

[kipman725]

I would advise against pots and instead use socceted resistors to reduce cost and increase perfomance. Also when one has finnished prototyping they could be soldered.

[CharlieLaub]

Originally Posted by kipman725

I would advise against pots and instead use socceted resistors to reduce cost and increase perfomance. Also when one has finnished prototyping they could be soldered.

I have some Marchand boards that use this approach and I don't like it. When I want to tweak the frequency, I have to come up with an entirely new set of 4 resistors, close in value, and then solder those on to a new header.

[CharlieLaub]

Hi Bob,

See my replies below.

Originally Posted by BobEllis

Hi Charlie,

I understand your desire for a prototyping board that can also be used for a finished product. However, the board real estate required for the pots and resistors to replace them would mean that you'll end up with a fairly large board. It might make more sense to make a prototyping board and a finished board. I'd like to see a board with SMD parts - even though it would be a challenge to assemble at my age and I have a boatload of through hole OPA2134s.

It's not really a prototype board. It's a finished board with all components on it. The ICs will be socketed so people who are in to that kind roll ICs until they are happy. Resistors that will be unchanged could be SMDs, but few people want to tackle soldering SMD devices so that might hinder bare PCB sales. I'm envisioning the pots to be vertical trimmer types. These don't take up much room, and the pin spacing should allow them to be removed and a resistor put in their place (in case someone wants that). I'm just trying to allow for flexibility, it's not a requirement to replace the pots.

Originally Posted by BobEllis

You should have some bandwidth limiting on the input. You may have thought of this already but didn't mention it. Allow for a decent size cap - you'll see why below.

Sure, both LP and HP filtering of the input would be good. I can probably work this in, but it just adds more components.

Originally Posted by BobEllis

You might want to breadboard a state variable filter and a Sallen-Key filter to see if you can hear a difference and like the state variable sound. I've heard that state variables don't sound as good, although there are plenty of folks who love the state variable Marchand XOs.

Jens' AF-4 implementation of the Sallen-Key gives adjustable Q while holding the gain to 1 using two opamps per 12 db/octave section. You could use pots for frequency and Q setting.

Have you simulated to see what happens when your pots don't track? Might be an issue, might not. You may be able to get some frequency spread to adjust phase adjusting them individually, but then the math gets more interesting.

I won't be using any dual pots. It's all done with trimmer pots, single gang, linear type. Tracking error is eliminated, but you have to separately adjust two pots to be the same resistance. I don't like dual pots of any kind, because of the tracking error issues.

Originally Posted by BobEllis

If you cannot get phase adjustment in the state variable filter, I strongly suggest that you provide and all pass section or two (another argument for SMD) You can use a pot for the prototype stage to set the shift. If you set up the baffle step input buffer with jumpers to allow the right connections you can use it as an all pass in a three way set up.

I thought about delay using all-pass first and second order sections. This is getting a little fancy, and I think that 95% of users would not really use this or know how to, but this could be built on another add-on board that people could purchase separately. You just can not do all of this on one PCB. It will be too large and too expensive. If the initial board is well received, I would move on to make other things such as the delay board.

Originally Posted by BobEllis

An option to make your settings adjustable would be to use sockets on the board to allow replacement of resistors or connection of pots, like you would probably use for your add on board. They can usually be snapped apart to allow single terminals to be placed on 10mm pitch.

Are you planning a single channel or stereo board? If single channel, 4 second order filters on the add on board seems overkill. How about 2 second orders, an EQ and an all pass? I know it sounds like I am an all pass junkie, but take a look at the filter Linkwitz shows in his how to section. You need a bunch of them.

As I mentioned, everything is adjusted via pots and these can be exchanged for resistors, but you then need to solder and unsolder them to change anything. You could just stick with the pots for future adjustability.

This is currently planned as mono boards. The idea is to use these in active speakers, so there would need to be one channel in each speaker. If you want a stereo pair, you will need to get two.

I don't think that four second order sections is overkill and it should work out well in terms of PCB size and IC requirements. For instance, if you want to increase a 2-way crossover to 4th order you need to cascade one 2nd order section with the HP, and one with the LP. That would leave two left over, for another 2nd order HP/LP and perhaps another BP filter for the equalizer section. If you want to make a 3-way, you need even more sections. Note that you can split up the 2nd order sections any way you want between filter sections and equalizer sections.

Originally Posted by BobEllis

Let's see what you come up with and I'm sure lots of others will chime in with their wish lists, too.

[CharlieLaub]

Originally Posted by BobEllis

Jens' AF-4 implementation of the Sallen-Key gives adjustable Q while holding the gain to 1 using two opamps per 12 db/octave section. You could use pots for frequency and Q setting.

That project is from 2005 or so. It seems that most of the info is now dead links. That was one of the reasons that I wanted to put this forward.

He may use 2 op-amps per SK filter, but with 3 per SVF I can get independent adjustability of all parameters, and better range of Q, increased circuit stability, etc. Also, with SK, you can only execute one function (e.g. HP or LP) at a time, so you can get mismatch between the two depending on component errors. The SVF executes LP and HP simultaneously, so these will always be matched, and the only error will be in the frequency or Q of BOTH.

-Charlie

[BobEllis]

Originally Posted by CharlieLaub

That project is from 2005 or so. It seems that most of the info is now dead links. That was one of the reasons that I wanted to put this forward.

He may use 2 op-amps per SK filter, but with 3 per SVF I can get independent adjustability of all parameters, and better range of Q, increased circuit stability, etc. Also, with SK, you can only execute one function (e.g. HP or LP) at a time, so you can get mismatch between the two depending on component errors. The SVF executes LP and HP simultaneously, so these will always be matched, and the only error will be in the frequency or Q of BOTH.

-Charlie

By prototype board I meant one for dialing in the crossover.

That's one of the advantages of SK (and MFB or any separate implementation of high and low pass filters) - you can intentionally mismatch frequencies, slopes and Q to deal with driver response irregularities and roll off as well as do some or all of the needed phase adjustment without an all pass filter. I guess it amounts to design philosophy.

As for the all pass, users can always jumper it out if not being used (or as in Jens' design not jumper it in). After trying the same crossover with and without phase correction, I won't be doing without again. Even a couple where I just ballparked the offset before measuring worked out better than without. I wholeheartedly agree with Linkwitz' statement the without phase correction an active crossover is marginally useful.

Input LP can be as simple as an RC. Linkwitz uses a 3 pole filter (RCRCRC) but that's probably overkill.

Another useful feature would be a level adjust on the high pass. You should buffer the output, whether opamp or discrete follower. If you use a dual opamp, you'll have a spare for an all pass.

[CharlieLaub]

Originally Posted by BobEllis

By prototype board I meant one for dialing in the crossover.

That's one of the advantages of SK (and MFB or any separate implementation of high and low pass filters) - you can intentionally mismatch frequencies, slopes and Q to deal with driver response irregularities and roll off as well as do some or all of the needed phase adjustment without an all pass filter. I guess it amounts to design philosophy.

OK, I see now. Yes, that is definitely a valid approach - sort of a hangover from passive crossover design IMHO. I prefer matched crossovers, and then flattening frequency response first by driver level adjustments, then by active baffle step compensation followed by one or more bands of equalization.

Mismatching frequencies also will mismatch the phase in the crossover region, and this can lead to response irregularities, although they may not be significant or audible.

Originally Posted by BobEllis

As for the all pass, users can always jumper it out if not being used (or as in Jens' design not jumper it in). After trying the same crossover with and without phase correction, I won't be doing without again. Even a couple where I just ballparked the offset before measuring worked out better than without. I wholeheartedly agree with Linkwitz' statement the without phase correction an active crossover is marginally useful.

True. The only problem I find is that you need to know how much offset you are compensating for. A single first order all pass can only do so much in terms of delay and one or more second order stages would be better. Multiple delay stages could be implemented in an adjustable way, but I think that would be better left for another board.

I don't really agree with SL WRT offset compensation. Look how many speakers with passive crossovers are out there that have nothing like that. Offset compensation via tweeter delay is something that has the potential to get you another 5% of performance, but only if there is a problem to solve in the first place. Odd frequency response irregularities in the crossover region (if they happen to be there at all) on account of the offset, can be remedied, but I am not sure that this is as much of a problem as SL suggests.

Originally Posted by BobEllis

Input LP can be as simple as an RC. Linkwitz uses a 3 pole filter (RCRCRC) but that's probably overkill.

Another useful feature would be a level adjust on the high pass. You should buffer the output, whether opamp or discrete follower. If you use a dual opamp, you'll have a spare for an all pass.

Yes, the input LP is definitely a good idea. Also a HP to reject radio frequency interference is equally good. First order for these is probably fine.

I'd plan to have an overall level (volume) control as part of the crossover. On the other hand, I prefer that the adjustment of the level of each driver be done at the amp, and this is pretty easy to do using a voltage divider pot or the like.

-Charlie

[BobEllis]

Just because you don't see a separate passive all pass section doesn't mean the designer hasn't adjusted the phase to a minimum error. When you see a nice deep reverse null it means the phases are pretty darned close. That rarely happens with a symmetrical filter.

You made my case for multiple all pass sections. The first time I added phase correction my son noticed without being asked. It was a fairly dramatic change. I'm a believer. YMMV.

[CharlieLaub]

I've manged to fit the circuit on a 3 x 3.5 inch PCB. Below is a screenshot of a 3-D rendering of the board, showing some component outlines.

Each board is a single channel and has:

* An expandable 2-way 2nd order crossover having adjustable frequency and Q
* A baffle step compensation circuit with adjustable step dB and center frequency
* An expandable fully parametric EQ with one on-board band

Once this board is in production, I plan to design and produce an expansion board that will allow the user to add additional 2nd order filter blocks and/or EQ bands.

Hopefully this board will be ready early next year. This will be offered as a kit, including PCB and parts. U-build it. The caps used are 1% Vishay and 5% Wima types, and resistors are 1% MF type.

-Charlie