I have a larger project for a multi-amp 4 way system and I want to design the best 4 way active crossover filter possible, and keep it buildable as a DIY project, so I want to avoid any SMT type of stuff. Although careful choice of opamps can allow 2 versions of PCBs made, one through-hole and the other could be SMT.
I have all my speakers made and I plan to have a kind of short rack on wheels to house all the power amps, the crossover and all the electronics needed, including a micro-controller based management of the whole thing to automate it so no human intervention is needed on the rack itself, not even to turn it off.
The type of filter chosen is Linkwitz-Riley 24db/oct, making use of all-pass filters to correct the phase differences.
The tentative crossover frequencies chosen (for now) are:
150hz
1.5khz
8khz
Certain things remain to be debated, as far as eventual additional adjustable delays on some bands to compensate for physical alignments of speakers. My speakers are adjustable, as each of the 4 ways is in its own cabinet and they are on top of each other, thus can be physically adjusted, to a point. But perhaps having an extra adjustable delay to correct for some of the physical alignment that can't practically be made, could be useful.
The input needs to be balanced, to allow for long signal cables, while the amp racks would be positioned very close to the speakers, to minimize the speaker cable lengths. I expect the longest power cable length not to exceed about 2m, probably even less, at least for the tweeters, that are way on top of everything else. The bass power cables might not even be much longer than about 1m.
Sorry for the us based diyers, as I work only in metric
Attaching a rough synoptic of the crossover.
I have all my speakers made and I plan to have a kind of short rack on wheels to house all the power amps, the crossover and all the electronics needed, including a micro-controller based management of the whole thing to automate it so no human intervention is needed on the rack itself, not even to turn it off.
The type of filter chosen is Linkwitz-Riley 24db/oct, making use of all-pass filters to correct the phase differences.
The tentative crossover frequencies chosen (for now) are:
150hz
1.5khz
8khz
Certain things remain to be debated, as far as eventual additional adjustable delays on some bands to compensate for physical alignments of speakers. My speakers are adjustable, as each of the 4 ways is in its own cabinet and they are on top of each other, thus can be physically adjusted, to a point. But perhaps having an extra adjustable delay to correct for some of the physical alignment that can't practically be made, could be useful.
The input needs to be balanced, to allow for long signal cables, while the amp racks would be positioned very close to the speakers, to minimize the speaker cable lengths. I expect the longest power cable length not to exceed about 2m, probably even less, at least for the tweeters, that are way on top of everything else. The bass power cables might not even be much longer than about 1m.
Sorry for the us based diyers, as I work only in metric
Attaching a rough synoptic of the crossover.
Hello, I assume you are familar with Rod Elliotts 4-way active?
Project 125
- good for inspiration or comparison with your own design in any case.
I am actually planning to order a few boards from him, and see how it works out. Would probably use different opamps though.
Project 125
- good for inspiration or comparison with your own design in any case.
I am actually planning to order a few boards from him, and see how it works out. Would probably use different opamps though.
Hi,
Yes, of course, and many others. Some are a good starting point for my goal, but will require changes and appropriate calculations.
His project as described is somewhat close to what I'm aiming for, but not quite.
For one thing, I do not want any potentiometers in the signal's path. If a level adjustment must be made, I would rather do it with something like a VCA or whatever solid state means, but no contact or anything electromechanical.
I do want to have a limiter on the outputs, so I am thinking about using them as level adjustments as well.
I also want to avoid all possible occurences of capacitors in the signal's path. We always have far too many, and since my signal will come from a device that would have an output capacitor for sure, I want to avoid adding any more along the way.
The input symetric amp will also have to serve as a signal detector so I can feed that input into a tiny microcontroller. When the whole system is powered down, only the input stage will be powered to be able to detect an incoming signal, so the microcontroller would then use that info to properly power up everything in the system, progressively, amp by amp, commencing by powering up the xover filter that follows that input stage, which itself would have its output muted until everything that follows it is confirmed to be powered up properly, with no faults, and then things would get un-muted all the way to the speaker outputs, in proper order.
There is one thing that I have reserves on about rod's project: the phase of each outputs in regards to each other. That design, along with many others, will definitely serve as guides.
I am also reading Doug Self's book "The Design Of Active Crossovers", but that'll take some time. I have gathered a good amount of stuff about this subject, and I think I can arrive at something decent that will suit the needs.
Member
Joined 2009
Paid Member
With regards through-hole vs SMT perhaps I can share my experience.
Small multi-legged SMT parts are really nasty to work with. If you can get them down neatly first time that's good but if you need to repair or change anything it's really tough unless you have all the gear.
However, the layout flexibility and space saving from using SMT is really useful and often you have a better choice of parts available. I found that two-terminal SMT devices are very easy to work with so resistors and capacitors and diodes can be all SMT without any hassle. If you use fairly large SMT devices with larger pin spacing I think you can easily use more complex parts too. Avoid SMT capacitors, especially electrolytics as they don't lend theselves easily to hand-soldering.
I wish you luck with your project - it's more ambitious than I would dare to tackle !
Small multi-legged SMT parts are really nasty to work with. If you can get them down neatly first time that's good but if you need to repair or change anything it's really tough unless you have all the gear.
However, the layout flexibility and space saving from using SMT is really useful and often you have a better choice of parts available. I found that two-terminal SMT devices are very easy to work with so resistors and capacitors and diodes can be all SMT without any hassle. If you use fairly large SMT devices with larger pin spacing I think you can easily use more complex parts too. Avoid SMT capacitors, especially electrolytics as they don't lend theselves easily to hand-soldering.
I wish you luck with your project - it's more ambitious than I would dare to tackle !
I personally will stay away from SMT, because I am no longer able to handle such things. My eyesight has gone to hell in the past few years, and my dexterity is far from what it used to be. I love building stuff and I do want to continue, but there is just no way I can tackle the SMT. However, SMT does have lots of advantages, if someone else would do the building for me, I wouldn't mind the result, but then I would not be the one doing the building and that would be a big minus.
For someone young and healthy, doing SMT might not be too bad. I just wish I was 20-30 years younger, so I could do it.
Still I will consider doing the PCB design for both versions, just in case.
My project goes back many years ago, I was already working on it in the early 80s. I'm just behind schedule, but eventually it'll get there. Like I said, I already the big stuff, the speakers, so I won't let them rot somewhere for ever, unused. So I want to build this stuff whenever I can get it done, and finally enjoy the results.
I know it won't be easy and it'll take time. I have quite a few projects in line and I'm not able to handle them all at once, so I'm taking this slowly and one at a time.
This project has a chance to get somewhere somewhat soon, and I should be able to do some work on it over time, here and there. It's only one part of a larger project, but this would be a good step towards the ultimate goal.
I have already done quite a bit of work on the mechanical side of it, with that wheeled short rack being mostly drawn in MCAD. I have some of the work done on the power amps, and ideas on those to continue.
It is a somewhat complex project overall, nothing unsurmountable.
I won't bother designing and building an equalizer, since there are decent ones on the market and they're cheap enough to not warrant the DIY effort, however I plan to make a pink noise generator and a spectrum analyzer, so this whole system can be properly tuned in place, wherever it is put to work.
This is not for living room hifi, it's for something much bigger. My speaker system (4 ways), has sizeable cabinets and the woofers are 18", and even the low mid range ones are 15".
I know that short rack will get heavy, with 4 power amps and the big toroidal transformers that go with them, the xover won't be that big compared to the rest of it inside, especially the cooling tunnels with regulated forced ventilation. With all that automation with microcontrollers so there would be hardly any knobs to adjust, that does make it a rather ambitious project.
But for the moment, I'll stick to the 1st stage, the xover, as the thread's title states...
For someone young and healthy, doing SMT might not be too bad. I just wish I was 20-30 years younger, so I could do it.
Still I will consider doing the PCB design for both versions, just in case.
My project goes back many years ago, I was already working on it in the early 80s. I'm just behind schedule, but eventually it'll get there. Like I said, I already the big stuff, the speakers, so I won't let them rot somewhere for ever, unused. So I want to build this stuff whenever I can get it done, and finally enjoy the results.
I know it won't be easy and it'll take time. I have quite a few projects in line and I'm not able to handle them all at once, so I'm taking this slowly and one at a time.
This project has a chance to get somewhere somewhat soon, and I should be able to do some work on it over time, here and there. It's only one part of a larger project, but this would be a good step towards the ultimate goal.
I have already done quite a bit of work on the mechanical side of it, with that wheeled short rack being mostly drawn in MCAD. I have some of the work done on the power amps, and ideas on those to continue.
It is a somewhat complex project overall, nothing unsurmountable.
I won't bother designing and building an equalizer, since there are decent ones on the market and they're cheap enough to not warrant the DIY effort, however I plan to make a pink noise generator and a spectrum analyzer, so this whole system can be properly tuned in place, wherever it is put to work.
This is not for living room hifi, it's for something much bigger. My speaker system (4 ways), has sizeable cabinets and the woofers are 18", and even the low mid range ones are 15".
I know that short rack will get heavy, with 4 power amps and the big toroidal transformers that go with them, the xover won't be that big compared to the rest of it inside, especially the cooling tunnels with regulated forced ventilation. With all that automation with microcontrollers so there would be hardly any knobs to adjust, that does make it a rather ambitious project.
But for the moment, I'll stick to the 1st stage, the xover, as the thread's title states...
Dear Spookydd,
Your PDF mentions 24dB/Octave LR filtering. Not to spoil the party, but it has been said a million times before, but here it comes once again: it is the acoustic output that should be LR 4, not the electric transfer function.
In the analogue domain it is quite a task to match analogue active electronics filters with individual loudspeaker's SPL to produce the desired acoustic results within a couple of dB's .
Douglas Self will not come to your aid: not a single design Self presents in 600+ pages takes individual woofer, mid and tweeter SPL curves into consideration. Same for Rod Elliot: forget about all that if you are after a serious design. The Linkwitz analogue filter circuitry will give you some insight in the complexity of such a venture in the analogue active world.
It is for very good reasons therefore more and more designers are moving to DSP with its unparalleled flexibility and accuracy: any curve/filter slope is designed with a handful of mouse-clicks.
Good luck,
Eelco
Your PDF mentions 24dB/Octave LR filtering. Not to spoil the party, but it has been said a million times before, but here it comes once again: it is the acoustic output that should be LR 4, not the electric transfer function.
In the analogue domain it is quite a task to match analogue active electronics filters with individual loudspeaker's SPL to produce the desired acoustic results within a couple of dB's .
Douglas Self will not come to your aid: not a single design Self presents in 600+ pages takes individual woofer, mid and tweeter SPL curves into consideration. Same for Rod Elliot: forget about all that if you are after a serious design. The Linkwitz analogue filter circuitry will give you some insight in the complexity of such a venture in the analogue active world.
It is for very good reasons therefore more and more designers are moving to DSP with its unparalleled flexibility and accuracy: any curve/filter slope is designed with a handful of mouse-clicks.
Good luck,
Eelco
I see your point, but I don't want to do DSP, as mentioned. So I started this thread with only analog design in mind.
Of course it's a complex thing and the main issue about matching the filters with the speakers is a big issue, but there is only so much we can tackle and we can give ourselves some options to compensate for certain things, like the phase and spl levels.
My speakers already exist and I will attempt to make the best match humanly possible. Not easy, with some tech data missing and no actual measurements.
For info, my spl on each speaker should be something like this:
low: 103 (18" speaker alone, and don't know the real one as the cabinet used is a rear expo horn load)
mid-low: 105 (15" spkr)
mid-hi: 110 (1" driver with horn)
hi: 105 (jbl2405)
The only thing I can think of to allow for matching to the spl is a level adjustment on each output of the filter, but I don't want to have potentiometers in the signal's path, so this will have to be done with a VCA and I'm thinking of making double use of the limiter for that.
One thing is certain, the goal for this project is analog and not DSP.
Of course it's a complex thing and the main issue about matching the filters with the speakers is a big issue, but there is only so much we can tackle and we can give ourselves some options to compensate for certain things, like the phase and spl levels.
My speakers already exist and I will attempt to make the best match humanly possible. Not easy, with some tech data missing and no actual measurements.
For info, my spl on each speaker should be something like this:
low: 103 (18" speaker alone, and don't know the real one as the cabinet used is a rear expo horn load)
mid-low: 105 (15" spkr)
mid-hi: 110 (1" driver with horn)
hi: 105 (jbl2405)
The only thing I can think of to allow for matching to the spl is a level adjustment on each output of the filter, but I don't want to have potentiometers in the signal's path, so this will have to be done with a VCA and I'm thinking of making double use of the limiter for that.
One thing is certain, the goal for this project is analog and not DSP.
Hello Spookydd,
I fully understand your points. For starters you must have in box measured SPL curves of the unfiltered individual drivers. These must then be combined with the filters, in order to produce the desired target slopes. Get both measuring and simulation software.
But maybe you already have been through all that. Look here for showing the way:
http://www.linearx.com/files/pdf/FilterShopApp_05.pdf
Eelco
I fully understand your points. For starters you must have in box measured SPL curves of the unfiltered individual drivers. These must then be combined with the filters, in order to produce the desired target slopes. Get both measuring and simulation software.
But maybe you already have been through all that. Look here for showing the way:
http://www.linearx.com/files/pdf/FilterShopApp_05.pdf
Eelco
Well I don't have any way of measuring anything, so although it would be great, it's not in the card at present.
However, I do have among the projects a pink noise generator and a spectrum analyzer, which would go a long way towards the measuring and adjustment possibilities goals.
What I have doesn't look even close to such setups. As I mentioned, this is not aimed at living room hifi, so my speakers are nothing like described there.
Each of my speakers are individual, so it is quite possible to physically adjust their position to compensate for phase. However, as I also mentioned, I think adding an extra adjustable delay/phase correction on each output would add to the possibilities of adjustment, in addition to the physical positioning of the speakers, in case the needed physical positioning isn't practical, both the delay feature in the filters and the physical positioning should allow for proper adjustment, as long as we can measure the response as needed.
This is why I'm also considering making a spectrum analyzer and the pink noise generator.
My sound system is not like home hifi. The speakers are a little big and they will always be positioned at least 8 or 10 meters away from the sound source, so this mandates using balance lines and also why I want the rack of amps and filters automated and placed as close as possible to each speaker stack.
For now, I am using overly long power cables and only a single power amp which is near the mix table and not near the speakers. Then I have 3 way passive filters, which prevents the use of the low-mid speakers.
It works but I want to make it better, as I planned it to begin with, long ago.
I found a software based spectrum analyzer that although is likely not as accurate as professional measuring equipment, should be sufficient to compare levels in the whole audio spectrum, and should allow tuning a system with its amps, speakers and xovers.
Using a computer software based analyzer instead of having to make an electronic device will surely make things much easier and saves a ton of money and time.
Making a pink noise generator is so simple and cheap, this won't be much to do.
So giving the xover the few features needed to make adjustments, plus whatever adjustments the speakers physically allow, should make it possible to tune a system in place without having to go overboard and spend a fortune.
Using a computer software based analyzer instead of having to make an electronic device will surely make things much easier and saves a ton of money and time.
Making a pink noise generator is so simple and cheap, this won't be much to do.
So giving the xover the few features needed to make adjustments, plus whatever adjustments the speakers physically allow, should make it possible to tune a system in place without having to go overboard and spend a fortune.
I am considering using the THAT 4301 IC for the output limiter.
It's a bit pricey but what are the alternatives?
Anyone? Suggestions?
Being a 4 way filter, putting a limiter on each output can quickly add up at such prices, but I don't know of many possibilities. This IC combines the VCA and the RMS Detector, not to mention the 3 opamps, so this is a single chip solution to make the limiter, and I hope it can also be used for level adjustment, since it has the VCA and an external adjustment should be feasible (digitally if at all possible, not using any potentiometer in the sound path).
It's a bit pricey but what are the alternatives?
Anyone? Suggestions?
Being a 4 way filter, putting a limiter on each output can quickly add up at such prices, but I don't know of many possibilities. This IC combines the VCA and the RMS Detector, not to mention the 3 opamps, so this is a single chip solution to make the limiter, and I hope it can also be used for level adjustment, since it has the VCA and an external adjustment should be feasible (digitally if at all possible, not using any potentiometer in the sound path).
Seems like quite an ambitious and interesting crossover project!
I don't know much about the THAT ICs, but regarding spectrum analyzer and speaker measurements, I can recommend you take a look at ARTA. I have just started using it, it gives you a lot of possibilities for little money. With a good external sound card (interface), you will actually have a quite decent FFT and spectrum analyzer, and you can also measure Thiele parameters.
About capacitors: Yes way too many is not good, but the way I look at it, better one extra, than not enough. Many have run into bigger problems when trying (on sheer princple) to eliminate all caps in a circuit.
About potentiometer for level adjustments: Perhaps not ideal, but they can always be replaced by fixed resistors, once you have found the correct levels. Simple and effective solution I think.
I don't know much about the THAT ICs, but regarding spectrum analyzer and speaker measurements, I can recommend you take a look at ARTA. I have just started using it, it gives you a lot of possibilities for little money. With a good external sound card (interface), you will actually have a quite decent FFT and spectrum analyzer, and you can also measure Thiele parameters.
About capacitors: Yes way too many is not good, but the way I look at it, better one extra, than not enough. Many have run into bigger problems when trying (on sheer princple) to eliminate all caps in a circuit.
About potentiometer for level adjustments: Perhaps not ideal, but they can always be replaced by fixed resistors, once you have found the correct levels. Simple and effective solution I think.
If you are serious about this then you absolutely need to take note of what Boden has said, as you will not succeed with your goal if you do not
I would also comment that best is very difficult to quantify, as there are so many tradeoffs, in the end it is what is best for you based on what tradeoffs you can live with and your setup. If you ignore the idea of tailoring the electrical filters to achieve the desired acoustic slope then I do not believe you will (short of amazing good fortune, or drivers that are flat way beyond the crossover frequencies) get the best possible results.
Tony.
Well, I just went to take a peek, and from what I see, it's a dead end for me, because it's a windoze only thing and I don't do windoze.
There is a right balance and care can be taken to only have a strict minimum number of them only where they are truly needed.
I've seen designs where they put a potentiometer for level adjustment and put a cap before and after it, and they have more caps in other places, in the signal's path. Far too many.
In this design, I want a balanced input and the signal would come straight from an equalizer, which would have already taken care of blocking dc with likely too many caps as well, so no need to add more, and then if need be, only one could be placed at the output of the last stage and right before the power amp...
I've seen what caps do to the signal in simulations, and it's not pretty most of the time. We put a lot of efforts into making circuits with very low distortion and noise, and we trash this by using way too many caps. What's the point?
Maybe, but since I will have a limiter on the output, which has a VCA, then the VCA can be used also for the level control, and that level control via the VCA can be done using a digitally made dc signal from a system with up/down buttons with small increments to give a fine tuning ability. I'd say at the minimum an 8bit resolution, perhaps even more.
This level adjustment would only be set at the time when the system is being tuned with the pink noise and analyzer, once the speakers have been physically positioned and a little more fine tuning can then be done with the level adjustment. So there would be no "knob" to touch once it's all set up properly.
Since I plan to make use of microcontrollers in this system, why not have a tiny one for that purpose? Not too difficult.
Let's take a hypothetical mid driver that is required to pass 300Hz to 1200Hz, just two octaves.
If it has a smooth flat response from 2 octaves below 300Hz to 2octaves above 1200Hz, i.e. from 75hz to 4800Hz, then you can apply a 24dB/octave passband filter that only has errors in the roll-off regions that are below -30dB. The phase and level errors out at 2octaves will be virtually inaudible.
But where can you find such a 6octave driver to give you a 2octave passband?
That's the problem. You would need to find four 6octave drivers to cover just 8octaves of audio. You are still missing 2octaves (you need 10octaves). That means two of the drivers need a flat response over 7octaves and the other two drivers can manage with only 6octaves of flat response.
It's the near impossibility of hitting that target of 4drivers with near perfect response over 6 to 7octaves that requires your active crossover to correct non flat responses in your passbands.
Find the drivers that best meet your target, Then find what corrections are required to give the correct acoustic roll offs.
Only then can you finalise your 4way crossover.
If it has a smooth flat response from 2 octaves below 300Hz to 2octaves above 1200Hz, i.e. from 75hz to 4800Hz, then you can apply a 24dB/octave passband filter that only has errors in the roll-off regions that are below -30dB. The phase and level errors out at 2octaves will be virtually inaudible.
But where can you find such a 6octave driver to give you a 2octave passband?
That's the problem. You would need to find four 6octave drivers to cover just 8octaves of audio. You are still missing 2octaves (you need 10octaves). That means two of the drivers need a flat response over 7octaves and the other two drivers can manage with only 6octaves of flat response.
It's the near impossibility of hitting that target of 4drivers with near perfect response over 6 to 7octaves that requires your active crossover to correct non flat responses in your passbands.
Find the drivers that best meet your target, Then find what corrections are required to give the correct acoustic roll offs.
Only then can you finalise your 4way crossover.
Capacitors used to couple the signal when selected correctly do not add any audible distortion to the audio signal.
Capacitors used to filter the signal cannot be avoided, since using inductors is worse.
Capacitors used for filters must be selected to best perform that function.
Some will need to be polypropylene, or better. But some can be other lower quality plastics and still perform adequately. Read D.Self.
Capacitors used to filter the signal cannot be avoided, since using inductors is worse.
Capacitors used for filters must be selected to best perform that function.
Some will need to be polypropylene, or better. But some can be other lower quality plastics and still perform adequately. Read D.Self.
Years ago, I built the 3-way crossover given in Linkwitz's article:
SB1980-3way
I found that the project wasn't difficult, and the article includes everything you need to change it to be 4-way and to calculate the xover frequencies. Certainly, substitute modern op amps for what was available to Linkwitz back in the '70s. When I built this, TL07x were available, and worked just fine.
SB1980-3way
I found that the project wasn't difficult, and the article includes everything you need to change it to be 4-way and to calculate the xover frequencies. Certainly, substitute modern op amps for what was available to Linkwitz back in the '70s. When I built this, TL07x were available, and worked just fine.