REW is free software for doing acoustic measurements. It's an invaluable tool when designing speakers.
REW - Room EQ Wizard Room Acoustics Software
The calibration curve is to compensate for the unlinearities of the microphone, so you get a linear measurement.
You need a PC and an amplifier of some sort. You're designing a high sensitivity system, so if you don't have anything, I'd reccomend you to get a one or two 4-channel UcD34 modules from Hypex. They're on end of product sale right now at €65, and that's a steal. The module is a 4x30W amp with SMPS power supply. You need to add a regulator, buffer and enclosure.
REW allows you to measure with gating, so that it basically cuts off the measurement early enough that room reflections aren't included.
See here for an explanation: https://www.minidsp.com/applications/acoustic-measurements/loudspeaker-measurements
You need to know what you are designing for, before you start designing, no matter what filter technology you're using. DSP filters are the bees knees for this, since they allow you to dial in crossover on a finished box.
If you don't know the response curves of anything, and start designing a crossover for wild guesses, you are in essence throwing parts and efforts at the problem until you hit reasonably close to a solution.
The Crossover frequency, phase and level controls are by no means the only thing you do in a crossover.
You do baffle step compensation, spl curve correction (EQ) and voicing. Without measurements, you are still shooting blind.
You don't need to get the MiniDSP 4x10HD.
If you are using it for only the design phase, a MiniDSP 2x4 is sufficient, and that one costs $105. The kit form is $80. You will need a protection capacitor for the compression driver. That should also be used on the finished speaker.
If you use it for measurement only, one of them is enough for measuring a 4-way construction, but if you want to listen to a finished system, you'll need two of them for doing more than a 2-way speaker.
You are spending a serious amount of time and resources on doing a 4-way system here. You need to spend a little resources on doing the right stuff, otherwise, you'll spend way more time and money on barking up the wrong tree.
I realize that the goal of a hobby is to convert spare time (and some money) into enjoyed time, but it's a bonus if you get something useful out of it.
Johan-Kr
Last edited:
Look again at your mid and bass drivers.
The mid goes down to 50Hz and has a 105dB sensitivity.
The bass goes down to 57Hz and has a sensitivity of 103dB.
I don't see any point in augmenting a mid with a bass driver that cannot go as low, nor go as loud.
The treble and the mid cover most of the audio range. That gives you a 2way, not a 4way.
The mid goes down to 50Hz and has a 105dB sensitivity.
The bass goes down to 57Hz and has a sensitivity of 103dB.
I don't see any point in augmenting a mid with a bass driver that cannot go as low, nor go as loud.
The treble and the mid cover most of the audio range. That gives you a 2way, not a 4way.
Cool! I'll have to take a close look at this.
Well, for me it'll have to be mac, no windoze.
The power amps that I have are getting too old and haven't been used for a long time, so when I get ready to do something of that nature, at least one amp will require a complete overhaul, changing all its caps...
With the costs involved, it'll be a while before it can be attempted.
It'll be cheaper and more useful for me to overhaul the old power amps. I don't have deep pockets, so I'll make use of what I already have.
Intriguing those modules, with switching power supplies, it makes them much more compact by avoiding the use of transformers.
Wow! Very cool!
I didn't know about that usb mike, and this is better than what I had in mind before for the system tuning. Along with the software based stuff, we don't need expensive spectrum analyzers any more.
The plan was to keep it sufficiently adaptable and do the tuning afterwards.
I was even thinking I'd have to get a spectrum analyzer, or at least build something, but now it should be easier with the computer based stuff.
Good thing, because it's way too expensive.
I had no idea this stuff could be in kit. That's better, at least it can save a little dough.
Why is that? The speaker will be directly hooked to its power amp later, and a cap would be extra stuff that's undesirable, just like passive filters.
I agree with that. There is a lot more involved than just a plain filter.
The hobby for me isn't just to design and build this, I also want to use it, so it's important that it does what it's meant to do and do it the best possible. Evidently the limited funds are a problem, and that make things more difficult and it can take more time.
I started on this project in the 80s, and of course it evolved a bit over time. Until now that 4 way set of speakers was configured with a 3 way passive filter (yikes!) and the low-mid was not used. It was just powered by a single stereo power amp, and I never had an equalizer either, so it went straight from the mixer to the power amp and the overly long power cables to the speakers. Not ideal.
To illustrate the point, the graph below (ignore the blue curve) is the transfer function of my active crossover. This when combined with the actual drivers responses results in a 4th order accoustic bessel rollof. The filters are based on 3rd order electrical but as you can see they are far from standard 3rd order slopes (at least the low pass). If I used standard 3rd order filters, or 4th order for that matter, I would not get anywhere near the response that I do with these built for purpose filters.
Tony.
Attachments
That can't be helped. But the whole bandwidth is very well covered, and the overlap is significant between all drivers.
The goal is to have each part of the spectrum reproduced by the most suitable driver.
The big driver (18in) may not go as low as the smaller 15in one, but I want to use it for the lowest part, and it's rear loaded with an expo horn, so although the spl of the naked speaker may be "only" 103, the horn surely changes that, including its response from the driver and it's cabinet combination. It's really geared for the low end, and it does it very well too.
For that low-mid 15" one, it may go down quite a bit, and it does make for a mean woofer, highly efficient, but it also is super clean and does the voices very well, crisp sound, I chose it on purpose, and there aren't any other like it on the market right now.
Although the 1" and horn driver above may also be capable of reaching 20khz, I want to let the slot tweeter do the top of the band instead. Those tweeters are so crisp and clean, it's a delight to hear them, and they reach higher than 20khz too.
Most of the energy being in the lower part of the spectrum, the power amps for the 2 upper drivers will be much less beefy.
The 1" driver can handle 70W above 800hz and 100W above 1k2, but I will cut it probably at 1k5, perhaps even a tiny bit higher, and it won't be doing the whole spectrum all the way up, that's the tweeter's job. And so the tweeter won't have to handle that much power either. It's capable of 40W above 6k. They recommend cutting at 7k or higher, and I was planning to cut at 8k.
Since so little power will be required for that tweeter, I'm thinking that a class-A amp might be a good idea. Those tweeters can reproduce the slightest detail.
It's true that only 2 of those drivers could very well be used together in a 2way, like for example the 1" and horn with the 15", would cover all of it easily, but I want to improve on that and use what I have.
They are built, it's not just a plan any more, and although they're been underutilized in the past, they will be much better utilized once the electronics is all built.
I have the mixer, and I always planned to just buy an equalizer instead of wasting time and money making one, because in that case, it's not worth it financially.
The equalizer will have to have symetric/balanced outputs, because it will be located always in the same place with the mixer, and the power amps and xover rack will be stuck as close as possible to the butt of the speakers, with the shortest power cables possible. The balanced line from the equalizer going to those racks might reach more than 20m in length.
And the power racks being remote from the mixer and equalizer, I don't want to have any kind of knobs or adjustments on those racks besides the ones to set from tuning and measurements, plus they must power themselves up and shut down automatically from the presence of modulation coming in. If no signal it sent out to the racks for perhaps some 10mins, they will properly shut themselves down and be put into standby mode, with only power to the microcontroller and the signal sensing stage remaining active. I probably will have the microcontroller and signal sensing stuff battery powered, so it goes totally off the grid when in standby.
REW is available for the Mac, too.
As you have several amps, one channel will be suficient for the measuring.
A UcD34 or two will probably be cheaper than recapping an amp. As you enjoy doing your own gear, you can make the input buffers yourself...
They will probably be a bit underpowered for the woofers, but with a stereo amp for the woofers, you can bridge two channels on each UcD34 to run the midwoofers, and then have one 30W channel for the tweeter and one for the mid.
Adaptability is where the DSP is king of the hill...
The protective cap is needed for active setups. In passive setups, the drivers are protected by the crossover.
It is a large cap (47-70uF or so) in series with the tweeter intended to protect the (expensive) compression driver from DC and overexcursion, in case of a amplifier or filter malfunction or mistake. It doesn't take much in the way of DC or low frequencies (say, if you by mistake send it a frequency sweep of 20Hz-20kHz, instead of 500-20k) to kill a compression tweeter. They can usually be repaired, but it's expensive...
Johan-Kr
Yes! Hurray! For once, the mac users were not left behind!
I found it and got it, but until I have a mike, I won't get anything out of it, but at least I can start getting familiar.
Thanks for the hints.
Yes, and by far, cheaper than buying more gear.
There are a bunch of caps to changes all over an amp, but they really aren't that numerous and it shouldn't break the bank to refresh those oldies. (from the early 80s mostly)
Well, since I will have to overhaul those power amps anyway, I won't let them go to waste. They are still good, just old, and they may even work if I power them up, but I won't. Not until they get all new caps. It's too risky, with more than 20 years of non-use. I just won't trust those caps, especially the big filters.
I have a 140W and a 150W (per channel), and both bridgeable, so they are plenty and they need to be brought back to service in any case. I already had planned to do this to several "old devices", not just power amps, like old computers for example. Because I also like retro computers and I plan to refresh them to bring them back to life. I have an Apple //e and a couple of ZX81, plus an Atari 520ST that all need their caps changed. All that stuff will get new caps when I get to them.
True. But I always wanted my analog filters. Maybe when I get this working, I'll look at the DSP stuff... But nowadays most of those things are better designed with FPGAs or similar, so that's totally not hobby stuff, with all that dense SMT stuff, there is no way I can tackle those things, at least not building them, but maybe get into design. I am looking into learning FPGAs, so maybe some day I might do something like that, but knowing I won't be the one actually building. By then my eyesight will have gone down further, and so far, no glasses have been worth anything...
Oh! So you're referring to protecting the drivers from the amps if it sends DC, or from clipping or some other overdrive conditions.
The amps themselves will have to be built to protect the speakers, by any means necessary, so the bare minimum is between the speaker and the amp, including the shortest possible cable.
A cap in series with a speaker is a high pass, and in any case it can only bring distortion in the fold and lower sound quality. The point is to avoid all passive stuff. I am always looking for the best possible damping factor, especially for the bass, so short cables, and try to avoid anything that can insert any bad contact over time.
True, and probably a good idea for testing purposes, but after that, they're extra fluff in the way. The xover should have limiters on its output, before the amps, and the amps should have DC protections. All my amps will have to be designed with solid state output muting relays, DC sensing, and even the rails will have solid state relays.
My speakers do cost a fair amount, and the amps must have all the conceivable protections to prevent speaker damage.
What you can do to power up the amps carefully, is to make yourself a short cable with a power connector on one end, a power outlet on the other, and a lamp with a 100W glow bulb (if you can get one, that is) in series with the live phase. The lamp will light up fully when the amp turns on, and go dim as the power supply caps is charged. If something is wrong, the lamp will burn out or stay powered up, but limit the current, so you can turn off the amp if it doesn't work as it should.
Johan-Kr
Johan-Kr
You may be right, but it is risky to ignore those aging caps.
One of my amps is older than the other and I think it may be from the late 70s. That's quite old, and in those days the caps were less reliable over time than they are now.
Plus they haven't been used for a very long time. I think it's probably closer to 30 years than 20 of non use.
It's not really that bad to change a set of caps, and much cheaper than buying more gear.
Those old amps are sturdy and never failed, and they can continue being useful.
One of them, the oldest one, got a few years of very bad storage, when it was sitting on its rear feet on the floor in a basement. That basement got very humid in those years and a layer of moist air rusted it up to about half way. It's not too badly damaged, but this was not good. Nothing to do with the caps, but it needs TLC now.
I tried looking at more sources for info about all-pass filters and wikipedia has a wiki for that:
https://en.wikipedia.org/wiki/All-pass_filter
And it so happens that they describe one that is exactly the same as what I tried using, and they also show the same formula that I used for it: 1/(2PiRC)
Perhaps the formula is not the wrong one for that topology but rather the topology I'm using isn't the one that I should.
https://en.wikipedia.org/wiki/All-pass_filter
And it so happens that they describe one that is exactly the same as what I tried using, and they also show the same formula that I used for it: 1/(2PiRC)
Perhaps the formula is not the wrong one for that topology but rather the topology I'm using isn't the one that I should.
This is an old good trick to use. This is good also for new power amps build, or anything else for that matter, when they are powered up for the first time, to hopefully prevent some damage (and avoid starting WW3).
But my old amps are just old and I won't take the chance. We do know for a fact that electrolytic caps age a lot, and even if they don't blow up in your face, at least they no longer have their original value.
Well, I've been running a lot of sims on many different topologies and comparing results. In 2 way, 3 way and 4 ways.
I found out certain things that pretty much dictates certain things and mandates avoiding some.
For example, an all-pass 2nd order filter, is supposed to have the exact same phase response as a 4th order filter (high or low pass), and I tried different schemes and found that there aren't any single opamp based all-pass topo (that I know of) that actually match both the phase and at the same time keep the level at 1:1.
The first all-pass that I tried with a single opamp could not match exactly the phase response, even when properly aligned on the crossover frequency, the slope of the phase response could not match right. It's close, but not enough.
That topo does give a unity gain, so no issue with the level, but the phase just can't match.
Having read many things, including what our guru Self says about those things, I suspect that, if I understood what Self says in his book, this all-pass single opamp (2nd order) topo doesn't allow for the Q to be what is needed to get the phase response just right.
Now I tried an other topo, still 2nd order all-pass, with a single opamp, and got the phase response exactly right, but then the level is wrong, with a gain at close to -6db. That gain not even being an exact -6db would make it tricky to re-amplify behind it to get back the 1:1 ratio. But then it would add an opamp and would defeat the gain of using a single opamp topo for the all-pass.
So after all those tests, I find that I must use a dual opamp topo for the all-pass, and then get a "proper" phase response to match the other 4ft order filter (low/high).
However, I found also that although the phase response mostly matches, it's not 100%, at least at each end of the spectrum. Some discrepancies show up at low frequencies in some topos, and again at hf, but well beyond the upper audio range and at such a low level (in the stop band) that it would not really interfere with the other filters in the pass band.
Still, once all outputs are summed, we do see the influence on the global phase response, and a little bit on the level in some frequencies. So the global summed reponse isn't 100% flat, but quite close though, as the variations are small in magnitude, only a few 10s of mdb max.
Having simulated topologies used by actual makers on the market, the likes of furman, jbl, dbx, aphex, rane, etc... And others from various sources, like linkwitz, elektor, esp, etc...
I find that hardly any of them even try to do a phase linear type of arrangement, and have not found a single maker on the market using such a thing. Only in magazines or similar, do we find attempts at making a phase linear arrangement.
I made quite a few simulations of phase linear arrangements, and compared that to many others not paying attention to phase, and I came to the conclusion that for anything more than a 3 way, it's really pointless to even try to do a 4 way phase linear. The part count gets way up there, and the small discrepancies I mentioned about the phase not being quite right tend to magnify with the increased level of complexity from the higher number of bands.
I will post my last simulation in the next message to illustrate.
I found out certain things that pretty much dictates certain things and mandates avoiding some.
For example, an all-pass 2nd order filter, is supposed to have the exact same phase response as a 4th order filter (high or low pass), and I tried different schemes and found that there aren't any single opamp based all-pass topo (that I know of) that actually match both the phase and at the same time keep the level at 1:1.
The first all-pass that I tried with a single opamp could not match exactly the phase response, even when properly aligned on the crossover frequency, the slope of the phase response could not match right. It's close, but not enough.
That topo does give a unity gain, so no issue with the level, but the phase just can't match.
Having read many things, including what our guru Self says about those things, I suspect that, if I understood what Self says in his book, this all-pass single opamp (2nd order) topo doesn't allow for the Q to be what is needed to get the phase response just right.
Now I tried an other topo, still 2nd order all-pass, with a single opamp, and got the phase response exactly right, but then the level is wrong, with a gain at close to -6db. That gain not even being an exact -6db would make it tricky to re-amplify behind it to get back the 1:1 ratio. But then it would add an opamp and would defeat the gain of using a single opamp topo for the all-pass.
So after all those tests, I find that I must use a dual opamp topo for the all-pass, and then get a "proper" phase response to match the other 4ft order filter (low/high).
However, I found also that although the phase response mostly matches, it's not 100%, at least at each end of the spectrum. Some discrepancies show up at low frequencies in some topos, and again at hf, but well beyond the upper audio range and at such a low level (in the stop band) that it would not really interfere with the other filters in the pass band.
Still, once all outputs are summed, we do see the influence on the global phase response, and a little bit on the level in some frequencies. So the global summed reponse isn't 100% flat, but quite close though, as the variations are small in magnitude, only a few 10s of mdb max.
Having simulated topologies used by actual makers on the market, the likes of furman, jbl, dbx, aphex, rane, etc... And others from various sources, like linkwitz, elektor, esp, etc...
I find that hardly any of them even try to do a phase linear type of arrangement, and have not found a single maker on the market using such a thing. Only in magazines or similar, do we find attempts at making a phase linear arrangement.
I made quite a few simulations of phase linear arrangements, and compared that to many others not paying attention to phase, and I came to the conclusion that for anything more than a 3 way, it's really pointless to even try to do a 4 way phase linear. The part count gets way up there, and the small discrepancies I mentioned about the phase not being quite right tend to magnify with the increased level of complexity from the higher number of bands.
I will post my last simulation in the next message to illustrate.
Here is a 4 way phase linear simulation, based mostly on a 3 way published in an old elektor (87).
I simulated the 3 way version from elektor and the results are mostly the same as an other previous elektor publication of a 3 way without the phase linear topo.
What is gained by trying to be phase linear doesn't seem worthwhile. At least it's not too much more complex in a 3 way, but with a 4 way the part count gets way up there.
I can see that for each band in the 4 way, we have 6 or 7 opamps in the signal's path.
There are some advantages, such as using only one type of band filter, low or high. So in this case I used only high pass. And the rest is all delay/all-pass.
It's supposed to be "phase linear", but I'm not sure what this is supposed to be like on a plot, as I can see the plotted summed response shows the phase going way beyond 1000 degrees at hf. This is quite a delay, and I'm questioning this "phase linear" concept.
The combined plot of the 4 curves show the obvious flaw in the response in the high frequencies, especially for the low output, which shows a level coming back up instead of rolling off. And this is probably the major contributor to the slightly elevated level above about 8khz in the summed response. It's only in the order of about 20mdB, so it's really negligible, and this goes into a subwoofer anyway, which isn't capable of reproducing this, and the level is low on the low output at those frequencies.
The combined plot also shows that for the 3 higher bands, the phase response is pretty much superimposed, but a good part of the phase response from the low output above 1khz diverges.
I've been doing simulations on the SVF (State Variable Filter) approach, which has been used a lot by many makers out there, including the likes of rane, furman, behringer, etc...
This can make it somewhat easier to allow making the corner frequencies adjustable, with multiple gang pots, but on a 4 way, the 4 gang pots are quite hard to get and they should be highly precise. So that's not in the diy realm.
So I've been looking into other methods of doing the adjustments, not using any pots or electromechanical stuff.
There is no need to make the corner frequencies widely adjustable when we know the end goal, like what I'm looking to make. I already know basically what the corner frequencies are going to be, more or less, and so making this adjustable would only require small variations, for example for the 150hz lowest corner freq, it could be made with 4 steps, at perhaps about 25hz intervals, which should be enough to adjust. I would not bring that too low or too high anyway.
There are many ways to do the switching of resistors that would be used instead of a pot. I've looked at jfets, mosfets, digital switches in ICs, etc...
I'm mostly concerned about the unwanted influences that those various switching methods would bring. Extra capacitance introduced can cause issues with the frequency cornering. Inherent Ron in the switches add to the resistor values setting the frequencies, and the Ron isn't guaranteed to be the same for all switches.
Using the good old 4066 or 4016 type of switches, has been done in some builds, even in some equalizers, but I am wary of those, for several reasons, such as their admissible voltages and their Ron values. Using single voltage supply devices to switch signals that are handled by dual voltage supply others, with a higher rail voltage, can only be a cause for concern to me.
And I'm not sure they aren't a high source of distortion, and noise.
So now I am thinking about discrete and not integrated solutions to handle the switching. Such as jfet or mosfets.
Having looked at the fairly high Rdson in jfet, I'm thinking the mosfets with much lower Rdson would be much better, and perhaps they wouldn't be too noisy and too much of a distortion source.
The debate is open.
I simulated the 3 way version from elektor and the results are mostly the same as an other previous elektor publication of a 3 way without the phase linear topo.
What is gained by trying to be phase linear doesn't seem worthwhile. At least it's not too much more complex in a 3 way, but with a 4 way the part count gets way up there.
I can see that for each band in the 4 way, we have 6 or 7 opamps in the signal's path.
There are some advantages, such as using only one type of band filter, low or high. So in this case I used only high pass. And the rest is all delay/all-pass.
It's supposed to be "phase linear", but I'm not sure what this is supposed to be like on a plot, as I can see the plotted summed response shows the phase going way beyond 1000 degrees at hf. This is quite a delay, and I'm questioning this "phase linear" concept.
The combined plot of the 4 curves show the obvious flaw in the response in the high frequencies, especially for the low output, which shows a level coming back up instead of rolling off. And this is probably the major contributor to the slightly elevated level above about 8khz in the summed response. It's only in the order of about 20mdB, so it's really negligible, and this goes into a subwoofer anyway, which isn't capable of reproducing this, and the level is low on the low output at those frequencies.
The combined plot also shows that for the 3 higher bands, the phase response is pretty much superimposed, but a good part of the phase response from the low output above 1khz diverges.
I've been doing simulations on the SVF (State Variable Filter) approach, which has been used a lot by many makers out there, including the likes of rane, furman, behringer, etc...
This can make it somewhat easier to allow making the corner frequencies adjustable, with multiple gang pots, but on a 4 way, the 4 gang pots are quite hard to get and they should be highly precise. So that's not in the diy realm.
So I've been looking into other methods of doing the adjustments, not using any pots or electromechanical stuff.
There is no need to make the corner frequencies widely adjustable when we know the end goal, like what I'm looking to make. I already know basically what the corner frequencies are going to be, more or less, and so making this adjustable would only require small variations, for example for the 150hz lowest corner freq, it could be made with 4 steps, at perhaps about 25hz intervals, which should be enough to adjust. I would not bring that too low or too high anyway.
There are many ways to do the switching of resistors that would be used instead of a pot. I've looked at jfets, mosfets, digital switches in ICs, etc...
I'm mostly concerned about the unwanted influences that those various switching methods would bring. Extra capacitance introduced can cause issues with the frequency cornering. Inherent Ron in the switches add to the resistor values setting the frequencies, and the Ron isn't guaranteed to be the same for all switches.
Using the good old 4066 or 4016 type of switches, has been done in some builds, even in some equalizers, but I am wary of those, for several reasons, such as their admissible voltages and their Ron values. Using single voltage supply devices to switch signals that are handled by dual voltage supply others, with a higher rail voltage, can only be a cause for concern to me.
And I'm not sure they aren't a high source of distortion, and noise.
So now I am thinking about discrete and not integrated solutions to handle the switching. Such as jfet or mosfets.
Having looked at the fairly high Rdson in jfet, I'm thinking the mosfets with much lower Rdson would be much better, and perhaps they wouldn't be too noisy and too much of a distortion source.
The debate is open.
Attachments
At the moment, that's all I can do. I have no means of properly measuring responses on the speakers, and no funds to get all that needed gear for that. Plus my speakers are stored away and I have no room anywhere to unpack and set them up.
Besides, to properly do any useful measurements, the best would be in a big anechoic chamber, which I'll never have. The next best thing is to do the measurements outdoors, away from other sound sources preferably, which isn't practical either at the moment, although when the time comes, that's likely how I'll do it.
Doing any measurements indoors is totally pointless.
Now compare to going out to buy gear on the market, it can't be any worse, if someone gets a filters here, amps there, and speakers elsewhere, it's just a bunch of stuff that then needs to be set up and adjustments need to be made to make all this stuff work about alright.
For those setting up sound stage out there, bringing power amps, filters and and speaker gear, in a different venue every time, then of course everything has to be redone every single time, and never perfect.
I'm not doing an indoor hifi with only one listener who must always sit in the exact same spot every time. So there is no point in even trying to aim adjustments at one point in space, aligning phase of each speaker in detail and all that, since it's not going to be used anywhere near that way.
So what I'm trying to do is my own build of something that isn't so different from what you can buy on the market, except it's DIY!!!
The only thing I'm planning to buy, when the time comes, is a dual 31 bands equalizer, as this is pointless to build as a DIY project. It would cost more to make one, and probably not even get near the same performance.
I want to build my amps, filters, limiters, and the rest of the electronics to automate it all and make it one system that works.
Now the filters just need to have some flexibility to be adjustable enough so some level of coherency can be achieved, with measuring gear, but each time in a different venue.
Who cares if the speakers aren't 100% properly aimed at one single point in front of them, since nobody will be there in that spot to listen?
In a large room with a bunch of people (dancing), it's totally useless to have the speakers finely tune to aim sound at one point.
If I understand it right, that measuring gear with that mike and software to measure the response from speakers, is capable of ignoring the echo from a room, by catching the sound before it reflects from walls and whatever.
If this is true and works right, then the fact that it's done indoors doesn't matter.
Then each and every box with its speaker could be tested and measured, indoors, without having to worry about the environment.
This would be helpful and make things more practical.
But I can't go buying a bunch of gear right now, and I can't even unpack my existing gear to set that up for measurement, so all I can do is design something, and make it as adaptable as I can.
In the case, let's say, that at some point I decided to change a speaker (in pair) in the system, then whatever measurements were done before would be useless, and everything would have to be done again, and the same gear would have to be used, and be adjusted to the new config.
This is what's expected from the gear on the market anyway, and they definitely did not design the gear with anything specific in mind.
So what I'm trying to do isn't that much different, but I do know more about the final setup and the goal.
It won't be perfect, it can't be, so all I can do is design things the best I can.
I want limiters each band for all power amps, to not let them clip, but as I'm finding out, limiters can be quite a tricky thing.
I guess this subject would deserve its own thread. I didn't find any thread about specifically peak limiters.
I've been looking at many limiter designs. Most are of marginal quality in my opinion, and so far I don't think I've seen anything really that good. They work, but bring down the overall quality, even when not limiting.
I simulated a fairly simple limiter with opamps and a fet, and it seems to work fine, but I see the fet brings in distortion, even when not activating, and too many extra caps introduced in the signal path.
When I look who the gear on the market is designed, I can't see how we can have the best quality and performance. If we add extra things like noise gate, expander, or whatever, it really brings down the quality and introduces shitloads of extra stuff in the signal path.
I looked at that dbx 3bx expander, which has a 3 way crossover inside, with 3 separate limiters, and then the 3 bands are re-merged together after expansion.
When we add the input filters, input and output buffers, other small details like muting and whatever switches, one single piece of gear can insert a huge amount of stuff in the signal path, and the level gets lowered, then re-amplified several times over, and the distortion adds up, plus noise...
I want to keep things as spartan as possible, have the needed features without any extra unnecessary stuff.
If for example an expander, an equalizer and a noise gate are inserted between a mix table and the crossovers, that is a whole lot of extra electronics, with many extra things that are redundant, and could be eliminated if everything was combined together.
No need for the repeated input filters, plugs and cables, input and output buffers with a level adjustment most of the time, with attenuation and re-amplification...
Think about it!
If this is true and works right, then the fact that it's done indoors doesn't matter.
Then each and every box with its speaker could be tested and measured, indoors, without having to worry about the environment.
This would be helpful and make things more practical.
But I can't go buying a bunch of gear right now, and I can't even unpack my existing gear to set that up for measurement, so all I can do is design something, and make it as adaptable as I can.
In the case, let's say, that at some point I decided to change a speaker (in pair) in the system, then whatever measurements were done before would be useless, and everything would have to be done again, and the same gear would have to be used, and be adjusted to the new config.
This is what's expected from the gear on the market anyway, and they definitely did not design the gear with anything specific in mind.
So what I'm trying to do isn't that much different, but I do know more about the final setup and the goal.
It won't be perfect, it can't be, so all I can do is design things the best I can.
I want limiters each band for all power amps, to not let them clip, but as I'm finding out, limiters can be quite a tricky thing.
I guess this subject would deserve its own thread. I didn't find any thread about specifically peak limiters.
I've been looking at many limiter designs. Most are of marginal quality in my opinion, and so far I don't think I've seen anything really that good. They work, but bring down the overall quality, even when not limiting.
I simulated a fairly simple limiter with opamps and a fet, and it seems to work fine, but I see the fet brings in distortion, even when not activating, and too many extra caps introduced in the signal path.
When I look who the gear on the market is designed, I can't see how we can have the best quality and performance. If we add extra things like noise gate, expander, or whatever, it really brings down the quality and introduces shitloads of extra stuff in the signal path.
I looked at that dbx 3bx expander, which has a 3 way crossover inside, with 3 separate limiters, and then the 3 bands are re-merged together after expansion.
When we add the input filters, input and output buffers, other small details like muting and whatever switches, one single piece of gear can insert a huge amount of stuff in the signal path, and the level gets lowered, then re-amplified several times over, and the distortion adds up, plus noise...
I want to keep things as spartan as possible, have the needed features without any extra unnecessary stuff.
If for example an expander, an equalizer and a noise gate are inserted between a mix table and the crossovers, that is a whole lot of extra electronics, with many extra things that are redundant, and could be eliminated if everything was combined together.
No need for the repeated input filters, plugs and cables, input and output buffers with a level adjustment most of the time, with attenuation and re-amplification...
Think about it!
I ran many various simulations, with various topologies.
Although the state variable topology would be very nice, finding the 4gang pots isn't an easy thing to do, expensive and not guaranteed to be good enough.
So I don't think using the state variable topo will be helpful.
I don't want to do DSP, so analog filters remain the only solution.
I tried sallen-key filters with all-pass/delay cells, to try the phase-linear approach, as also described in some old elektor article, but that didn't seem much better than plain old regular method like what Rod Elliott proposes, so I went back to that type of topology and simulated a 4 way filter.
I used the LT1122 opamp to simulate this, as it was easier and that model comes built in with ltspice. Simplifies things.
The crossover frequencies came down to:
120Hz
1k5
8k
The caps are chosen and the resistors calculated from that. I have it all in a spreadsheet, which makes it easy if a change needs to be made, and to find the best resistor combinations to get the closest values to what was calculated.
I simulated with simplified common values, like on the E24/E48, and the errors don't really show much, but still, picking much closer values can be done, using all available values all the way to the E96 set.
All it's missing is output buffers, and reluctantly I suppose simple pots can be put there for adjustable output levels for each channel.
The output buffers can be balanced line drivers, so this can be made into a stand alone unit. With a built-in psu, it's a full fledged device.
Without sufficient financial means to go all the way, this can suffice to get it done and have something that works. Maybe not for a 100% flat response as is, depending so much on speakers (and the room). I think with a pink noise generator and some kind of spectrum analyzer, some adjustments can be made to make it work.
In any case, this beats any passive filter method.
And this could be kind of viewed as a prototype, to try it out and debug it.
Although the state variable topology would be very nice, finding the 4gang pots isn't an easy thing to do, expensive and not guaranteed to be good enough.
So I don't think using the state variable topo will be helpful.
I don't want to do DSP, so analog filters remain the only solution.
I tried sallen-key filters with all-pass/delay cells, to try the phase-linear approach, as also described in some old elektor article, but that didn't seem much better than plain old regular method like what Rod Elliott proposes, so I went back to that type of topology and simulated a 4 way filter.
I used the LT1122 opamp to simulate this, as it was easier and that model comes built in with ltspice. Simplifies things.
The crossover frequencies came down to:
120Hz
1k5
8k
The caps are chosen and the resistors calculated from that. I have it all in a spreadsheet, which makes it easy if a change needs to be made, and to find the best resistor combinations to get the closest values to what was calculated.
I simulated with simplified common values, like on the E24/E48, and the errors don't really show much, but still, picking much closer values can be done, using all available values all the way to the E96 set.
All it's missing is output buffers, and reluctantly I suppose simple pots can be put there for adjustable output levels for each channel.
The output buffers can be balanced line drivers, so this can be made into a stand alone unit. With a built-in psu, it's a full fledged device.
Without sufficient financial means to go all the way, this can suffice to get it done and have something that works. Maybe not for a 100% flat response as is, depending so much on speakers (and the room). I think with a pink noise generator and some kind of spectrum analyzer, some adjustments can be made to make it work.
In any case, this beats any passive filter method.
And this could be kind of viewed as a prototype, to try it out and debug it.
Attachments
Rather than daisy-chain all the filters, you can first split at 1.5kHz, then feed each output to another splitter. This shortens the max number of active stages of any path. Though there are arguments that the highest frequencies should have the shortest path as in your design. Either way you could modularize using 3 identical PCBs, with multiple footprints for the capacitors to allow different sizes/numbers of caps in a flexible way. Simplify the testing by powering everything up together, all outputs muted, with a test signal(s) on the input. Check the individual outputs are looking good, then switch to input signal and unmute all. This means far fewer control signals and relays (only one power up, only one unmute) - but the microcontroller can still say what's wrong so long as it can read the 4 outputs individually. Powering up in stages adds complexity that I don't think is warranted.