Up until now, all my custom speaker builds have incorporated active crossovers and multi-channel amplification.
While I am still convinced this is by far the best way to implement a speaker, I am planning a couple builds that (for practicality sake) will use a passive crossover.
I have read a lot of info on the subject but the OCD in me leaves me with some questions yet.
Some of this, I have asked previously.. but I want to be sure I am 100% spot-on regarding the answers.
1) Speaker sensitivity - the electical measurement itself
I understand the basics here: feed a speaker 1 watt at 1k and measure the SPL with a calibrated mic @ 1 meter distance. But details matter.
Question 1.a: Is it not more accurate to measure this AFTER the speaker is actually mounted in the baffle in which it will reside? This affects the final SPL after all, correct?
Question 1.b: In all reality, the TRUE sensitivity (that counts) is when you perform the "what is 1 watt" based on the real driver impedance, correct? In other words - I should do an impedance sweep (using DATS).. take the exact impedance of that driver @ 1k (while mounted in the actual baffle).. then calibrate the signal (measuring voltage) until the product comes out to exactly 1 watt. Is this a more correct way to get a precise measurement of sensitivity? Or should I use the "nominal" impedance (also calculated by DATS) - regardless of the impedance-frequency relation that is "seen" by the amp?
Question 1.c: I imagine the sensitivity of a driver actually fluctuates as frequency changes to some degree since the impedance certainly does, no? Is this taken into account when dealing with a passive xover?
2) Crossover points.
I have asked part of this previously... but will post the one question here along with the others just to be sure everything is answered in one spot for the sake of the algorythms.
Question 2.a: I assume that the calculation for the crossover component values (and frequencies) should be done based on the measured impedance of each driver at the crossover frequency. Correct? This is what seemed to be confirmed to me previously. So for examplle: I plan to crossover a particular driver at 1k.. I should first measure what the exact impedance for that driver is (impedance sweep while mounted in the actual baffle). So if a driver with nominal impedance of 4 ohms is actually 5.5 ohms at 1k, then the 5.5 ohms is what should be used in the calculations for the crossover - correct?
Question 2.b: What is generally the most preferred alignment for typical 3-way passive designs? L-R? Also - provided I use drivers well within their limits (and far enough away from their Fs).. is 1-st order not ideal? I hear this is what my old Vandersteens use (they sound pretty damn good to me). For my own purposes, I will be using these speakers (I will build) for close and near-field listening. I understand 1st order is not ideal for this?
Question 2.c: How much should I worry about time alignment? I hear this is always better to do physically when possible (stepped baffle) - which has its own issues. But how many here are doing this stuff in the passive crossover section?
I have a huge assortment of film caps coming from AliExpress (have had excellent success with over the years). A dozen different values (4 of each value) specifically for crossovers. I also wind my own inductors. So I can certainly "play around" with the crossover as needed here. My assortment of power resistors is a little more limited, but I may get an assortment of those on the way as well (for matching sensitivities where needed).
All the advice is welcome!
-Dean
While I am still convinced this is by far the best way to implement a speaker, I am planning a couple builds that (for practicality sake) will use a passive crossover.
I have read a lot of info on the subject but the OCD in me leaves me with some questions yet.
Some of this, I have asked previously.. but I want to be sure I am 100% spot-on regarding the answers.
1) Speaker sensitivity - the electical measurement itself
I understand the basics here: feed a speaker 1 watt at 1k and measure the SPL with a calibrated mic @ 1 meter distance. But details matter.
Question 1.a: Is it not more accurate to measure this AFTER the speaker is actually mounted in the baffle in which it will reside? This affects the final SPL after all, correct?
Question 1.b: In all reality, the TRUE sensitivity (that counts) is when you perform the "what is 1 watt" based on the real driver impedance, correct? In other words - I should do an impedance sweep (using DATS).. take the exact impedance of that driver @ 1k (while mounted in the actual baffle).. then calibrate the signal (measuring voltage) until the product comes out to exactly 1 watt. Is this a more correct way to get a precise measurement of sensitivity? Or should I use the "nominal" impedance (also calculated by DATS) - regardless of the impedance-frequency relation that is "seen" by the amp?
Question 1.c: I imagine the sensitivity of a driver actually fluctuates as frequency changes to some degree since the impedance certainly does, no? Is this taken into account when dealing with a passive xover?
2) Crossover points.
I have asked part of this previously... but will post the one question here along with the others just to be sure everything is answered in one spot for the sake of the algorythms.
Question 2.a: I assume that the calculation for the crossover component values (and frequencies) should be done based on the measured impedance of each driver at the crossover frequency. Correct? This is what seemed to be confirmed to me previously. So for examplle: I plan to crossover a particular driver at 1k.. I should first measure what the exact impedance for that driver is (impedance sweep while mounted in the actual baffle). So if a driver with nominal impedance of 4 ohms is actually 5.5 ohms at 1k, then the 5.5 ohms is what should be used in the calculations for the crossover - correct?
Question 2.b: What is generally the most preferred alignment for typical 3-way passive designs? L-R? Also - provided I use drivers well within their limits (and far enough away from their Fs).. is 1-st order not ideal? I hear this is what my old Vandersteens use (they sound pretty damn good to me). For my own purposes, I will be using these speakers (I will build) for close and near-field listening. I understand 1st order is not ideal for this?
Question 2.c: How much should I worry about time alignment? I hear this is always better to do physically when possible (stepped baffle) - which has its own issues. But how many here are doing this stuff in the passive crossover section?
I have a huge assortment of film caps coming from AliExpress (have had excellent success with over the years). A dozen different values (4 of each value) specifically for crossovers. I also wind my own inductors. So I can certainly "play around" with the crossover as needed here. My assortment of power resistors is a little more limited, but I may get an assortment of those on the way as well (for matching sensitivities where needed).
All the advice is welcome!
-Dean
It seems to me you are on the right track and often answer your own questions correctly. May I suggest, however, that you get a good book on crossover design, such as Vance Dickason's "Loudspeaker Design Cookbook", published by the KCK Media Corp., first printed in 2006. ISBN: 1-882580-47-8. That book has everything you could ever possibly want to know about designing passive crossovers.
Hi!
The best investment you can make when building passive crossovers is a calibrated analog mic.
You need to have time aligned frequency/phase response and impedance for each speaker so as to load them in a crossover simulator tool such as XSim or Vituix. Time aligned is needed in order to make sure relative phases among the speakers have same reference.
If you are not planning to have a calibrated mic, then you follow the post indicated in post #3 as an alternative.
Instead of a single impedance measurement, you will have the complete curve - much more precise. And yes, the impedance deeply interacts with the crossover elements. When you don't have the curves, them you should look for the impedance around the crossover point as an approximation.
So, you should use at least 2nd order as a general rule and sometimes a mix with 3rd order or notch filters.
In addition, off axis can be affected if a lot of overlap is used, which is the case of 1st order. It all depends.
Regarding frequency choice, it depends on each speaker response. Typically you could start with something around 500Hz and 3.5kHz for a 3way.
The choice of crossover frequencies will impact many aspects, including the size of inductors/capacitors. The lower frequency between woofer/mid the higher inductance and capacitance will be needed, which add costs and size.
The crossover can change phases to align all the frequencies to the point of listening as long as you have time-aligned measurements, so your curves will reflect any physical disaligment. Playing with the crossover, you can compensate everything.
Remember what affect phase rotation:
-Physic position since one speaker is behind or ahead the other
-Electric crossover components - the electrical signal phase applied to each speaker is changed by the crossover
-Acoustic effects - the speaker changes phase along the frequency range.
When you make actual measurements, physic and acoustic effects on phase are incorporated into the curves.
And when you load the curves into the simulator and place the electrical components, you add the electric influence on top of everything.
Sumary:
1. Simulate the box to get the desired bass response - this is based on woofer's T/S and WinISD, for example, in order to get the volume and port dimensions (if ported)
2. Build the box and auxiliary mid-range box if necessary
3. Install the speakers and bring the individual speaker wires to the outside, so they can be individually fed by the amplifier
4. Choose a power level that is compatible with the tweeter (let's say 0.5W to 1W) so as to not damage it
5. Position the mic
6. Adjust level in the software that will create the curves (REW for example)
7. Don't change any level, box position or mic position so you keep physical aligment untouched
8. Take each speaker frequency/phase response
9. Take each speaker impedance
Load the 6 files, if 3-way, to a crossover simulator such as XSim.
Play with components to get the flat response.
The best investment you can make when building passive crossovers is a calibrated analog mic.
You need to have time aligned frequency/phase response and impedance for each speaker so as to load them in a crossover simulator tool such as XSim or Vituix. Time aligned is needed in order to make sure relative phases among the speakers have same reference.
If you are not planning to have a calibrated mic, then you follow the post indicated in post #3 as an alternative.
Yes, cause the final enclosures (you might have a small enclosure for mid-range) and baffle affect frequency, phase and impedance measurements.
When you measure the speakers in the final enclosure, you sweep from 20 to 20kHz using a constant VOLTAGE, cause this is the real situation when you are listening to music. Along the frequencies, current and power will change due to impedance change. Speakers are built to be SPL flat (not power flat) as possible when applying a constant voltage along the frequencies.
Sure impedance changes a lot, specially around the resonance frequency. The way to deal with all these variables is to measure individual responses (freq/phase/impedance) and than simulate. All the complexities will be counted and your work will be to fine tune the capacitors, inductors and resistors. Knowledge, experience and creativity will have to be used.
Yes and no. Using the method of getting actual frequency reponse, phase response and impedance curves you don't have to deal with that.
Instead of a single impedance measurement, you will have the complete curve - much more precise. And yes, the impedance deeply interacts with the crossover elements. When you don't have the curves, them you should look for the impedance around the crossover point as an approximation.
It will depend on your speaker responses. First order can be used for woofers (or subwoofers) and mid-ranges but not for tweeters. 1st order doesn't provide a good protection for tweeters (mid and bass can damage tweeters).
So, you should use at least 2nd order as a general rule and sometimes a mix with 3rd order or notch filters.
In addition, off axis can be affected if a lot of overlap is used, which is the case of 1st order. It all depends.
Regarding frequency choice, it depends on each speaker response. Typically you could start with something around 500Hz and 3.5kHz for a 3way.
The choice of crossover frequencies will impact many aspects, including the size of inductors/capacitors. The lower frequency between woofer/mid the higher inductance and capacitance will be needed, which add costs and size.
When you can physically align all the speakers, it's better. But this is not absolutely necessary and normally difficult to achieve.
The crossover can change phases to align all the frequencies to the point of listening as long as you have time-aligned measurements, so your curves will reflect any physical disaligment. Playing with the crossover, you can compensate everything.
Remember what affect phase rotation:
-Physic position since one speaker is behind or ahead the other
-Electric crossover components - the electrical signal phase applied to each speaker is changed by the crossover
-Acoustic effects - the speaker changes phase along the frequency range.
When you make actual measurements, physic and acoustic effects on phase are incorporated into the curves.
And when you load the curves into the simulator and place the electrical components, you add the electric influence on top of everything.
Sumary:
1. Simulate the box to get the desired bass response - this is based on woofer's T/S and WinISD, for example, in order to get the volume and port dimensions (if ported)
2. Build the box and auxiliary mid-range box if necessary
3. Install the speakers and bring the individual speaker wires to the outside, so they can be individually fed by the amplifier
4. Choose a power level that is compatible with the tweeter (let's say 0.5W to 1W) so as to not damage it
5. Position the mic
6. Adjust level in the software that will create the curves (REW for example)
7. Don't change any level, box position or mic position so you keep physical aligment untouched
8. Take each speaker frequency/phase response
9. Take each speaker impedance
Load the 6 files, if 3-way, to a crossover simulator such as XSim.
Play with components to get the flat response.
As suggested, you probably don't need to know the sensitivity. If you do your measurements of different drivers at the same level, you have the relative sensitivity covered.
Let us know if you plan to use a simulator, since it will change our answers to your questions.
This is difficult to answer without a full polar measurement approach. Otherwise, many choose to use L-R but you can remain flexible.
I prefer not to compromise physically. I don't use the time alignment based approach. You will see what needs to be done when you work to bring together the responses.
With your previous active designs, have they been based only on electrical parameters.. ie are you up to speed working together with acoustic measurements to develop filters?
Thanks so far on the wealth of info!
@kurtis Richter - I just purchased the book you referenced. I have heard it mentioned time and time again - so I am going to take some good advice to heart 🙂
To address a few of the questions you all have already asked:
1) I already own DATS V3 and am used to doing test sweeps, basic measurements, etc.
2) I also already own a calibrated Dayton Onnimic V2. I am also decently familiar w/ REW and taking basic measurements. I even have a couple decent SPL meters from Radio Shack from way back in the 90's.
3) I am also about to dump the $$ on a decent audio interface w/ XLR, phantom power, etc as well as another calibrated [XLR] mic so I can have a true loopback so I can have a proper timing reference. Likely an entry level Earthworks mic with either Focusrite or M-Audio interface.
4) For basic box design/simuation, I have usually used WinISD. Sometimes I use bassbox. Beyond that, nothing more really. I have experimented with hornresp and Martin Kings TL tables as well back when I was on a TL kick.
5) I am going to start playing around with Xsim unless someone has a better recommendation for xover simulation.
Most by design work has either been SUPER BASIC (not even worth mentioning) or purely active. Most the speaker projects I have built have used active filters (based on Elliot Sound Products designs), in combination with: modified china plate amps (TPA3116, 2355, etc), switching power supplies, etc.
Some of my recent work has implemented some more purely-analog approaches using active filters in combination with gainclone-style chip amps, etc.
Yes, I typically have done basic measurements of the drivers before working on the active filters. But in purely-active setups.. this is much simpler since the filter network is not interacting with reactive components (it simply drives a high impedance endpoint what does not really change much with frequency). Any effect of the reactive components in my previous builds is easily dealt with via a simple active filter to tame-out.
My workflow has been something along these lines:
1) Buy the drivers to more/less suite what I am trying to build
2) Break in the woofer 48 hours with 10hz sinewave, measure ts and use that data to design the bass portion
3) Build the speaker
4) Measure actual in-speaker responses (direct, no crossover).. use that info along with any other "knowns" to build the needed active filters.
5) Final test/measurement... sometimes I need to re-visit and build additional filters, etc.
I am by far any sort of expert - even on the active stuff. But I have been very happy with the results of most my work thus far.
Problem is: even some of my projects using active components... Sometimes, I hate to implement ANOTHER channel of amplicication (like in a 3 way design)... when the two channels of 40 or 50 watts is already more than enough. Example: crossing the woofer <-> mid and then also crossing the mid <-> tweeter and having to amplify 3 channels. I often find myself thinking: damn, I really should just have a high pass, use a single 2 channel chip amp and then split the mid <-> tweeter using passive components. But I have always stayed away from this due to my fear of screwing something up. Honestly - purely active domain filters seems (to me) a foolproof way if one is not afraid of circuits and soldering.
I have had many friends comment on how great my active builds sound and want something similar - but want to use their own amplifiers. I do not exactly do this for a living (software engineer by trade). But it is always fun to build something that sounds great over a few beers. And for most people, a more traditional design using passive xovers is certainly the way to go.
I imagine this is why Paul McGowen once said in his videos something along the lines of "If every manufacturer had the choice, they would almost certainly build purely-active loudspeakers". But as is always the case - everthing in this hobby is a compromise :-D
@kurtis Richter - I just purchased the book you referenced. I have heard it mentioned time and time again - so I am going to take some good advice to heart 🙂
To address a few of the questions you all have already asked:
1) I already own DATS V3 and am used to doing test sweeps, basic measurements, etc.
2) I also already own a calibrated Dayton Onnimic V2. I am also decently familiar w/ REW and taking basic measurements. I even have a couple decent SPL meters from Radio Shack from way back in the 90's.
3) I am also about to dump the $$ on a decent audio interface w/ XLR, phantom power, etc as well as another calibrated [XLR] mic so I can have a true loopback so I can have a proper timing reference. Likely an entry level Earthworks mic with either Focusrite or M-Audio interface.
4) For basic box design/simuation, I have usually used WinISD. Sometimes I use bassbox. Beyond that, nothing more really. I have experimented with hornresp and Martin Kings TL tables as well back when I was on a TL kick.
5) I am going to start playing around with Xsim unless someone has a better recommendation for xover simulation.
Most by design work has either been SUPER BASIC (not even worth mentioning) or purely active. Most the speaker projects I have built have used active filters (based on Elliot Sound Products designs), in combination with: modified china plate amps (TPA3116, 2355, etc), switching power supplies, etc.
Some of my recent work has implemented some more purely-analog approaches using active filters in combination with gainclone-style chip amps, etc.
Yes, I typically have done basic measurements of the drivers before working on the active filters. But in purely-active setups.. this is much simpler since the filter network is not interacting with reactive components (it simply drives a high impedance endpoint what does not really change much with frequency). Any effect of the reactive components in my previous builds is easily dealt with via a simple active filter to tame-out.
My workflow has been something along these lines:
1) Buy the drivers to more/less suite what I am trying to build
2) Break in the woofer 48 hours with 10hz sinewave, measure ts and use that data to design the bass portion
3) Build the speaker
4) Measure actual in-speaker responses (direct, no crossover).. use that info along with any other "knowns" to build the needed active filters.
5) Final test/measurement... sometimes I need to re-visit and build additional filters, etc.
I am by far any sort of expert - even on the active stuff. But I have been very happy with the results of most my work thus far.
Problem is: even some of my projects using active components... Sometimes, I hate to implement ANOTHER channel of amplicication (like in a 3 way design)... when the two channels of 40 or 50 watts is already more than enough. Example: crossing the woofer <-> mid and then also crossing the mid <-> tweeter and having to amplify 3 channels. I often find myself thinking: damn, I really should just have a high pass, use a single 2 channel chip amp and then split the mid <-> tweeter using passive components. But I have always stayed away from this due to my fear of screwing something up. Honestly - purely active domain filters seems (to me) a foolproof way if one is not afraid of circuits and soldering.
I have had many friends comment on how great my active builds sound and want something similar - but want to use their own amplifiers. I do not exactly do this for a living (software engineer by trade). But it is always fun to build something that sounds great over a few beers. And for most people, a more traditional design using passive xovers is certainly the way to go.
I imagine this is why Paul McGowen once said in his videos something along the lines of "If every manufacturer had the choice, they would almost certainly build purely-active loudspeakers". But as is always the case - everthing in this hobby is a compromise :-D