Indeed, fully complementary crossovers with consideration of individual driver behaviour was the aim when crossovers were first used. There has been much done since to improve the state of the art.
What benefit do you see here, as I think it would be less than ideal in a reflective environment. In a live sound environment there is more space to work. The distances can be calculated geometrically based on closer measurements if necessary.
Agree. Particularly with low crossovers in the 100 - 500 Hz range. Measure both drivers individually, then both drivers together. Then using a crossover software such as Xsim, simulate 3 drivers: one is the LF driver, one is the HF driver, and one is the measured sum of both drivers. Use the software to simulate the "both drivers together" response. Vary the delay of the LF or HF drivers until the simulated both-drivers matches the measured both-drivers.
In my case, crossing a 12 inch woofer to a 6 inch mid at 200 Hz, I estimate the voice coil locations and came up with 3.5 inches. i.e the woofer was 3.5 inches behind the mid. However using the above technique, the true value came out to be 9 inches. This is 5 inches behind the woofer magnet (!). It proves that the location of the voice coil is only an approximate estimate of the acoustic center. And it varies with frequency. At 700 Hz, the best match was with a 4 inch delay
With my Hypex Fusion system, I originally used the null method to define the delay. At 2 kHz with LR4, this worked real well. I got a very deep reverse null with 37 mm of delay. At 200 Hz it was a different story. The room response made it very hard to find a null. I used fast sweep sine waves, slow sweep sine waves, pink noise. I tried measurements at 1 meter, at the listening position, and I tried ground plane measurements. I never did get a pronounced null that I could differentiate from the room response. Using the software simulation method finally worked.
Last edited:
Thanks TMM, I could work with that.
What would you recommend as a measuring position and would you allign three drivers at once or in pairs (T/M and M/W). The latter seems easier
What would you recommend as a measuring position and would you allign three drivers at once or in pairs (T/M and M/W). The latter seems easier
I do pairs of adjacent drivers, primarily because it's easier and secondarily so the amplifier doesn't get upset driving a whole bunch of drivers in parallel with no crossover. Theoretically this can lead to poorer phase alignment in your measurements between drivers that weren't measured together (e.g. T/W in a TMW system) as any error in each measurement adds up, but this shouldn't matter as there should be very little overlap between them with the crossovers in place.
This is true, and I use HolmImpulse. However, it is always a good idea to do it this way as an idiot-check. It'll also compensate for physical mistakes such as bumping the speaker or mic when changing the connections between drivers while measuring. Nothing worse than realising after you've built the crossover that you stuffed up your phase data somewhere along the line and your speakers don't perform as designed.
Last edited:
Thanks and this seems reasonable. All my drivers are seperately powered so TMW is manageable. I'm using Hypex FA-123 plate amps
I am planning to measure outside in the yard.
Would you measure are listening distance (=2.8 meter) at TM level?
This may be more difficult since I don't have the technical details of my drivers (there stock B&W CM8 S2 drivers of which I haven't found all the details).
So, I'll have to just measure time and again and see how it looks.
Last edited:
Hi Allen,
What are some examples where you think the state of the art has improved?
Not in terms of how xovers are implemented (passive vs dsp etc),
but in terms of the science behind how the types of IIR filters work (BW, LR, Bessel, etc)?
The benefit i see in measuring at listening distance is that it's simply easier and more accurate at assuring the correct timing between acoustic centers, than measuring close and then making calculations to adjust.
It doesn't matter live vs indoors...space isn't an issue, nor are reflections, when measuring to find the acoustic center differences. (as long as no low-pass filter is in place)
Our measurement programs look for peak impulse energy to establish timing, which occurs well before reflections enter the impulse.
Peak energy, with any linear spaced digital measurement, will be inevitably skewed towards high frequency response.
But fortunately peak is a summation of all contributing frequencies, so peak summation tends to occur towards the middle of the range where a driver's mag response is flat (which means phase is too.)
Ime, the good news is, acoustic centers don't vary near as much vs freq, as common wisdom suggests.
But folks often screw up distance measurements because they try to make them with low-pass filters in place....that no worky. 😉
Makes it appear the acoustic center moves with freq ....
Hi all, a few more 2 cents of thoughts....
I don't think measuring a summed response of two drivers without a xover provides valid data.
Our measurement programs are forced to make some kind of phase curve out of the information gathered. So the two different phase curves of the drivers are being blended into some kind of amalgamation.
You can see this with measurement software that shows coherence, the statistical validity of the measurement by frequency.
Where two drivers overlap, without phase or time alignment, coherence will read lower. (Same thing from floor bounce, reflections etc.)
The null method works, but with a major caveat....it will show nulls both in-time and out-of-time.
It's because the null method says phase is 180 degrees out, but it does not say are you on the correctly timed null.
For instance, if xover is at 2kHz, the period is 1/2ms. If you have found a timing that produces a null, move the timing difference up 1/2ms, or down 1/2ms, (or multiples thereof) and you will still get a null at 2kHz.
You can watch response away from xover freq change as you bump up or down, but the null at 2kHz will remain pretty much constant.
Which is why phase traces are the only real way to align imo, as they show time as well.....
.... by taking individual traces, and time aligning them where they lay on top of each other. Hint: use timings found without low-pass filters in place.
Last thought i guess.... i really don't see why i would ever want to sim or use software to tune, when I can measure and tune in real time.
The only time i use software for EQ & xovers is when using FIR, and then a coefficient file must be built.
If I'm tuning via active IIR dsp, i just twist knobs so to speak, on each driver individually, watching how the measurement responds.
When each driver is corrected, tying them together with complementary xovers is a no brainer.
All that's needed to finish is to use timings found via hint, and adjust to equal levels. Takes about an hour to do a 3-way.
I don't think measuring a summed response of two drivers without a xover provides valid data.
Our measurement programs are forced to make some kind of phase curve out of the information gathered. So the two different phase curves of the drivers are being blended into some kind of amalgamation.
You can see this with measurement software that shows coherence, the statistical validity of the measurement by frequency.
Where two drivers overlap, without phase or time alignment, coherence will read lower. (Same thing from floor bounce, reflections etc.)
The null method works, but with a major caveat....it will show nulls both in-time and out-of-time.
It's because the null method says phase is 180 degrees out, but it does not say are you on the correctly timed null.
For instance, if xover is at 2kHz, the period is 1/2ms. If you have found a timing that produces a null, move the timing difference up 1/2ms, or down 1/2ms, (or multiples thereof) and you will still get a null at 2kHz.
You can watch response away from xover freq change as you bump up or down, but the null at 2kHz will remain pretty much constant.
Which is why phase traces are the only real way to align imo, as they show time as well.....
.... by taking individual traces, and time aligning them where they lay on top of each other. Hint: use timings found without low-pass filters in place.
Last thought i guess.... i really don't see why i would ever want to sim or use software to tune, when I can measure and tune in real time.
The only time i use software for EQ & xovers is when using FIR, and then a coefficient file must be built.
If I'm tuning via active IIR dsp, i just twist knobs so to speak, on each driver individually, watching how the measurement responds.
When each driver is corrected, tying them together with complementary xovers is a no brainer.
All that's needed to finish is to use timings found via hint, and adjust to equal levels. Takes about an hour to do a 3-way.
That would be inconvenient, but would it prevent success if the filtered driver as a system became the new point of reference? Or to put it another way, do we need a distance measurement or can we extrapolate a virtual delay reference through the phase response?
No, not really.
I think many people do that, try to 'extrapolate a virtual delay reference through the phase response'.
And that virtual delay moves with every change in freq, order, and type of xover.... resulting in needing to change delays as well.
Too many variables to juggle with ease and expect good odds, for me anyway.
I think it makes more sense to measure and match phase with the physical acoustic centers which don't move,
than to try to move virtual acoustic centers around with changing phase.
And that virtual delay moves with every change in freq, order, and type of xover.... resulting in needing to change delays as well.
Too many variables to juggle with ease and expect good odds, for me anyway.
I think it makes more sense to measure and match phase with the physical acoustic centers which don't move,
than to try to move virtual acoustic centers around with changing phase.
Perhaps the processes or methods are conflating the discussion. Specifically separating delay and phase? Before the line between design and implementation, it doesn't matter.
Not sure what you're saying Allen.....
What do you mean by a line between design and implementation ?
To me, delay and phase are entirely separate animals.
Delay being a fixed time constant, and phase being a relative waveform relationship between all frequencies.
For optimal results, i think neither should be used to correct the other.
What do you mean by a line between design and implementation ?
To me, delay and phase are entirely separate animals.
Delay being a fixed time constant, and phase being a relative waveform relationship between all frequencies.
For optimal results, i think neither should be used to correct the other.
Accurately measuring acoustic offset with a umik or other single channel measurement system is possible, Jeff Bagby (RIP) wrote a terrific guide on how to do this:
How to use OmniMic and PCD to find the Relative Acoustic Offset -
Techtalk Speaker Building, Audio, Video Discussion Forum
combined with his measurement whitepaper
http://audio.claub.net/software/FRD...curate In-Room Frequency Response to 10Hz.pdf
You have all you need to get good data to use with HFD or other crossover modelling software
No problem! For what it's worth, I have found that the part where you overlay the three curves and adjust offset until they match is easiest using XSIM rather than PCD as the guide uses, but YMMV
You may see the need to separate the two due to the way you implement a cross, but not everyone does or needs to.
Consider the case of group delay, introduced due to a natural rolloff, but not due to an actual delay...
Yep, not everyone sees the difference between time and phase...
too bad really, cause i think that's the major obstacle to a simpler, deeper, understanding of xovers....
Working with linear phase made it very easy to see the difference.
Before moving to linear phase xovers, I never could see the time vs phase distinction when juggling IIR xovers.......fwiw....
But now, i see the distinction holds just as true for IIR as it does for linear phase.
Group delay is probably the prime example of folks confusing time with phase.
Group delay is about phase, not time.
Group delay is not a constant...it varies with frequency, and is simply a derivative of phase expressed in terms of time at a particular frequency.
It's an awful, confusing misnomer imo.....
too bad really, cause i think that's the major obstacle to a simpler, deeper, understanding of xovers....
Working with linear phase made it very easy to see the difference.
Before moving to linear phase xovers, I never could see the time vs phase distinction when juggling IIR xovers.......fwiw....
But now, i see the distinction holds just as true for IIR as it does for linear phase.
Group delay is probably the prime example of folks confusing time with phase.
Group delay is about phase, not time.
Group delay is not a constant...it varies with frequency, and is simply a derivative of phase expressed in terms of time at a particular frequency.
It's an awful, confusing misnomer imo.....
We shouldn't make this more complicated than it is. If one studies waveforms that have had various filters applied, and sees how phase can go, for example, even beyond 360 degrees (and what that looks like) I think you'll find that the two are not that different.
Does anyone know what is applicable for the UMIK-1 microphone?
Q2: if I export the .frd files from REW into WIN-PCD I can't see any measurements. The bar (where you load the .frd file) turns green but no graph is shown (as in, the graph stays empty). Any help here?
Last edited:
I've found this XSIM tutorial which seems more helpful and actually works with loading measurements.
Speaker acoustic center - How to find it | Audio Judgement
Although I haven't made the actual measurements yet, I did the steps using old measruements and it worked out well.
I'm hoping (if the weather permits) to measure outside or in the garage of a family member.
So I'll hopefully be able to (finally!) first fix the acoustic offset (at tweeter level (= ear height) and listening distance (2.8 m)) and then design the crossovers.
Just a quick question, would it be wise to design the crossovers at listening distance as well? It seems too far away.
Or just measure everything at 1 m for sake of simplicity...
So many options!
Speaker acoustic center - How to find it | Audio Judgement
Although I haven't made the actual measurements yet, I did the steps using old measruements and it worked out well.
I'm hoping (if the weather permits) to measure outside or in the garage of a family member.
So I'll hopefully be able to (finally!) first fix the acoustic offset (at tweeter level (= ear height) and listening distance (2.8 m)) and then design the crossovers.
Just a quick question, would it be wise to design the crossovers at listening distance as well? It seems too far away.
Or just measure everything at 1 m for sake of simplicity...
So many options!
Last edited: