So you measured in an anechoic chamber with a USB mic and no acoustic timing reference?
If that is true, then your measured phase data is crap and you are down to Z setting and "minimum phase checkbox = checked" on driver tab. X and Y can remain ZERO if you have not moved the mic (because the off-axis behaviour has been captured due to the mic being in a fixed - e.g. on tweeter axis position).
I would however still use your farfield data only for crossover design. If your first reflection is about 22.5msec according to your impulse responses, you're good down to ~ 44Hz. So no need to splice any response.
If that is true, then your measured phase data is crap and you are down to Z setting and "minimum phase checkbox = checked" on driver tab. X and Y can remain ZERO if you have not moved the mic (because the off-axis behaviour has been captured due to the mic being in a fixed - e.g. on tweeter axis position).
I would however still use your farfield data only for crossover design. If your first reflection is about 22.5msec according to your impulse responses, you're good down to ~ 44Hz. So no need to splice any response.
Last edited:
You can use the acoustic timing reference. However the loopback option is so simple. One channel goes from output to input, the other to the amp and from the mic... On the other hand for a portable acoustic reference you'd need a speaker mounted on your mic boom (and possibly another amp).
@Dave Bullet Yes I know now I have been an idiot and not taken the right measurements but I guess failure is one way to learn. I was in the lucky position of the chamber only costing me a couple of days leave so not too expensive.
@AllenB will look into that, I am running a 6 channel amp and a Minidsp so the acoustic methods easy to rig up but if the loop back method is again easy with the Behringer which has two channels then I will try and give that a go.
Thanks everyone for the input. All very much appreciated.
@AllenB will look into that, I am running a 6 channel amp and a Minidsp so the acoustic methods easy to rig up but if the loop back method is again easy with the Behringer which has two channels then I will try and give that a go.
Thanks everyone for the input. All very much appreciated.
Just found this - https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_REW.pdf
Another Doh moment!!!!!
Looks like I have been doing something else wrong which I suspect screws up the phase/timing even more.
As I am running a Minidsp active crossover at the moment I set the output channel in REW to use that I.e. plays direct from REW to the minidsp which then feeds the six channel Poweramp. This is useful as I can plug in all three driver and use Minidsp to mute those I am not measuring. The mic is then set in REW via the Behringer interface so no connection between the system making the sound and that measuring it. What an idiot!
I think I have figured out the loop back process and this needs the Poweramp to be plugged into the Behringer so all in/out goes via that. Can I use a stereo integrated amp so long as I have it in direct mode (no tone controls) and don’t touch the volume knob once set up?
I may get some time this evening to investigate, I have an old speaker so can play around with that rather than potentially damaging the good stuff.
Every day is a school day.
Another Doh moment!!!!!
Looks like I have been doing something else wrong which I suspect screws up the phase/timing even more.
As I am running a Minidsp active crossover at the moment I set the output channel in REW to use that I.e. plays direct from REW to the minidsp which then feeds the six channel Poweramp. This is useful as I can plug in all three driver and use Minidsp to mute those I am not measuring. The mic is then set in REW via the Behringer interface so no connection between the system making the sound and that measuring it. What an idiot!
I think I have figured out the loop back process and this needs the Poweramp to be plugged into the Behringer so all in/out goes via that. Can I use a stereo integrated amp so long as I have it in direct mode (no tone controls) and don’t touch the volume knob once set up?
I may get some time this evening to investigate, I have an old speaker so can play around with that rather than potentially damaging the good stuff.
Every day is a school day.
I'm not familiar with VituixCAD but I have a couple of suggestions that maybe others can agree or dismiss.
Measurment method:
1) In REW you really want to use the ASIO drivers for the Behringer interface. This is usually more reliable for measurments. It also fixes you to only using that device for all in and out.
2) Use the left analog output of the behringer to feed the MiniDSP and on to your amp and Speaker.
3) loop the right output of the behringer to the right input of the behringer. This is your timing reference channel. You could loop it through the MiniDSP on the way, but since it will simply introduce a fixed time delay that is not really a concern.
4) your mic goes to the left Input channel of the behringer.
CAD
If the timing data of your current data is good (I'm unclear about that since I don't understand if you had any loopback?) You could just ignore simulation of the vertical axis. What could you do about it anyway? The speaker is already built, right? Just choose if you want to xover between mid and tweeter as low as possible for the widest vertical listening window or at a frequency wavelength equal to the driver spacing * 1.2 for best power response.
As I'm also in the UK may I ask which chamber you used?
Measurment method:
1) In REW you really want to use the ASIO drivers for the Behringer interface. This is usually more reliable for measurments. It also fixes you to only using that device for all in and out.
2) Use the left analog output of the behringer to feed the MiniDSP and on to your amp and Speaker.
3) loop the right output of the behringer to the right input of the behringer. This is your timing reference channel. You could loop it through the MiniDSP on the way, but since it will simply introduce a fixed time delay that is not really a concern.
4) your mic goes to the left Input channel of the behringer.
CAD
If the timing data of your current data is good (I'm unclear about that since I don't understand if you had any loopback?) You could just ignore simulation of the vertical axis. What could you do about it anyway? The speaker is already built, right? Just choose if you want to xover between mid and tweeter as low as possible for the widest vertical listening window or at a frequency wavelength equal to the driver spacing * 1.2 for best power response.
As I'm also in the UK may I ask which chamber you used?
@Tenson Thanks for the input, much appreciated. The Chamber is where I work so not available to the public, managed to squeeze some time in there between jobs.
Quick Update -
Connected the Behringer output to the input and did a calibration, all looked good with slight roll off below 20hz and above 10khz but only 0.2dB. Looks pretty well behaved.
I then rigged up an old integrated amp (Denon PMA6.5) and an even older surround speaker from a CRT TV (Toshiba SS-V90) that looks like a full range driver. Used these just in case something went bang.
As @Tenson and many others suggested I connected :
Frequency plots
I then looked at each impulse/step response and read off the position of the first peak and converted these to distance using 343m/s for the speed of sound.
Pretty please with those results once normalized around the 100mm position. Obviously my starting position was about 20mm out but other than that they line up well.
Thanks again for all of the help, looks like I am getting there. Will see if I can measure a proper driver (have an old Mackie HR824 I can strip the drivers out or connect up at the rear) later to see if it works Ok with that as I had some issues getting the TV speaker loud enough.
Quick Update -
Connected the Behringer output to the input and did a calibration, all looked good with slight roll off below 20hz and above 10khz but only 0.2dB. Looks pretty well behaved.
I then rigged up an old integrated amp (Denon PMA6.5) and an even older surround speaker from a CRT TV (Toshiba SS-V90) that looks like a full range driver. Used these just in case something went bang.
As @Tenson and many others suggested I connected :
- Left Out to Amplifier > Speaker
- Right out > Channel 1 (R) in (set to line in, half gain)
- Channel 2 (L) in to Mic
Frequency plots
I then looked at each impulse/step response and read off the position of the first peak and converted these to distance using 343m/s for the speed of sound.
Pretty please with those results once normalized around the 100mm position. Obviously my starting position was about 20mm out but other than that they line up well.
Thanks again for all of the help, looks like I am getting there. Will see if I can measure a proper driver (have an old Mackie HR824 I can strip the drivers out or connect up at the rear) later to see if it works Ok with that as I had some issues getting the TV speaker loud enough.
Last edited:
There must be a soccer field near you somewhere? It is your national sport. You won't bother any neighbours either.
@Nico Ras Only problem is getting mains power out to the middle of the field for the amplifier. Also, most open spaces in the UK are owned by the Government (local council) and so their use, in particular any amenities (power) is strictly policed. The UK does not have a culture of people doing strange things in the middle of football fields so I suspect I will end up with either a lot of attention or at worst reporting to the authorities plus during the day they are often in use for dog walker, small children playing, people exercising etc.. And last but not least most football grounds are next to main roads and so traffic noise may be as much an issue as indoor reflections.
I've done it in fields. The main issue is wind, both blowing over speakers perched on stands and also noise in the measurement. And yes, curious people ask what I'm doing every 10 mins. Also the ground bounce is still a bas**rd. Ground plane measurements have never worked well for me, unless you splice with free field, but then we have issues of phase etc again.
Another update -
So loopback seems to be working fine. I did a test with my old speakers with and without loopback, results for delay are similar but not the same. Measurements were done in my 10ftx9ft study with lots of hard surfaces so a fair amount of noise but it was mainly to test the system was working.
I then measured my main speaker which has protection capacitors in place for active crossovers but the results with loopback seem to confirm the paired driver measurements results (20uS Mid/Tweeter and 270uS Woofer/Mid).
So it looks like approx 30uS for Mid/Tweeter and 260uS for Woofer/Tweeter, so pretty close. The conversion to mm is also pretty close to my Pythagoras calcs as well. So all seems to stack up.
Now just need to find time to move everything into either the lounge or the garage and do some 360 measurements with and without the capacitors and see if I can get any more time in the anechoic chamber to get a clean set (probably not this side of Chistmas).
Just a thought, if I added a delay in REW to my original incorrect measurements to get them in line with the 0/30/260uS offset from this test would that make a usable set with the right offset ? Bit of a painful job but could be done.
So loopback seems to be working fine. I did a test with my old speakers with and without loopback, results for delay are similar but not the same. Measurements were done in my 10ftx9ft study with lots of hard surfaces so a fair amount of noise but it was mainly to test the system was working.
I then measured my main speaker which has protection capacitors in place for active crossovers but the results with loopback seem to confirm the paired driver measurements results (20uS Mid/Tweeter and 270uS Woofer/Mid).
So it looks like approx 30uS for Mid/Tweeter and 260uS for Woofer/Tweeter, so pretty close. The conversion to mm is also pretty close to my Pythagoras calcs as well. So all seems to stack up.
Now just need to find time to move everything into either the lounge or the garage and do some 360 measurements with and without the capacitors and see if I can get any more time in the anechoic chamber to get a clean set (probably not this side of Chistmas).
Just a thought, if I added a delay in REW to my original incorrect measurements to get them in line with the 0/30/260uS offset from this test would that make a usable set with the right offset ? Bit of a painful job but could be done.
Minimum phase means it removes all delay relayed phase shift. So it's like the bare boans phase of the driver.
PS: I wasn't implying you were an idiot, sorry. I was just trying to confirm your measurement setup so that we could assess what / how you could use those measurements effectively. The anechoic chamber measurements are still useful by the sounds of it. There's never any harm IMHO of taking measurements in 2 different ways and applying the same XO simulation, to see if they align with the same output (acoustic behaviour). It gives confidence in consistency at least.
This is correct.
@Ugg10 I noticed you said
"I think I have figured out the loop back process and this needs the Poweramp to be plugged into the Behringer so all in/out goes via that."
This may well exceed the power expected for what is effectively a line level input and may fry the input. I don't know about how flexible / tolerant the behringer is here. A line level loopback is all you need, especially if you have taken an amplifier calibration (and low level) which it seems you have done to eliminate any non-linearity from the amp in the measurements.
The phase related delay (if any) from the amp won't matter as it will be baked into all driver measurement at all angles, so there's no need to chase phase right to the output as such. The start marker in your gate is more important. The closer to the impulse (of the closest driver so you don't chop it off for drivers further away) will minimise the phase wrap in the measurements (especially as frequency goes up). Whilst this won't matter to XO software (where you put "L" / start marker) - it just makes it easier on the eye to see the phase without excessive wrap.
I believe this is done via Controls -> Offset t=0. enter the offset in msec, e.g.:
I don't know how you apply this in bulk however
that sounds positive, so if I -
- Import the original non-loopback anarchic measurements into Vituix
- Check the minimum phase button on the driver set up tab
- Import the impedance measurements
- Add the new loop back offsets into the driver set up tab
- Add the Y off set of the drivers as identified by @hifijim in post #2
then I am ready to design the crossover understanding that I only have a 180deg spin which will affect the results somewhat. Hopefully the difference between 20/270uS (pairs alignment) which I have at the moment and 28/256uS is not massive.
I guess the only questions is that I get slightly different offsets if I use the top of the first peak on the impulse vs the step traces - which one should I be using or something different.
It's good to cross check.
That said, most wouldn't bother calculating the offset unless they don't have accurate measurements so there's no need to complicate matters moving forward.
Diffraction is also delayed.. so going minimum phase renders it inaccurate. It isn't 'extracted', it get's stomped on.
Minimum phase is one thing, removing overall delay from the impulse is another thing.
All the rest are correct except the above I believe. If you have:
a) measured all drivers on the same (say tweeter) axis - then the Y offset "roll off" will be baked into your measurement. Since the midrange and woofer will be consistently off (Y) axis in relation to your on tweeter axis measurement. If you put a negative Y on this, you will be adding Y axis rolling off to your already rolled off measurement
b) IF you are using a baffle diffraction sim for your woofer (you don't need to as I mentioned you're good down to 44Hz by the looks of it) - then if you position the mic at tweeter axis and take the woofer BDS curve, it will have the Y offset baked in also.
Edit: so why have a "Y" axis?
1. If the designer has measured each driver on the driver axis (not a fixed axis - e.g. tweeter)
2. One is importing manufacturer curves and using them (which are all on-axis to the specific driver on the specific baffle)