Most of the tutorials I've read are suggesting to blend or splice the two measurements in the low hundreds Hz range. ARTA actually gives you a suggested frequency to splice at calculated from the gating time. Averaging the two measurements would still leave the error in the measurements, just dumbed down. Splicing the two removes the error. I could have misunderstood this though, wouldn't be the first time!
Good thread, very informative. Probably too late to the party but when it comes to measurements my experience says the procedure needed of taking measurements depends on the software you are going to use for the simulation. The different software tools out there will assume differently where the mic is.
For example. Taking the measurements of a 2-way in order of tweeter then woofer then tweeter + woofer all at the same mic position will work just fine for some tools. I know Lspcad expects the measurements to be on axis of the driver being tested. It will correct the response as off axis as required based on the "y" value entered for its physical distance from the reference point (usually tweeter axis)
In the simulation when you enter the x,y,z values for the driver the software will automatically apply any corrections to the response based on the drivers physical location. So for a small 2-way the difference is small and may not even matter. If you are measuring a tall multi-way where the lowest driver is say 24" away from the tweeter the correction to the woofers response made by the software is much greater. So in this case if you kept the mic at tweeter level and took the response from the woofer it will be quite different from the response taken from the woofer directly on axis. Now using the response taken at tweeter height then entering the "y" value for the the woofer will give the software the wrong reference point and it will alter the woofers response to be much greater off-axis then it really is.
This is where you need to know what the software is doing with the physical locations. Some tools will only alter the phase driver so keeping the mic at reference point is fine. Other tools go deeper.
Sorry if this post was out of line of the thread. Its just knowledge I picked up along the way and it seemed relevant.
This is turning into a real Canucklehead fest. 
And I don't think the party is over yet.
The xo program is XSim which doesn't use x, y or z (zed) coordinates but only inputs the differences in the path lengths from the mic to the acoustic centers of the drivers. From there one can separately calculate the relative acoustic centers and then re-calculate the path length differences if you wish to change the listening distance in the simulations. A slight drawback in the software in my opinion.
And I don't think the party is over yet.The xo program is XSim which doesn't use x, y or z (zed) coordinates but only inputs the differences in the path lengths from the mic to the acoustic centers of the drivers. From there one can separately calculate the relative acoustic centers and then re-calculate the path length differences if you wish to change the listening distance in the simulations. A slight drawback in the software in my opinion.
@jReave,
All I can say or add is that using the method(s) that you outlined has yielded bang on results every time for me.😀 My only dilemma is deciding which software to use for crossover design between PCD 7 or Xsim. There is feature sets I like in each.🙂
Concerning x, y, or z coordinates, can a person use PCD to establish these then export the summed frequency response curve to a selected folder then import into Xsim using the Get file option in the curves drop down menu? I'm not convinced going to that extreme is necessary but someone else might think so. 😀
Hope you are doing well,
Rich
All I can say or add is that using the method(s) that you outlined has yielded bang on results every time for me.😀 My only dilemma is deciding which software to use for crossover design between PCD 7 or Xsim. There is feature sets I like in each.🙂
Concerning x, y, or z coordinates, can a person use PCD to establish these then export the summed frequency response curve to a selected folder then import into Xsim using the Get file option in the curves drop down menu? I'm not convinced going to that extreme is necessary but someone else might think so. 😀
Hope you are doing well,
Rich
The room actually has nothing to do with it. When you merge nearfield and gated far field responses you are creating a quasi-anechoic response, i.e. a response with no room effects. The line between high and low is determined by the baffle width, see the attachment for a full explanation.
Whatever works for you is fine, but don’t expect everyone else to agree.
I’m sorry, but that is simply balderdash. Why do you think BSC circuits were developed in the first place? Because crossovers developed without BSC compensation sounded “shouty” and uneven.
The days of separate BSC circuits are pretty much gone. With today’s crossover tools the BSC is handled in the main crossover in the majority of cases. So no resistor is needed. And if you start with an accurate quasi-anechoic response you’ll find that you won’t need an inductor anywhere near the size you indicate.
Because the probability of a tone control at a set frequency adequately compensating for BSC across its frequency spectrum is very low.
Yes indeed. Please read the attachment to understand my point of view for crossover development. I don't doubt that your method works for you. But I also know other people who just as strongly advocate using on-line calculators and then tuning by ear. But I certainly wouldn't advocate that method either.
No worries ... I know that what I do works for me.
BTW... your response splicing technique looks pretty darned close to what I get when I average measurements in REW.
The reason I got into this at all was to deal with issues of crossover inefficiency, borderline designs and needless complexity. This was moved by watching a very expensive amplifier I had just lost money repairing die again right after hooking it up to a set of home brew speakers. When I traced the crossover and did a circuit analysis I realized it had been designed lego-style in a simulator and hit an impedance low of 0.1 ohms. I ended up having to fix the amp again and design a new crossover in the process.
Not wanting this to become a common problem, I thought to help people see the problems and prevent future occurrences. Yes, a poorly designed crossover can kill an amplifier ... and don't even bother trying to tell me that a Crown DC300a is poorly designed.
I thought people would welcome it ... but I was wrong. So far this has a moderator stalking me, gotten me sin binned 3 times and now has people contriving all levels of objection to most everything I say. So, plainly, it just ain't worth the aggravation.
Good luck on your future projects, guys ... But I think it's time for me to log out and find a better way to waste my time.
I'm not sure if I fully understand what you are saying here. Explain a little more if you feel like it.
Hi jReave,
Sorry about not explaining my question very well. Let me try giving a little more detail. So, in Jeff's article finding relative acoustic offsets in PCD once a person has inputted all the necessary coordinates (x, y, z) and matched the simulated summed response to the measured response can this (black summed line) than be exported from PCD for use in other programs like Xsim? I mentioned using the get file method the curves tab but now that I think about it your method of creating unattached speakers in Xsim would work equal as well. I don't no if doing some of the processing in PCD would increase accuracy or be of any benefit at all. Hopefully, I gave enough information this time around.😀
Best Regards,
Rich
Sorry about not explaining my question very well. Let me try giving a little more detail. So, in Jeff's article finding relative acoustic offsets in PCD once a person has inputted all the necessary coordinates (x, y, z) and matched the simulated summed response to the measured response can this (black summed line) than be exported from PCD for use in other programs like Xsim? I mentioned using the get file method the curves tab but now that I think about it your method of creating unattached speakers in Xsim would work equal as well. I don't no if doing some of the processing in PCD would increase accuracy or be of any benefit at all. Hopefully, I gave enough information this time around.😀
Best Regards,
Rich
Here's another batch of measurements. I've still got some issues with the microphone input levels, it's recording too quietly. Hopefully these files will work though. I added some sound absorbing foam insulation behind the drivers for these measurements.
I wasn't sure but I thought that that was what you meant.
Ok, so you've matched the measured summed response to the simulated summed response in PCD. Therefore if you now go into XSim with the individual driver measurements, there is no difference between using the measured summed response or the summed response you just created in PCD because - they are the same.
But once you have now derived the z offsets in PCD, you can just do the math as I outlined previously and get the path length differences, ie. driver delays, for XSim that are now going to be accurate for any mic/listening distance you choose, not just at 1m. Notice that in PCD or WinPCD (and VCAD?) you are able to change the listening position whereas in XSim, the listening distance is fixed by the delay variables that you've input. And in a 3-way, changing the listening distance from 1m to 3m (as an example) will change the phase relationship between drivers and therefore the summed response in the xo regions as well. It's small difference but one that might be worth looking at, or at least knowing about at a minimum.
Are you referring to WinPCD? Its so much nicer to use the the old spreadsheet version. David Ralph's Speaker Pages - Windows Passive Crossover Designer (WinPCD)
Lol guess I should read the whole post before replying. My bad.
Lol guess I should read the whole post before replying. My bad.
Last edited:
jwilhelm,
I've gone over your new measurements and I'd say congratulations are in order. Everything I've looked at so far looks to be good. I've dropped the files into XSim and come up with delays for the mid and woofer that are exactly the same as what you came up with using the other technique. Which means you did everything correct the 1st time and my concerns with that method were unfounded.
I'd say drop the files into XSim and you should be good to go with a legitimate xo. To be clear, at 1m on the tweeter axis the mid delay is .83" and the woofer delay is 2.5". Doing the calculations, this means the relative AC offset for the mid is 13.16mm and for the woofer is 15mm. And here is the reason why my original delay values are different - the AC's turn out to be fairly different than where I physically measured them from the spec sheets, especially for the woofer. This is perhaps one of the biggest weaknesses in doing simulations without measurements - the AC's can be quite wrong using the rules of thumb concerning where they are supposed to be for the different drivers.
Now if you wanted to do your xo sims in XSim using a more correct listening distance than 1m which is where you took your measurements from, go back to my geometry equations and input the right data. For eg, if the listening position is 2.5m away, the delay for the mid would be .64" and the delay for the woofer would be 1.38". Interestingly, as you increase the listening distance, the differences in path lengths gets smaller due to the change in the angles from the ears to the center of the drivers.
I've gone over your new measurements and I'd say congratulations are in order. Everything I've looked at so far looks to be good. I've dropped the files into XSim and come up with delays for the mid and woofer that are exactly the same as what you came up with using the other technique. Which means you did everything correct the 1st time and my concerns with that method were unfounded.
I'd say drop the files into XSim and you should be good to go with a legitimate xo. To be clear, at 1m on the tweeter axis the mid delay is .83" and the woofer delay is 2.5". Doing the calculations, this means the relative AC offset for the mid is 13.16mm and for the woofer is 15mm. And here is the reason why my original delay values are different - the AC's turn out to be fairly different than where I physically measured them from the spec sheets, especially for the woofer. This is perhaps one of the biggest weaknesses in doing simulations without measurements - the AC's can be quite wrong using the rules of thumb concerning where they are supposed to be for the different drivers.
Now if you wanted to do your xo sims in XSim using a more correct listening distance than 1m which is where you took your measurements from, go back to my geometry equations and input the right data. For eg, if the listening position is 2.5m away, the delay for the mid would be .64" and the delay for the woofer would be 1.38". Interestingly, as you increase the listening distance, the differences in path lengths gets smaller due to the change in the angles from the ears to the center of the drivers.
Nice to hear I've been getting some things right! 1M is a lot closer than she'll ever be to these speakers, likely 2.5M on axis sometimes, but likely 10M away on the other side of the room with them blasting more often than anything.
When you say relative AC offset, you mean the actual driver offset to the baffle (E and F in your example)?
When you say relative AC offset, you mean the actual driver offset to the baffle (E and F in your example)?
Last edited:
Yes exactly, E and F in the diagram. The equations are much easier to work with if you drop them into a spreadsheet program btw. Can't remember if I mentioned that or not.
Years ago, I opened one up and started entering different little equations into the program. Calculating box volume is the one I probably use most frequently, either in inches or mm and so liters or cubic feet as well. Other frequent conversions and unusual volume formulas like cylinders or truncated pyramids are in there too.
I'd probably do the sims with delays equivalent to about 2.5-3m for a possible serious listening chair if she is into that. If not and she might be standing and moving around on the other side of the room most of the time, you may want to look at the off-axis responses at larger distances in another xo program like PCD or WinPCD. Up to you.
A word regarding target xo points and slopes. They are intended as starting points and are not locked in stone. Even when I try to pinpoint the xo frequencies, I try to state it as a range of frequencies and not just a single frequency. Sometimes you just have to go where the drivers want to go. Design always involves compromises. In the end, your ear will tell you what works or what doesn't and when things don't, then you may have to change a driver or 2 if necessary or go right back to the drawing board sometimes. When I'm in the design phase and am picking potential drivers, if I can I'll try to pick more than 1 of each driver and then when I run the sims I can see if some of them work better together than others.
Years ago, I opened one up and started entering different little equations into the program. Calculating box volume is the one I probably use most frequently, either in inches or mm and so liters or cubic feet as well. Other frequent conversions and unusual volume formulas like cylinders or truncated pyramids are in there too.
I'd probably do the sims with delays equivalent to about 2.5-3m for a possible serious listening chair if she is into that. If not and she might be standing and moving around on the other side of the room most of the time, you may want to look at the off-axis responses at larger distances in another xo program like PCD or WinPCD. Up to you.
A word regarding target xo points and slopes. They are intended as starting points and are not locked in stone. Even when I try to pinpoint the xo frequencies, I try to state it as a range of frequencies and not just a single frequency. Sometimes you just have to go where the drivers want to go. Design always involves compromises. In the end, your ear will tell you what works or what doesn't and when things don't, then you may have to change a driver or 2 if necessary or go right back to the drawing board sometimes. When I'm in the design phase and am picking potential drivers, if I can I'll try to pick more than 1 of each driver and then when I run the sims I can see if some of them work better together than others.
I've ordered another microphone, the UMM-6 I've been using isn't working out the best. They are sold right out at parts Express, but I managed to buy a EMM-6 from Solen but Canada Post seems to be really taking their time delivering it. In the meantime I put together a simple crossover design with Douglas Blake's help to get the speakers running initially. I plan to do a second design crossover with proper acoustic alignment and baffle step compensation, ect after they are finished and assembled and I can get some better measurements.
