I've had a play with the crossover as per peoples' kind and constructive suggestions. Apologies that it's done in Xsim, but I've yet to come to terms with Vituix.
I have to admit I'm surprised at the low efficiency of this two way: the Slapshots MTMs are 90dB, so I expected six dB less from this project, but it looks more like 75 (i.e. 15 dB less), if that makes sense.
As they'll be for a small room, it probably doesn't matter that much and I'll be driving them with a 60w RMS receiver.
As always, ideas and comments welcome, and thank you
@a4eaudio said:
...(By the way, the SPL in the measurements is not 100% accurate, we are probably really looking at 81dB but your earlier comment about it being low sensitivity is correct.)
My SPL was not calibrated precisely. I do know I was about 2dB lower than what 2.83v would have been. Below is the Dayton Audio infinite baffle SPL, which they claim is 89dB. Below that is my on-axis measurement on your baffle - because I know it was about 2dB too low I have adjusted it up 2dB in REW (meaning all of the data I posted and you are using is about 2dB lower than the plot in red). Thus, in reality you should shift all of your data up 2dB in xsim (I don't know if you can do that, but you can do it in VituixCad in about 10 seconds ) My guess is that your original thinking is correct and this should be about 6dB less sensitive than the Slapshots.
Interesting to note the rising response of the woofer from 300 to 1,000Hz when put in a cabinet - direct comparison makes it easy to see the effect of baffle step.
One thing that surprised me, and I went back and checked everything in case I made a mistake, was the total drop in SPL below 1kHz was closer to 8dB than 6dB.
Here is what I had in mind for baffle step loss on a 9-inch (218mm) wide baffle.
Note, however, that this is with 218mm radius edges, i.e., this is on a sphere. With only 18mm roundovers below, there is diffraction at the baffle edges causing a nearly 2dB hump before the roll-off. This is pretty clearly the hump from 400Hz to 2kHz that we are struggling with in the xo.
Here is the Merger Tool in VituixCAD. The yellow line is the nearfield port, the green line the nearfield woofer, and the redline the summed output of the two. Note, the red line is before applying the diffraction model above, and is similar to the line you might see on a manufacturer's infinite baffle measurement. This is a pretty flat curve, actually rising slightly below 500Hz, and gives the impression of a nice even bass response.
However, applying the diffraction model above adds the baffle step loss to the red line which is now falling below 600Hz. In fact it is falling about 6dB below 600Hz and about 8dB below the hump at 700Hz like our diffraction model indicated. So much for our nice smooth bass response - and this is why we need that big 2.5 to 3.5mH inductor to bring the curve down and then pad our tweeter down to meet it.
Now we want to merge all of that nearfield stuff with the farfield, the blue line below. (The baffle step loss is pretty obvious with the blue line, but due to the gated measurement is pretty meaningless below about 250Hz.) Here, the nearfield stuff has been brought up 18dB so that it has a pretty good overlap between 200Hz and 500Hz. In particular, it is blending the nearfield and farfield over 1 octave centered at 350Hz, indicated by the magenta lines and selected in the "Transition" box.
And the actual merge process blends the red line and the blue line to produce the white line, which is our on-axis woofer measurement. As you can see the list of all of our on- and off-axis measurements above the graph, this is actually going to create the merged response for all of our off-axis measurements 0 to 180 degrees by just clicking the Merge button and selecting the folder for VituixCAD to save them.
However, there is a little uncertainty just how precise we are below 100Hz. So maybe we can do something else, like use the Mic in a Box method.
Please don't go to any more trouble with measurements, you've helped me so much already. I'm not too worried about things below 100Hz as I'll most probably be using a sub with these.
I need to order a few parts and wire up a couple of options and see how they sound
Please don't go to any more trouble with measurements, you've helped me so much already. I'm not too worried about things below 100Hz as I'll most probably be using a sub with these.
No trouble. I bought the mic because I wanted to try the Mic in a Box method and specifically measure the speakers I took to SDC this year. But those are boxed up right now, so this gave me a chance to try it out.
Here is the best sim I came up with. I agree with dcibel, that I hope you use your own design, but hopefully seeing some different topologies and component values gives you some ideas to play with.
Excellent Vcad merger tutorial and low parts count model! You got it down to 8 parts! And since you provided the merged woofer frd's in your zip file, all I had to do when creating my 11 part model was to load them into Vcad's driver tab. It was almost as easy as setting up a simple XSim model.
David, how did you select your driver spacing (ie the 6.5" C-2-C)? Did you use the Directivity Index in VituixCAD to determine the best spacing or was it more of a visual thing.
@PWRRYD said:
David, how did you select your driver spacing (ie the 6.5" C-2-C)? Did you use the Directivity Index in VituixCAD to determine the best spacing or was it more of a visual thing.
Well, I first measured harmonic distortion at...just kidding, I used 6-1/2" C2C because that is what Geoff specified Note, since it is just a prototype at this point I did decrease and increase the y-offset and it didn't matter much and in fact 6-1/2 works well. When I get home I can change it and share the difference with a smaller and larger C2C.
The C to C and dimensions are based on our 'Slapshot' MTMs and were guessed without any calculations etc.
When I eventually get around to doing a 5.1. system, I could use the new speakers as rears but in the meantime they will be used as a stereo pair in a study.
@PWRRYD said:
...Did you use the Directivity Index in VituixCAD to determine the best spacing or was it more of a visual thing.
Craig,
I haven't looked at the details to calculate the directivity index, but I know that it is pretty insensitive to even modest changes in the xo -- Once you have the baffle width, edge treatment and driver sizes, it is mostly baked in. That is not necessarily true of in-room response, power response, etc.
So a new "rule of thumb" that some are following with respect to C2C is 1.2 to 1.4 times the wavelength at the crossover. I believe this is due to the potential for better power response. In this case the crossover is around 2.5kHz and 1.2x that length is pretty much 6-1/2 inches (165mm). This tweeter has a pretty big faceplate so the closest we can physically get is really about 145mm. This doesn't have much visible effect, so let's say we had an even smaller faceplate and could get a C2C spacing of 125mm (~4-7/8 inch). The GIF below shows the difference. Original 165mm does not have the magenta overlays. At 125mm CTC the original is the dot-dashed magenta overlays and the new Power Response is blue and the new Power Directivity Index is red. I also included the vertical polar map.
Well, I can't get the GIF to work. The dot-dashed magenta overlays above are 165mm C2C while the polar map above is for 125mm C2C. The polar map below is 165mm.
Another important Vcad chart to look at when evaluating driver spacing distance is the vertical polar chart. Use the slider to change the polar chart frequency slightly above and below the xover point and then watch to see how the shape of the vertical lobe changes as you move from a seated to a standing listening postion (zero degrees seated to +10 degrees or so standing). For my xover, very little change occurs in the 2.5kHz vertical lobe as I change the driver spacing distance from 125mm to 165mm (see charts below). But the tilt and shape of this lobe can change unexpectedly as you tweak your xover parts, so you need to keep an eye on it. You have to be very careful not to inadvertently create a partial null at the zero degree seated position.
Thanks David, I feel like I have not been wasting time when I modify my tweeter flanges for tighter vertical spacing. Also it looks less like a dog. More like Mr. Magoo.
I don't find much value in those single frequency slice circular polar plots. They provide much less information than what's available in the vertical directivity polar map, as well can lead to some bad decision making, such as placing drivers as close together as possible is better, because the polar lobe is larger. However this arrangement may also result in worse power response which is of much greater importance honestly. It's amazing how much of the audio you hear at 2.5m+ listening distance is indirect sound. Determining if the polar lobe is forward facing is already available in the phase plot and directivity plot, so there's no new information gained in these charts. My recommendation - turn focus to power & DI chart, and directivity maps.
@dcibel said:
I don't find much value in those single frequency slice circular polar plots. They provide much less information than what's available in the vertical directivity polar map, as well can lead to some bad decision making, such as placing drivers as close together as possible is better, because the polar lobe is larger. However this arrangement may also result in worse power response which is of much greater importance honestly. It's amazing how much of the audio you hear at 2.5m+ listening distance is indirect sound. Determining if the polar lobe is forward facing is already available in the phase plot and directivity plot, so there's no new information gained in these charts. My recommendation - turn focus to power & DI chart, and directivity maps.
Comments
Mystery solved, thank you gentlemen
Geoff
I've had a play with the crossover as per peoples' kind and constructive suggestions. Apologies that it's done in Xsim, but I've yet to come to terms with Vituix.
I have to admit I'm surprised at the low efficiency of this two way: the Slapshots MTMs are 90dB, so I expected six dB less from this project, but it looks more like 75 (i.e. 15 dB less), if that makes sense.
As they'll be for a small room, it probably doesn't matter that much and I'll be driving them with a 60w RMS receiver.
As always, ideas and comments welcome, and thank you
Geoff
My SPL was not calibrated precisely. I do know I was about 2dB lower than what 2.83v would have been. Below is the Dayton Audio infinite baffle SPL, which they claim is 89dB. Below that is my on-axis measurement on your baffle - because I know it was about 2dB too low I have adjusted it up 2dB in REW (meaning all of the data I posted and you are using is about 2dB lower than the plot in red). Thus, in reality you should shift all of your data up 2dB in xsim (I don't know if you can do that, but you can do it in VituixCad in about 10 seconds
) My guess is that your original thinking is correct and this should be about 6dB less sensitive than the Slapshots.
Thanks, that makes sense.
Interesting to note the rising response of the woofer from 300 to 1,000Hz when put in a cabinet - direct comparison makes it easy to see the effect of baffle step.
Geoff
Well, since you bring it up...
One thing that surprised me, and I went back and checked everything in case I made a mistake, was the total drop in SPL below 1kHz was closer to 8dB than 6dB.
Here is what I had in mind for baffle step loss on a 9-inch (218mm) wide baffle.
Note, however, that this is with 218mm radius edges, i.e., this is on a sphere. With only 18mm roundovers below, there is diffraction at the baffle edges causing a nearly 2dB hump before the roll-off. This is pretty clearly the hump from 400Hz to 2kHz that we are struggling with in the xo.
Here is the Merger Tool in VituixCAD. The yellow line is the nearfield port, the green line the nearfield woofer, and the redline the summed output of the two. Note, the red line is before applying the diffraction model above, and is similar to the line you might see on a manufacturer's infinite baffle measurement. This is a pretty flat curve, actually rising slightly below 500Hz, and gives the impression of a nice even bass response.
However, applying the diffraction model above adds the baffle step loss to the red line which is now falling below 600Hz. In fact it is falling about 6dB below 600Hz and about 8dB below the hump at 700Hz like our diffraction model indicated. So much for our nice smooth bass response - and this is why we need that big 2.5 to 3.5mH inductor to bring the curve down and then pad our tweeter down to meet it.
Now we want to merge all of that nearfield stuff with the farfield, the blue line below. (The baffle step loss is pretty obvious with the blue line, but due to the gated measurement is pretty meaningless below about 250Hz.) Here, the nearfield stuff has been brought up 18dB so that it has a pretty good overlap between 200Hz and 500Hz. In particular, it is blending the nearfield and farfield over 1 octave centered at 350Hz, indicated by the magenta lines and selected in the "Transition" box.
And the actual merge process blends the red line and the blue line to produce the white line, which is our on-axis woofer measurement. As you can see the list of all of our on- and off-axis measurements above the graph, this is actually going to create the merged response for all of our off-axis measurements 0 to 180 degrees by just clicking the Merge button and selecting the folder for VituixCAD to save them.
However, there is a little uncertainty just how precise we are below 100Hz. So maybe we can do something else, like use the Mic in a Box method.
Fascinating stuff, thank you
Please don't go to any more trouble with measurements, you've helped me so much already. I'm not too worried about things below 100Hz as I'll most probably be using a sub with these.
I need to order a few parts and wire up a couple of options and see how they sound
Cheers
Geoff
Thanks @a4eaudio - great explanation of some obscure aspects of modeling and measurement.
No trouble. I bought the mic because I wanted to try the Mic in a Box method and specifically measure the speakers I took to SDC this year. But those are boxed up right now, so this gave me a chance to try it out.
Here is the best sim I came up with. I agree with dcibel, that I hope you use your own design, but hopefully seeing some different topologies and component values gives you some ideas to play with.
Thank you!
Geoff
Nice, simple topology. Off axis looks great to me.
Excellent Vcad merger tutorial and low parts count model! You got it down to 8 parts! And since you provided the merged woofer frd's in your zip file, all I had to do when creating my 11 part model was to load them into Vcad's driver tab. It was almost as easy as setting up a simple XSim model.
David, how did you select your driver spacing (ie the 6.5" C-2-C)? Did you use the Directivity Index in VituixCAD to determine the best spacing or was it more of a visual thing.
Well, I first measured harmonic distortion at...just kidding, I used 6-1/2" C2C because that is what Geoff specified
Note, since it is just a prototype at this point I did decrease and increase the y-offset and it didn't matter much and in fact 6-1/2 works well. When I get home I can change it and share the difference with a smaller and larger C2C.
The C to C and dimensions are based on our 'Slapshot' MTMs and were guessed without any calculations etc.
When I eventually get around to doing a 5.1. system, I could use the new speakers as rears but in the meantime they will be used as a stereo pair in a study.
Geoff
Craig,
I haven't looked at the details to calculate the directivity index, but I know that it is pretty insensitive to even modest changes in the xo -- Once you have the baffle width, edge treatment and driver sizes, it is mostly baked in. That is not necessarily true of in-room response, power response, etc.
So a new "rule of thumb" that some are following with respect to C2C is 1.2 to 1.4 times the wavelength at the crossover. I believe this is due to the potential for better power response. In this case the crossover is around 2.5kHz and 1.2x that length is pretty much 6-1/2 inches (165mm). This tweeter has a pretty big faceplate so the closest we can physically get is really about 145mm. This doesn't have much visible effect, so let's say we had an even smaller faceplate and could get a C2C spacing of 125mm (~4-7/8 inch). The GIF below shows the difference. Original 165mm does not have the magenta overlays. At 125mm CTC the original is the dot-dashed magenta overlays and the new Power Response is blue and the new Power Directivity Index is red. I also included the vertical polar map.
Well, I can't get the GIF to work. The dot-dashed magenta overlays above are 165mm C2C while the polar map above is for 125mm C2C. The polar map below is 165mm.
Thank you David for such a detailed reply.
Another important Vcad chart to look at when evaluating driver spacing distance is the vertical polar chart. Use the slider to change the polar chart frequency slightly above and below the xover point and then watch to see how the shape of the vertical lobe changes as you move from a seated to a standing listening postion (zero degrees seated to +10 degrees or so standing). For my xover, very little change occurs in the 2.5kHz vertical lobe as I change the driver spacing distance from 125mm to 165mm (see charts below). But the tilt and shape of this lobe can change unexpectedly as you tweak your xover parts, so you need to keep an eye on it. You have to be very careful not to inadvertently create a partial null at the zero degree seated position.
125mm spacing distance:
145mm spacing distance:
165mm spacing distance:
Thanks David, I feel like I have not been wasting time when I modify my tweeter flanges for tighter vertical spacing. Also it looks less like a dog. More like Mr. Magoo.
I don't find much value in those single frequency slice circular polar plots. They provide much less information than what's available in the vertical directivity polar map, as well can lead to some bad decision making, such as placing drivers as close together as possible is better, because the polar lobe is larger. However this arrangement may also result in worse power response which is of much greater importance honestly. It's amazing how much of the audio you hear at 2.5m+ listening distance is indirect sound. Determining if the polar lobe is forward facing is already available in the phase plot and directivity plot, so there's no new information gained in these charts. My recommendation - turn focus to power & DI chart, and directivity maps.
Have ordered some bits and raided my parts bin, so the fun will start in a week or so: I'm going to try two options for the XO and see how I go
Thanks again to all!
Geoff
It is just a small slice. A sliver of info
Indeed, and those eyes are already staring at you in the vertical directivity map.