As an example, here's 1/3 on the tweeter, 1/6 on the mid, and 1/12 on the woofer
versus zero smoothing:
versus zero smoothing:
It also occured to me just now:
Zero out the driver position in the Y axis. You used a single microphone location for all the measurements I think? If that is true, the Y axis offset is already in your measurements.
Zero out the driver position in the Y axis. You used a single microphone location for all the measurements I think? If that is true, the Y axis offset is already in your measurements.
That polar shows that you are using bad off axis data, so I would ignore anything related to directivity shown here..
Just follow the instructions for VCad2 given by Kimmo:
Perform a dual channel measurement (do not change volume levels between measurements)
if you want a reliable power response, measure each driver on its axis, at least 0-90° every 10-15°
Export the FR in REW according to kimmo's instructions
Design your baffle in Diffraction tool, export diffraction FR at 10m and 1m.
Merge NF and FF responses
Load the all the FRs in main program and set the driver coordinates (Tw is always 0,0,0 and so Zin your case. Yvaries for mid and wf) and start with a template like this: first set filter type and slope and then activate all the PEQ you desire (smooth all the peaks and, possibly, ignore all the dips).
In menu View, select Preference rating > press the button Full space and check the box Show target to get the desired target response: in the picture below the speaker is optimised to get a good PIR
Do I need to enter driver coordinates? And what is modelled if coordinates are not entered?
Last edited:
@shadowplay62
VCAD beats me every time.
So, I measured in REW off axis response of each driver and loaded into VCAD. I then managed to build crossovers and filters.
Then, in 'diffraction tool' I set the shape of the baffle and add one driver at a time and then export the frequency response one dirver at a time? then murge that with what I measured? but I measured in that very baffle already. Why do I need this step?
I cannot seem to add more than one driver to the baffle in the 'diffraction tool'. Is it one at a time only?
NF and FF merge: of what? FF of mid and tweeter and NF of woofers? In my sim above I used woofers NF but did not do any merge operation, just loaded the FRD on axis into the woofer driver.
VCAD beats me every time.
So, I measured in REW off axis response of each driver and loaded into VCAD. I then managed to build crossovers and filters.
Then, in 'diffraction tool' I set the shape of the baffle and add one driver at a time and then export the frequency response one dirver at a time? then murge that with what I measured? but I measured in that very baffle already. Why do I need this step?
I cannot seem to add more than one driver to the baffle in the 'diffraction tool'. Is it one at a time only?
NF and FF merge: of what? FF of mid and tweeter and NF of woofers? In my sim above I used woofers NF but did not do any merge operation, just loaded the FRD on axis into the woofer driver.
Last edited:
Why would you use a baffle diffraction tool at this point? Your measurements already include any diffraction. Maybe there's something the software needs, but it doesn't seem logical to me.
I see no reason for using the woofer near field either. But, I'm not up on this software.
I see no reason for using the woofer near field either. But, I'm not up on this software.
You should simulate baffle before making the crossover.
Baffle simulation allows to
- merge nearfield and farfield responses correctly and
- simulate the speaker at any distance different from 1 m (usually 3 m or so)
Yes.
Of any driver that needs frequency response data below the one allowed by gated measurement. Usually woofer and (depending on your reflection free measurement time) midrage.
Tipically, you measure FF at 1m and NF at 5-10mm, then the NF FR does not include the effect of the baffle step loss for Woofer and Midrange.
Prepare all the FRs for W and M and load them in the Merger tool
Once you have all the responses, load them in the main program, set the coordinates for the drivers and calculate your XO.
PS
In the Room tab of the main program it's possible to adjust the listening distance: the program is able to simulate the XO for that distance even if you measured at 1m.
Alma, you probably have measurements that are good enough to create the next iteration. You can play around with more complicated stuff in the version after that.
Suggestions for your first VCAD sim:
1. It looks like you aimed for a sloped on axis response. Normally, one aims for flat here. If you've heard about flat being bad, that's in room. Take a look at the blue power response line in the middle left window of the six pack. Note that it falls away with an increasing slope as the frequency goes up. If you like that, cool. But if you don't know what you like, I'd suggest you aim to make that a linear or flat downward slope. (the downward tilting line is flatter). So what I would do for that, is lift the midrange and tweeter levels a bit, with an aim for flat on axis.
2. There is a bit of directivity narrowing around 9khz. This is the bump I mentioned and recommended leaving alone, and just listen off axis. If you bump up the tweeter, and allow the 9k hump to exist, the dip in the power response should fill in. So your in room will smooth out. Then you tilt them a bit off axis, so you're not listening to the bump. Viola, listening axis is nice and flat and in room is nicely linear and downward sloping.
3. Your LT numbers look wonky. The specifics don't actually matter, but.... It looks like you're trying to shape the bottom of the response curve, which I would suggest you don't do. Measurements down this low are basically garbage unless you do them near field. Even in a big *** church. I guessed at the natural F pole and Q of your enclosure when I built my model. It's probably pretty close.
So F0 of 72hz (give or take a little) and a Q0 of .8 (give or take). When I plug in an Fc of 30 or 40hz and a Qc of .7, the resultant curves look roughly correct to me in sim. So I would recommend you use something close to the F0 and Q0 I used, then change the Fc and Qc to build a shape you want. That way you're not reacting to dips and bounce cancellations and whatnot that are artifacts of your measurements rather than being real.
Suggestions for your first VCAD sim:
1. It looks like you aimed for a sloped on axis response. Normally, one aims for flat here. If you've heard about flat being bad, that's in room. Take a look at the blue power response line in the middle left window of the six pack. Note that it falls away with an increasing slope as the frequency goes up. If you like that, cool. But if you don't know what you like, I'd suggest you aim to make that a linear or flat downward slope. (the downward tilting line is flatter). So what I would do for that, is lift the midrange and tweeter levels a bit, with an aim for flat on axis.
2. There is a bit of directivity narrowing around 9khz. This is the bump I mentioned and recommended leaving alone, and just listen off axis. If you bump up the tweeter, and allow the 9k hump to exist, the dip in the power response should fill in. So your in room will smooth out. Then you tilt them a bit off axis, so you're not listening to the bump. Viola, listening axis is nice and flat and in room is nicely linear and downward sloping.
3. Your LT numbers look wonky. The specifics don't actually matter, but.... It looks like you're trying to shape the bottom of the response curve, which I would suggest you don't do. Measurements down this low are basically garbage unless you do them near field. Even in a big *** church. I guessed at the natural F pole and Q of your enclosure when I built my model. It's probably pretty close.
So F0 of 72hz (give or take a little) and a Q0 of .8 (give or take). When I plug in an Fc of 30 or 40hz and a Qc of .7, the resultant curves look roughly correct to me in sim. So I would recommend you use something close to the F0 and Q0 I used, then change the Fc and Qc to build a shape you want. That way you're not reacting to dips and bounce cancellations and whatnot that are artifacts of your measurements rather than being real.
@Busdriver02
thank you for these inputs. Some feedback:
I do not know yet if I prefer sloped or flat - I think I will like a slight slope, but I have not verified that, so I will start with flat. I understand that if I set up for the top/green curve to be flat, the in-room curve (blue) will be slightly sloping and I should try to fit it into the target area (shaded blue area). Right? I will try to fit into that target and then see if I like that or prefer anything different to that.
OK, will see how it works vs trying to EQ these dips/peaks in DI out. There are two kinks in the DI - the 9 to 12kHz and then there is the ripple area between 500Hz and 2kHz. How can I improve the that region?
What I did n this sim is just scrolled on each of the four parameters randomly trying to make the curve as flat as I could and as low extension as I could (but not lower than 30Hz). I have not tried these LT version in the actual speaker. Instead I have default Q and 70Hz -> 30Hz frequencies using the Hypex Assym Shelf. I then measure the speakers one by one from listening position and EQ three room peaks out with narrow notch filters, and then check with both playing. This brings huge improvement to bass. I am absolutely astonished how deep these relatively compact speakers play.
Yes, I am aware of the need to ignore the peaks/dips of the in-room measurements below 200Hz as these are room modes. I instead look at the general fall off of the response and push it up using LT.
I discovered yesterday that I can enter the room dimensions in REW and it will give me the frequencies for all the main room modes and even show the impact on the response depending on listening position. It was amazing to see how accurate that was to what I saw in measurements. This should help me to distinguish peaks that are r00om-related when analyzing the in-room measurement and then apply EQ notches.
thank you for these inputs. Some feedback:
I do not know yet if I prefer sloped or flat - I think I will like a slight slope, but I have not verified that, so I will start with flat. I understand that if I set up for the top/green curve to be flat, the in-room curve (blue) will be slightly sloping and I should try to fit it into the target area (shaded blue area). Right? I will try to fit into that target and then see if I like that or prefer anything different to that.
OK, will see how it works vs trying to EQ these dips/peaks in DI out. There are two kinks in the DI - the 9 to 12kHz and then there is the ripple area between 500Hz and 2kHz. How can I improve the that region?
What I did n this sim is just scrolled on each of the four parameters randomly trying to make the curve as flat as I could and as low extension as I could (but not lower than 30Hz). I have not tried these LT version in the actual speaker. Instead I have default Q and 70Hz -> 30Hz frequencies using the Hypex Assym Shelf. I then measure the speakers one by one from listening position and EQ three room peaks out with narrow notch filters, and then check with both playing. This brings huge improvement to bass. I am absolutely astonished how deep these relatively compact speakers play.
Yes, I am aware of the need to ignore the peaks/dips of the in-room measurements below 200Hz as these are room modes. I instead look at the general fall off of the response and push it up using LT.
I discovered yesterday that I can enter the room dimensions in REW and it will give me the frequencies for all the main room modes and even show the impact on the response depending on listening position. It was amazing to see how accurate that was to what I saw in measurements. This should help me to distinguish peaks that are r00om-related when analyzing the in-room measurement and then apply EQ notches.
Alma,
1. Basically yes. In your design, the power response dropps above 3khz and falls below your shaded blue region. The amount of downward tilt is a subjective thing. Take a look here: Erin's Audio Corner at the KEF Concerto Meta and the KEF R3 Meta speakers. The Concerto has a much more downward sloped in room, vs the R3. In the case of the Concerto, they also intentionally padded down the tweeter to make sure the power response was smooth.
2. You can't change the DI at this point, you have to work with/around it. It's fixed. There is also bound to be measurement artifacts in there, so don't go crazy looking at minute detail.
However, You can tune for any sort of compromise you want. You can also tune it for listening on any axis you want. If you look above the six pack graphs in VCAD there is a reference axis window that you can change. Play with that a bit to get a feel for what different listening axis might look like.
So as an example:
I'm still using the measurements I mentioned before (Church/Base2/Church)
This first one has a -0.5 db per octave power response slope, which produces a mostly flat on axis response, and I would listen 10 degrees off axis or so (second picture) as a starting point
This one is setup to produce a -1.0db per octave power response tilt, which gives a very slight downward tilt on axis. I would listen to this on axis.
3. Cool, just checking you have a plan.
1. Basically yes. In your design, the power response dropps above 3khz and falls below your shaded blue region. The amount of downward tilt is a subjective thing. Take a look here: Erin's Audio Corner at the KEF Concerto Meta and the KEF R3 Meta speakers. The Concerto has a much more downward sloped in room, vs the R3. In the case of the Concerto, they also intentionally padded down the tweeter to make sure the power response was smooth.
2. You can't change the DI at this point, you have to work with/around it. It's fixed. There is also bound to be measurement artifacts in there, so don't go crazy looking at minute detail.
However, You can tune for any sort of compromise you want. You can also tune it for listening on any axis you want. If you look above the six pack graphs in VCAD there is a reference axis window that you can change. Play with that a bit to get a feel for what different listening axis might look like.
So as an example:
I'm still using the measurements I mentioned before (Church/Base2/Church)
This first one has a -0.5 db per octave power response slope, which produces a mostly flat on axis response, and I would listen 10 degrees off axis or so (second picture) as a starting point
This one is setup to produce a -1.0db per octave power response tilt, which gives a very slight downward tilt on axis. I would listen to this on axis.
3. Cool, just checking you have a plan.
Last edited: