VituixCAD

That is probable, but as I said, Olive's original equation 9 with slope-sensitive SM_PIR (calculated as PEARSON^2 or RSQ in Excel which prefers slope=-1) and slope-sensitive NBD_ON (which prefers slope=0) are able to detect totally wrong tilts so it's valid for evaluation of sound balance and coloration - at least compared to modifications where slope-sensitivity is totally removed from SM_PIR without adding SL_ON or/and SL_LW.
Yeah, I know.
I think optimization by the score could be a cunning approach but the slope dependency of the original equation is very very problematic.
And I personally do not like the concept because it's kind of a redeem of a bodge job, I mean, the only imaginable reason I need to use optimization by the score is the faulty directivity of my design.
I am an idealist, sort of, so I would take the only rightest approach I think, simply flat on-axis (or maybe slight tilt or roll-off, depending on the taste) and a proper directivity control, which would result in a nice PIR and more importantly good perception.
So I am not a big fan of optimization by the score but I hope optimization by the modified equation may be sufficient (depends on constants, obviously) and results in preferable sound for VituixCAD folks as well.

I hope he meant that PR is not inclusive enough for designing. Original PR with eq 9 is easy and quite adequate for fine tuning sound balance and coloration with XO parameters, but designer should ensure that overall balance and other properties such as linearity/compression, multi-source effects i.e. diffraction, total directivity, smoothness of directivity with acoustical design etc. are taken care of.
I think he might have been fed up as well with inappropriate use of PR by some PR fanboy.
 

kimmosto

Member
2005-10-06 10:37 am
slope dependency of the original equation is very very problematic.
I totally agree this though I tried to defense original equation 9 earlier. Effect of slope-dependence preferring slope=-1 is very low compared to effect of smoothness so optimizing for example active gain of tweeter with eq 9 can lead to quite bad overall sound balance - especially with near field monitors weighting ON and LW. That kinda proves that original equation 9 is not valid also for scoring. It just "happens" to give decent correlation in Olive's study, with selected speakers in selected listening environment and setup; distance, angle etc.
So I am not a big fan of optimization by the score but I hope optimization by the modified equation may be sufficient (depends on constants, obviously) and results in preferable sound for VituixCAD folks as well.
I optimize sound balance using custom equation with slope target for ON or LW and at least monitor slope of PIR to avoid too bright/harsh sound if optimized parameters affect to slope. The latest build has also target for PIR slope which could be the only slope target in some...many cases. This is very good method for fine tuning XO. Designer just need to know what are proper slopes for selected acoustical concept, target environment and typical listening setup.
 
If you don't fully comprehend the settings in Preference Ratings, are you insinuating to just leave it on the default settings (in v 2.0.88.4)?
i.e. don't use equation 9 or 10?

Is there any way (a link?) to learn/understand the Preference Ratings settings?
I understand the meaning of the acronyms, but am totally lost with the significance of each, and how to alter the settings.
As a newbie, would it be best to just leave them on your default settings?
 
I totally agree this though I tried to defense original equation 9 earlier. Effect of slope-dependence preferring slope=-1 is very low compared to effect of smoothness so optimizing for example active gain of tweeter with eq 9 can lead to quite bad overall sound balance - especially with near field monitors weighting ON and LW. That kinda proves that original equation 9 is not valid also for scoring. It just "happens" to give decent correlation in Olive's study, with selected speakers in selected listening environment and setup; distance, angle etc.

I optimize sound balance using custom equation with slope target for ON or LW and at least monitor slope of PIR to avoid too bright/harsh sound if optimized parameters affect to slope. The latest build has also target for PIR slope which could be the only slope target in some...many cases. This is very good method for fine tuning XO. Designer just need to know what are proper slopes for selected acoustical concept, target environment and typical listening setup.
If the basic design is fixed so the inherence of each driver on the loudspeaker can not be changed, score optimization (with optimized constants for individuals) must be a very strong tool.
Realist me likes the score optimization but idealist me rejects it and forced me back to the drawing board. ha-ha

For me, "proper slopes for selected acoustical concept, target environment and typical listening setup" (and preference of individuals), were the hardest part but bunch of measurements of products on the VituixCAD helped a lot.


By the way, can I ask how to add measurements such as "****_hor_deg15 ver_deg15.txt" (I mean horizontally 15deg and vertically 15deg )?
I know I can manually edit hor and ver in the Drivers tab so that I can get a more precise off-axis simulation but I could not find the right filename format for automatic recognition.

And I'd like to add 2.5deg step or 7.5deg step (some turntable has 7.5deg increment) measurements for convenience but I could not input decimal point.
Could I ask to add decimal point input?
 
Adding an extra decimal character to auto-generated filenames is a big step. Character recognition within the filename of real numbers (vs integers) can be a headache. It is not as easy as it sounds. When I managed an engineering software team, we struggled with this because our files were used across platforms, and Unix/Linux/Windows all handle thigs a bit different from each other. Backward compatibility was also an issue. Of course Kimmo is very clever and he may have some rabbit to pull out of his hat. :cool:
 
  • Like
Reactions: 1 user

kimmosto

Member
2005-10-06 10:37 am
If you don't fully comprehend the settings in Preference Ratings, are you insinuating to just leave it on the default settings (in v 2.0.88.4)?
i.e. don't use equation 9 or 10?

Is there any way (a link?) to learn/understand the Preference Ratings settings?
I understand the meaning of the acronyms, but am totally lost with the significance of each, and how to alter the settings.
As a newbie, would it be best to just leave them on your default settings?
Very difficult to give exact instructions due to wide variety of possible speaker concepts, listening environments and positioning, but here are few simple tips:

1. Do not optimize crossover or anything else in construction with equation 9 and 10. Use custom equation instead.

2. Decent weighting factors to start with:
1659682019077.png

This is quite universal and flexible for both near field monitors and longer listening distances, but not necessarily the best for all cases.

3. Increase weight of SM_PIR to 2 if typical listening distance is longer than near field.

4. Increase weight of SM_ON to 2 if listening distance is near field only.

5. Do not check SL_ON and weight too much SM_ON and NBD_ON if any of the drivers has bad on-axis response while LW and PIR are possible to get smooth. For example typical coaxial drivers (hifi and PA) have too bad on-axis for any optimizing with it. Keep SL_LW or SL_PIR or both checked, drop SM_ON and NBD_ON to half...zero to optimize smoothness with SM_LW, SM_ER and SM_SP alone. PIR contains LW 22%, ER 44% and SP 44% so you can increase weight of all three by increasing SM_PIR alone for example to 2...3.

6. Some concepts in some environments and with positioning may sound too bright and harsh with flat horizontal on-axis. For example speakers with strong directivity or line arrays at long distance. You can improve sound balance by optimizing LW slope=-0.5...-0.7... or PIR slope -1.5...-1.7...; making sure that on-axis tilts down enough.

7. Keep LFX and LFQ unchecked or set factors to zero or check 'With sub'. LF response cannot be optimized with typical passive XO, and active XO is best to optimize manually. LFX and LFQ and not very smart for optimizing.

After optimizing it's easy to peek PR also with equation 9, and test how result changes when some XO components or parameters are changed. Typically result gets worse right away.
 
Last edited:
  • Thank You
  • Like
Reactions: 5 users
Very difficult to give exact instructions due to wide variety of possible speaker concepts, listening environments and positioning, but here are few simple tips:

1. Do not optimize crossover or anything else in construction with equation 9 and 10. Use custom equation instead.

2. Decent weighting factors to start with:
View attachment 1078273
This is quite universal and flexible for both near field monitors and longer listening distances, but not necessarily the best for all cases.

3. Increase weight of SM_PIR to 2 if typical listening distance is longer than near field.

4. Increase weight of SM_ON to 2 if listening distance is near field only.

5. Do not check SL_ON and weight too much SM_ON and NBD_ON if any of the drivers has bad on-axis response while LW and PIR are possible to get smooth. For example typical coaxial drivers (hifi and PA) have too bad on-axis for any optimizing with it. Keep SL_LW or SL_PIR or both checked, drop SM_ON and NBD_ON to half...zero to optimize smoothness with SM_LW, SM_ER and SM_SP alone. PIR contains LW 22%, ER 44% and SP 44% so you can increase weight of all three by increasing SM_PIR alone for example to 2...3.

6. Some concepts in some environments and with positioning may sound too bright and harsh with flat horizontal on-axis. For example speakers with strong directivity or line arrays at long distance. You can improve sound balance by optimizing LW slope=-0.5...-0.7... or PIR slope -1.5...-1.7...; making sure that on-axis tilts down enough.

7. Keep LFX and LFQ unchecked or set factors to zero or check 'With sub'. LF response cannot be optimized with typical passive XO, and active XO is best to optimize manually. LFX and LFQ and not very smart for optimizing.

After optimizing it's easy to peek PR also with equation 9, and test how result changes when some XO components or parameters are changed. Typically result gets worse right away.
@kimmosto -That is absolutely awesome.

This is my first design using VituixCAD, and I am really struggling (especially with Preference Ratings).
You have designed a brilliant but very powerful program, and simple DIYers like myself, really struggle with some of the more complex, but very important features.
However, I think I speak for many of the DIYers, we are willing to persist with trying to understand as much of the program as possible, which hopefully, will enable us to produce quality crossover designs.
Thanks again for producing such a program, and thanks for the patience in continuing to provide seemly basic information to newbies like me. :love:
 

kimmosto

Member
2005-10-06 10:37 am
If the basic design is fixed so the inherence of each driver on the loudspeaker can not be changed, score optimization (with optimized constants for individuals) must be a very strong tool.
Realist me likes the score optimization but idealist me rejects it and forced me back to the drawing board. ha-ha
Optimizing based on some math is balancing sound with less random subjective variation. Balancing with XO is needed no matter how good acoustic design is and how idealist, experienced and smart designer is in his own opinion. Every multi-way design has many possible combinations for XO parameters, and it's quite difficult to decide how to weight each response to 3D space and try to apply experiences to different speaker concepts. Using documented parameters is good method to control that task and use experiences for next projects in a controlled manner. In addition, quality and strength of directivity is not global fixed constant, and it's not enough to make excellent speaker. For example many DIYers think that Kef's coaxial speakers are something special because they have excellent directivity features. But actually those are just average and no problem to design better with worse directivity, vertical lobing etc.

I've known one local idealist for almost two decades. He planned speakers for many years. Endless discussion on forums and by e-mails, selecting the best concept, close to best drivers, flawless shape for the concept and so on. But sound was total crap in his room. Idealism and perfectionism limited to speaker alone is quite worthless without adequate experience and understanding what features in whole system make good sound reproduction.
 
  • Like
Reactions: 2 users
I've known one local idealist for almost two decades. He planned speakers for many years. Endless discussion on forums and by e-mails, selecting the best concept, close to best drivers, flawless shape for the concept and so on. But sound was total crap in his room. Idealism and perfectionism limited to speaker alone is quite worthless without adequate experience and understanding what features in whole system make good sound reproduction.
If you're serious about the quality of sound you want to hear, I believe you have to put as much work into your listening room as you do your speaker design.
Of course it goes without saying that the rest of the audio equipment must also match the quality of the speakers.