First VituixCAD Design - Please check

[rpb]

looks like i opened a can of worms... ;)

i'll keep the components as is prior to the listening distance update but will take this into account with the next dual 7"/1-1/8 build coming soon...

Keep in mind what you are building! It's a surround speaker. Build a x-over, and try it out. Odds are pretty good that you will like it just fine after a couple of voicing tweaks. I would again suggest getting set up to measure. It's fun, and will eliminate some of the guessing about the sim.

[DeZZar]

I meant no offence, we only have different definition of "optimized". With the information available, I would describe the above response as flat, but not optimized, where optimized is finding that balance of on and off axis result. I have no intention of suggesting an alternative crossover, I encourage the OP to keep learning the tools and work towards a design of his own.

nah none taken mate, I perhaps used the word optimized too liberally as I agree - its primarily on axis flat - but without real measurements its a simulation on top of a simulated data and so I just feel, without any obvious issues showing up, this is about as far as I can take it, in a reasonable timeframe, with the information available.

however, I do generally hope for a learning experience in every interaction here and seeing a contribution from someone else may provide that opportunity! But I appreciate we're all busy.

[dcibel]

DeZZar I meant no offence, we only have different definition of "optimized". With the information available, I would describe the above response as flat, but not optimized, where optimized is finding that balance of on and off axis result. I have no intention of suggesting an alternative crossover, I encourage the OP to keep learning the tools and work towards a design of his own.

rpb Your opinion is quite surprising to me. Are you suggesting that the research completed by the likes of Olive, Toole, Klippel and others that ultimately resulted in a measurement and data representation standard is all a waste of time and effort? ASR and EAC using Klippel NFS to evaluate speaker performance is a useless metric? From my own design efforts using these tools, what you are dismissing as totally useless has become my #1 goal. To each his own.

[fpitas]

Don't feel bad, this stuff is complicated. Sometimes it seems like a miracle that any speaker sounds good ;)

[zinger084]

looks like i opened a can of worms... ;)

i'll keep the components as is prior to the listening distance update but will take this into account with the next dual 7"/1-1/8 build coming soon...

[rpb]

The sim you provided is more than the OP asked for, so mission accomplished. My opinion is that looking at all the power, and DI graphs may make a designer feel good about his design, but is totally useless in the real world of DIY speakers.

[DeZZar]

Yes of course - I'm aware of the additional considerations and optimizations - its just that I am now at the limit of how much time I can dedicate to steering the OP in the right direction. Feel free to contribute a revision of your own.

[dcibel]

Accurate depiction of Power & DI response must be done in the far field, VituixCAD document states "do not set listening distance to <2m", which is for accuracy of these results, and follows that CTA-2034-A standard. These charts are quite a bit more real than old PCD spreadsheet, which used strictly single axis of measurement and piston model calculations, off-axis insight in PCD is truly only crossover interaction and nothing more.

[dcibel]

Updated XO optimised for 2.5m listening distance

Optimized for on-axis perhaps, don't be afraid of the actual optimizer in the software ;)

I usually optimize to listening window and in-room response. Listening window at -0.2dB/oct from 100Hz to 10kHz, and in-room response at -0.8 to -1dB/oct from 100Hz to 10kHz. Ctrl Click on the left and right side of the SPL and Power & DI chart to set the target slope. Select components to optimize, right click and select "optimize on". Open the optimizer, select listening window ant in-room response at 50/50 weight, click the optimize button and watch the magic happen!

Of course, it works with incredible accuracy and reliability with real measured data, this simulation is mostly a bunch of guesswork, estimated z offset, and "piston model" off-axis response determined by the diffraction tool lacks a lot of information so is unlikely to be incredibly accurate above 1kHz where it matters the most.

[DeZZar]

I'm not taking sides...

haha...no sides to take mate - I was only making an observation and asking for viewpoints.

Then I then assume that it can also show how the same xo summation would look at 2m, if nothing in the x-over is changed..

yeah, that's exactly what it's doing. Same as I assume as Bagby's spreadsheet does as well as PCD.

[rpb]

A difference yes, not a problem as such.

Given its such a variable how do you govern any existing design as being adequate for your listening distance? How do you know what distance they were designed for if this is baked into the crossover?

How do you correlate the simulation with your measurements if you simulate at say 2.5m, but are never (generally for development purposes) going to measure at anymore than 1m. All crossover iteration and design and finalization is generally happening with quasi anechoic measurements taken no more than a meter away....how do you get these results to correlate - if your sitting at your computer looking at a simulation at 2.5m and comparing it to 1m actual measurements....

Just curious what the views are actually - not challenging the theories - I get that.

In this example if the crossover was fine tuned for 2.5m, they would be unbearably forward in the midrange any closer under any other application.

I'm not taking sides, but just wanted to comment on how I assume the software is computing (predicting) things. I assume that the sim for 1m shows the summation that results at 1m. (. So far, so good???..)

Then I then assume that it can also show how the same xo summation would look at 2m, if nothing in the x-over is changed.. (Note, the relative Z offset is going to change slightly. More so with wide driver spacing.)

It may be that a small 4" change in listening height would have just as much of an effect., and possibly correct the 1m vs 2m difference.

I think that ultimately, you just want a very stable response on the listening axis. One that doesn't suddenly change if you slouch, or sit up very straight.

For my trial and error x-overs, I measure at somewhat random distances, and also look at the effect of measuring from a few inches higher, or lower, to help me decide if x-over changes might improve the results. I also listen at various stages in the x-over design when I start getting in the ballpark.

[DeZZar]

Updated XO optimised for 2.5m listening distance

[DeZZar]

I have to disagree that it is not a problem. Difference = problem, one result is correct, one is less so.

A 1.5db dip is what I was referring to as something I don't consider a problem in this instance, it can easily be accepted as equally good.

Ok, I can understand this becomes more and more important the greater the array of drivers, the further apart they are etc - makes sense.

[dcibel]

A difference yes, not a problem as such.

Given its such a variable how do you govern any existing design as being adequate for your listening distance? How do you know what distance they were designed for if this is baked into the crossover?

How do you correlate the simulation with your measurements if you simulate at say 2.5m, but are never (generally for development purposes) going to measure at anymore than 1m. All crossover iteration and design and finalization is generally happening with quasi anechoic measurements taken no more than a meter away....how do you get these results to correlate - if your sitting at your computer looking at a simulation at 2.5m and comparing it to 1m actual measurements....

Just curious what the views are actually - not challenging the theories - I get that.

In this example if the crossover was fine tuned for 2.5m, they would be unbearably forward in the midrange any closer under any other application.

I have to disagree that it is not a problem. Difference = problem, one result is correct, one is less so. Interaction of drivers in a large tower speaker at 1m will be rather different than at 2.5m due to simply the distance and angle of each driver relative to the listener. Only speakers intended for near listening as regular use should be designed at short listening distance. To reiterate my original statement, listening distance should be set at typical listening distance. You'll find that once you are at >2m, change in further distance has little effect on result, unless your speaker happens to be enormous. Don't design at 1m distance if you intend to listen at >2m.

Measurement distance and listening distance for design do not have to be the same, in fact they usually aren't. With VituixCAD the separation of the two is easily done. 1m distance is not a requirement for any measurement, rather a distance of at least 3x the baffle width so that the mic is adequately in the far field to capture diffraction effects of the baffle. The correct process for VituixCAD is to capture full 3D balloon data for each driver using dual channel measurement system. With the complete set of measurements, simulation at correct listening distance >2m is simply done by using the driver x,y offset values and interpolation of the measured off-axis data. Using the driver coordinates, relative distance to mic as well as driver angle is easily calculated. This is the only correlation that is needed, measurement at further distance will not be different from 1m, provided 1m is adequately in the far field for the size of speaker (ex big 15" driver may be on a 22" baffle, in which case 1m may still be too close. A small bookshelf may be okay to measure at only 600mm by comparison).

For more information, there is the measurement guides for VituixCAD for ARTA, REW, or SoundEasy. The CTA-2034-A standard is a good reference as well.

For single axis simulations, on-axis only like Xsim, listening distance becomes irrelevant, x,y,z for each driver should be zero with only delay included in the driver to account for acoustic offset. There is no power response, DI, off-axis detail to be had with this limited information.

[DeZZar]

Simulating at 1m distance can create problems at actual listening distance as Dezzar has found.

A difference yes, not a problem as such.

Given its such a variable how do you govern any existing design as being adequate for your listening distance? How do you know what distance they were designed for if this is baked into the crossover?

How do you correlate the simulation with your measurements if you simulate at say 2.5m, but are never (generally for development purposes) going to measure at anymore than 1m. All crossover iteration and design and finalization is generally happening with quasi anechoic measurements taken no more than a meter away....how do you get these results to correlate - if your sitting at your computer looking at a simulation at 2.5m and comparing it to 1m actual measurements....

Just curious what the views are actually - not challenging the theories - I get that.

In this example if the crossover was fine tuned for 2.5m, they would be unbearably forward in the midrange any closer under any other application.