When designing speakers, by which I mean, cabinets and crossovers etc as opposed to drive units, which should we give most credence to and why?
Each have their pitfalls:
Simulation may be misleading due to incorrect T/S model data supplied by the vendors.
Measurements deliver numbers - but these must be set into context of listening experience. Where are the relevant benchmark numbers everybody may agree to?
Listening experience does not deliver results that may be reproduced by third parties and are highly prone to personal bias.
So my verdict is
none of them alone
but a mix of all of them.
Simulation may be misleading due to incorrect T/S model data supplied by the vendors.
Measurements deliver numbers - but these must be set into context of listening experience. Where are the relevant benchmark numbers everybody may agree to?
Listening experience does not deliver results that may be reproduced by third parties and are highly prone to personal bias.
So my verdict is
none of them alone
but a mix of all of them.
I'd say Experience 😀 Ears are the ultimate judge, but easily affected by coffee and other non hearing related stuff unless there experience to guide. Measurements and simulations help validate each other unless the measurements are off. It requires some experience to get confidence to measurements. Simulation is misleading without proper measurements, or at least knowledge about the measurement data must be there.
At the beginning it is hard to trust any of them. When ear is experienced enough to not give doubt the only thing left is to measure and simulate until all three agree. Eventually one will trust the ear and the rest becomes repetition. Coffee break over 🙂
edit. bucks bunny was faster with the essence 🙂
At the beginning it is hard to trust any of them. When ear is experienced enough to not give doubt the only thing left is to measure and simulate until all three agree. Eventually one will trust the ear and the rest becomes repetition. Coffee break over 🙂
edit. bucks bunny was faster with the essence 🙂
Last edited:
For me, it's not a matter of ranking their credence....
it's a matter of what is best to rely on, when.
When I first get a build idea, I run sims until i think i have credible plans for a prototype.
After building the prototype, if measurements are not even close to the sim, i return to sims to try to figure out why.
If prototype measurements are in the ballpark of the sim, I'm done sims. And then use measurements to modify, refine, and tune the prototype.
It may take several prototype modifications or re-builds, if it's a more complicated design requiring some trial and error. (like a synergy for example).
But all decisions are made from measurements alone in this stage, until a final build is tuned and running.
Then it's time for first listen, and ears totally decide if the entire effort worked or not.
If ears say something is wrong, either on or off axis, i return to measurements to find out why, and how adjustments such as xover freq/order, etc, etc, help or hurt.
When ears like what they hear, I'm done, and Happy 🙂
Ears (and brain) are not very specific in terms of what they perceive and may be biased by the source material and other factors. Measurements are "unbiased" but are not part of the human hearing chain and therefore must be interpreted in terms of a variety of factors to be useful.
Personally I rely heavily on measurements, and modeling based on them, when developing the crossover on the as-built loudspeaker. Once I feel I have the crossover more or less finalized using this approach, I turn to extended listening sessions using tracks that I am well familiar with and know the tonal balance of. At this stage I make only subtle, broad-banded adjustments that tweak the overall "sonic character" of the loudspeaker in the room where they are set up. Bring them to another listening space that is sufficiently different and this process might need to be repeated, since the room most definitely contributes to what you hear.
Personally I rely heavily on measurements, and modeling based on them, when developing the crossover on the as-built loudspeaker. Once I feel I have the crossover more or less finalized using this approach, I turn to extended listening sessions using tracks that I am well familiar with and know the tonal balance of. At this stage I make only subtle, broad-banded adjustments that tweak the overall "sonic character" of the loudspeaker in the room where they are set up. Bring them to another listening space that is sufficiently different and this process might need to be repeated, since the room most definitely contributes to what you hear.
It depends on the depth of ones knowledge and understanding. A trained engineer would primarily use predictions to design supported by the odd measurement (ear or instrument I wouldn't distinguish) to confirm the validity of the modelling assumptions. Occasionally the modelling might require some empirical input (e.g. damping coefficients) because of the variability of materials like wood, joints, etc... but the reasoning about the design would still be based on scientific predictions and not measurements.
If one doesn't have a reasonable grasp of the science then understanding the physical processes and modelling assumptions in the simulations and how to use them to reason about the design becomes problematic and unreliable. More measurements can help with the reliability (subject to the experiments and measurement techniques being competent and reliable which is not always the case). Independent reasoning however becomes problematic if one has nothing reliable on which to base it. So people look to experience, others and measurements for guidance.
Good point. Start then with measuring the parameters and driver responses to make sure these are accurate to input into the sim, GIGO?
You have to have your measurement technique nailed where you can routinely get repeatable and accurate measurements or any hope of accurate simulations goes right out the window. That for me is the most important part of it to get right.
After you do your build you can easily take a measurement to confirm the accuracy of the simulations. They should be very close if all was done correctly.
Rob🙂
After you do your build you can easily take a measurement to confirm the accuracy of the simulations. They should be very close if all was done correctly.
Rob🙂
Then you have more faith in the measurements? If the two align why not finish with measurements?
Yes, the room is a big factor, but not really what I'm interested in, very difficult to simulate I'd imagine?
Interesting, this makes a lot of sense to me. Simulations seem very powerful these days and as long as the inputted data is accurate they should give an accurate result. I too wouldn't make much of a distinction between microphones and ears for measuring in this context, it seems to me they are both unreliable but for different, sometimes opposite, reasons, for example the measurement and audibility of comb filtering.
Yes! I build the boxes first, then install the drivers and take measurements of the individual drivers and then drop this into a crossover simulator. When all is complete I will verify with an actual measurement against the simulation.
Rob 🙂
You should give most credence to your ears. After all - if you don't like it, it doesn't matter how well you simulated or measured your speaker.
Simulations are very useful to invest in a "will these blend"? concept with chosen drivers. How easy or hard the (passive) crossover design will be. I use them exclusively before spending a cent.
Measurements are great to confirm simulations.... or restart the design process in your actual cabinet.
Credence is to ears on the end result. Credence is to simulation and measurement and repeat cycles to design an end result you are going to be happy with.
Simulations are very useful to invest in a "will these blend"? concept with chosen drivers. How easy or hard the (passive) crossover design will be. I use them exclusively before spending a cent.
Measurements are great to confirm simulations.... or restart the design process in your actual cabinet.
Credence is to ears on the end result. Credence is to simulation and measurement and repeat cycles to design an end result you are going to be happy with.
Depends on how you specify "xover simulation". Skipping possibility to control sound in radiation space with crossover frequencies, phase matching and filter orders, and compromise between design axis and off-axis/directivity/power is just ignorant and unprofessional with concepts and environments where these things matter. Usually everywhere, but of course we could name that process something else than "xover simulation".
Bollocks to that. You can make a very poor crossover using some of the active pro crossovers, for example. "Active" does not mean idiot proof, and the user still needs to know what does or will happen given a certain set of crossover parameters. The ears are a very poor "instrument" for selecting between the myriad choices that are available at the start of the development process.
15" coax has been first encounter for me to get best out of it when simulating.
mixed slopes with slightly different xo points has been better than good ol symmetrical LR4.
mixed slopes with slightly different xo points has been better than good ol symmetrical LR4.
I define it as going straight to taking raw measurements of each driver section on and off-axis.
And use those to determine starting xover points.
Then for each driver.... smooth in band, and also out of band based on xover points chosen.
Then put complementary steep linear phase xovers in place.
Make polarity, time, and levels alignments.
Check with polar measurements, make adjustments.
All measurement based...
Get's me there without sims...just my style...