By tough I meant the design process itself, especially because it's hard to measure anything properly in this region. Typically this is outside the capabilities of a common DIYer (would need to go to great lenghts), so shortcuts are needed.
And it's great to have these models so we know what to expect.
Last edited:
Ah was thinking the XO transition between say a 2 way with woofer and comp driver a top.
That's not so tough per se, IF you have the data. The truth is we often do this a bit blind.
This is the region that dsp crossover can be indispensable. I just let Bacchdsp with ORC sort it all out. Job done.
I would never let a DRC algorithm to "sort it out" if it's a crossover that needs to be done. Once you have a good crossover, then why not (say below 400 - 500 Hz).
Yeah simulations are great, to get some target that the response could be in! Also nearfield measurements and ground plane measurements help, but come with couple more opportunities to make humanly mistakes with the data...
The problem in the end is that if the sound is not as good as one expected, then what gives? Is it my hearing, some bias going on? Was my expectations unbased? Is there issues in my implementation and it's not what I simulated? is it my data used in simulation? or what it is? It's just great amount of uncertainty and it takes some time to solve everything out and validate it's what it is supposed to be and there is no significant error in the implementation and it just is something I didn't expect. Perhaps it is I just didn't know.
Even if all issues get spotted and sorted out there is still uncertainty about whether perceived sound is from positioning and all sort of things. Only after it's all clear as day one could try and implement another version to evaluate fairly, and so on. Absolute success is very hard to notice in a way, it can be quite an effort to validate a system is tuned to it's best. Gotta remember it's hobby and it's supposed to be fun, so as long as it's fine enough one should be happy 😀 And, every issue spotted and fixed ought to make it that much better 🙂
The problem in the end is that if the sound is not as good as one expected, then what gives? Is it my hearing, some bias going on? Was my expectations unbased? Is there issues in my implementation and it's not what I simulated? is it my data used in simulation? or what it is? It's just great amount of uncertainty and it takes some time to solve everything out and validate it's what it is supposed to be and there is no significant error in the implementation and it just is something I didn't expect. Perhaps it is I just didn't know.
Even if all issues get spotted and sorted out there is still uncertainty about whether perceived sound is from positioning and all sort of things. Only after it's all clear as day one could try and implement another version to evaluate fairly, and so on. Absolute success is very hard to notice in a way, it can be quite an effort to validate a system is tuned to it's best. Gotta remember it's hobby and it's supposed to be fun, so as long as it's fine enough one should be happy 😀 And, every issue spotted and fixed ought to make it that much better 🙂
But software like audiolense makes far better crossovers than humans. At least in the 100-1000hz range. Maybe 'crossover with roomeq' needs a new name?
So you measure each driver separately and feed those data into the software? How many points and how far apart?
Wait, sorry, how are you getting down to 550 Hz with the HF driver? Which driver/waveguide combo?
Do you mean software that combines crossover implementation with room EQ ?
Hope not...I think that's the biggest mistake anybody could make in tuning/setting up speakers.
With audiolense its either a single point or an average of a few positions. I've found a single measurement is best. But since I've got BacchDSP I take responses from favourite seating positions around the room. It can track you with a camera and adjust for where you are in the room - all without changing the tonal balance and all with crazy 3d realism. The ORC room correction is particularly impressive - apparently using psycho-acoustic processing.
The imagining side of it is only a small part of the benefit. I find that the regeneration of the natural acoustic is the most important factor.
I have to say that when I first heard about this software I assumed it was all hype and nonsense. But it turns out the Dr Choueiri is a bona fide genius audiophile and everything he says is true!
Audiolense gives you the option to do eq on each driver separately or just an overall system response. Initially I assumed that the 'per driver' would be way superior (because drivers can be individually time aligned) but if you make a 'rough but good enough' passive crossover the full range eq sounds remarkably similar. The long .fir filters sound best - I think 32k was the sweet spot.
BacchDSP's filters are much shorter.
BacchDSP's filters are much shorter.
I don't want to veer too far off the thread's central focus (or get into Bacch for sure, or even Audiolense other than perhaps it's DIY speaker crossover construction capabilities).
In a nutshell, I think crossovers are a function of the speakers' physical acoustic design. Anechoic acoustic design.
(if the speaker is being deliberately built for boundary interaction, those boundaries should be set up in the anechoic test bed).
The electro-acoustical tuning needs to be fully independent of any particular room.
Leave room-correction to a separate layer....DSP, acoustic treatment, however accomplished.
Room correction software is getting better, but still measurements in a small room are incredibly difficult to make reliably, repeatably,
.... especially spatially....super especially low frequency.
In complete frankness, and being a consummate measurement addict, I think most of the glorious in-room measurements we see on sites like ASR, are either luck to a spot, or BS. And even if a glorious in-room measurement is bona-fide, I'd like to hear the system before accepting its validity.
I'd certainly not trust software to be able to alter crossovers automatically in DSP.
I'm not sure I can conceive of a more relevant measurement than a HRTF taken at listening position.
Apart from diagnosing major faults im not clear how anechoic response can be very useful. Reminds me of those who judge a musician's rhythmic accuracy by correlation to a metronome.
Apart from diagnosing major faults im not clear how anechoic response can be very useful. Reminds me of those who judge a musician's rhythmic accuracy by correlation to a metronome.
It's still not clear how it can make a loudspeaker crossover, i.e. apply different filters to the individual drivers so that they acoustically sum to "flat".
You measure each drivers raw response. Then set the crossover target accordingly (audiolense has some nice slow then fast rolloff filters). Then you set a room curve and the software calculates a .fir filter for each driver. Stick them in Reaper or the marvellous Hang Loose Convolver and you're away. Mitch Barnett is the guy who knows all the subtleties of filter making, but it's possible to generate good sounding filters with my only moderate knowledge.