Just not comprehending all the good info? No need to be skeptical. No secrets held back here. Nothing but truth and a whole lot more.
Phase plots of my previous post are of speaker that follows rule of thumb where drivers are about 1/4 wavelength apart at crossover frequency, and thus function as single source; reflections remain coherent.
A room treats all sounds the same. If sounds entering room are realistic, at least chance exists that they will be perceived as such.
Regards,
Andrew
If you want average good speakers, maybe. But there really is more to it than that. It's like any skill, playing a sport or an instrument. Being good means following a few rules, being great means a lot more. As your knowledge grows, you start to notice things you never did before.
Hi Pano
You've been challenged by an upstart
Can you supply an nice example of how you obtain a better-than-average speaker by going beyond the rules of thumb? Are you suggesting that if we dismantled a highly-regarded commercial speaker we would begin to see strange things within it that caused us to scratch our heads? As I say, I'm sceptical. Within the drivers themselves, yes, I can imagine all sorts of subtle and clever things going on - or maybe it's just a case of basic theory put into practice solidly - but the boxes themselves..?
I am very impressed by your phase plots, and I like that rule of thumb, but is it as perfect as you suggest, especially at shallower crossover slopes?
Do you have an opinion on the so-called Stochastic Interleave function described here?
http://www.essex.ac.uk/csee/researc...Analogue and digital crossover alignments.pdf
Yes, usually. No speaker (or its parts) is perfect, so getting what you want becomes an exercise in compromise. Knowing what is important to keep, what can be lost, what blends well with what - are all things that take a lot of experience. You can get lucky with "design by the rules" and you certainly should start that way to understand what's going on and to have a decent chance at a good speaker. Learn how to draw before you learn how to paint.
If you are just talking about boxes, then it's simpler, tho not simple. If you are talking about driver choice, crossover design, box design, porting, stuffing, interaction with the room, etc. - then the balancing act gets tricky.
Shallow crossover slopes:
Simulation of 1kHz crossovers with 1st order Butterworth and 4th order Linkwitz-Riley with 1/4 wave offset at 1kHz:
The above pic shows ripple in 1st order filter due to summing of low pass filter attenuation band with high pass filter. As filters become steeper, ripple becomes single dip with width inversely related to filter steepness.
Continued use of low slope filters offers no advancement in speaker design. Their simplicity in design and use highly compromises coherent sound radiation.
I checked out Stochastic Interleave, and it is complex way of saying: Let's compromise on axis performance for improved off axis performance. IMO, junk.
Regards,
Andrew
I like it. What do you define as a low slope? Anything below 16th order, say?
Presumably with steep crossover slopes there is less reason to use it anyway..?
Can you give an example of something you have done that lifted a speaker beyond average? And what led you to think it would work before you did it?
To some degree, yes, but consistent directivity can be prioritized among the various speaker design tradeoffs if one wishes. In particular it's hard to acheive consistency between vertical and horizontal directivity in a point source speaker which achieves commercially viable SPLs. Line arrays solve this but tend to be too big for home audio, which leads to problems with phase coherency as the walls and listener are acoustically close. Making the room dead can address that, but such extensive room treatment is often not feasible. In a lot of ways it works better to chuck the whole problem and go with IEMs but one advantage DIY has here is that, since you're not usually trying to make a buck off the speakers, you can relax the SPL requirements in favor of improving directivity. When I did that I ended up here. Not really any rules of thumb in that design; it was all simulated beforehand and there was as much of a chain of ABX test results behind the design decisions as I could manage (though some things, like testing of different speakers, are really hard to make wholly blind in a home context).
For commercial builds where the design process is openly available try Geddes and Nao, both proprietors of which are active on this forum. In addition to the usually cited sources you might also find the papers from Harmon-Kardan's research group interesting. In DIY, the Ardor springs to mind as well, as do some of the builds from the moderators over at the misson possible forum on HTGuide. At the end of it, planet10 is right, though; there's enough we don't really understand about hearing and how listener to listener variation affects subjective sound quality that, at the end of it, as a designer you have to pick some tradeoffs, build something, and see how it turns out.
I would, however, argue that most performance is entirely predictable from physics based reasoning about a design. The main difficulties is the metrics needed to translate that into a prediction of subjective quality don't exist and that their establishment is beyond the resources of nearly all DIYers. So what you see in this forum is a lot of people either struggling with the learning curve to understand audio---the physics are not easy and neither is psychoacoustics---or struggling to figure out why something does or does not works well for them. Sure, it's messy. Get used to it.
Last edited:
Transition from ripple to dip is seen in pic with 4th order. Ripple is lobe behavior. Dip is smooth wavefront.
I see no reason to compromise on axis performance for off axis performance.
I see greatest advantage of steep slopes in controlling IMD.
Regards,
Andrew
I'm not Barleywater, but interesting paper. I'd expect the interleave to sound worse; the transition band is wider and less consistent whereas all the test results I've gotten favour narrowing the band and keeping the bounding volume of the radiating surface as acoustically small as is feasible (at the crossover's center frequency the radiating surface is obviously the union of both drivers but there's no well defined point in an XO's stopband where the suppressed driver is no longer audible, though I would tend to say it 40 to 60dB down is a reasonable choice). However, ya don't know for sure until you set it up both ways and ABX it. Fortunately digital makes ABXing a lot easier.
But how narrow would you make the band? Some people think that steep filters sound intrinsically worse because of pre-ringing, and/or that a generous degree of overlap is naturally a good thing so that there is a less abrupt transition between drivers. Personally, I think you can go pretty steep without a problem, but maybe my ears aren't golden enough.
A few examples might be:
- Where do you set the crossover frequency and why? Standard practice or your own findings and choices?
- Steep filters, shallow or a mix? How much overlap or spread?
- Driver placement and alignment. A pretty, symmetrical layout, or something more researched and refined?
- Harmonics. Do the drivers have similar harmonic signatures, or ones that work well together?
- Is your crossover point good harmonically for your drivers? Or does it sound strange? This can be counter-intuitive.
- Driver Q. Are they similar? If they are, the drivers tend to blend better and call less attention to themselves.
- Box tuning. By the book tuning, or something you find works better? It's surprising what works and doesn't with boxes.
- System Q. Optimally damped or over/under damped? Which do you prefer? Which works best with the other limitations of the speaker?
- On and off axis, or power response. What's important to you? How will the speaker be used?
The problem with speakers is that we have to listen to them.
They would be much easier to design if we didn't.
Causal filters do not prering but linear phase filters do, so the tradeoff depends in part on how much phase error you're willing to accept and whether the speaker you're building allows you to rake the drivers to compensate (works in boxes, not so much with open baffle). There are few single variable optimizations in audio and crossover selection tends to be one of the more multivariable ones since driver behaviour, box/baffle/nude driver directivity, and room response are involved in addition to the cleanliness of the crossover's impulse response and the acoustic size of the variation in the radiating surface. A good exercise is to set up a digital XO and spend an afternoon measuring and auditioning different crossover alignments and phase corrections.
Personally it's my experience it's not exactly easy to beat a fullranger with a multiway design especially at a comparable price point, though digital helps a lot there and I seem to be more fussy about a tight impulse response than average. I've yet to hear an odd order alignment where the quadrature sounded as tight in a linear phase crossover as the in phase summation of a Linkwitz-Riley alignment. I can't see myself ever using less than LR4 in multiway---I agree with the benefits claimed for shallower slopes but the better directivity control and avoidance of driver limitations (such as cone breakup or Bl nonlinearity) of LR4 has consistently been subjectively preferable for me (expect a bunch of posts disagreeing with this, though passive LR4 is certainly a pain)---but every time I've tried it LR8 hasn't sounded quite as good. Not entirely sure why; maybe it moves the sound location too rapidly in frequency, maybe it's greater preringing when corrected to linear phase, though I would tend to bet more on the transition being too fast. B5 and B7 (fifth and seventh order Butterworth) have that odd order offness to me, whether linear or warped phase, so I haven't used them even though a lot of people like odd order crosses. As a result, my go to crossovers are LR4 and LR6. Of the two LR6 usually sounds a bit better, mostly because I can pull the mid-tweeter cross a bit lower and get slighly more controlled directivity out of it. Though sometimes it's helpful in working around driver limitations.
I've not tried ninth order or higher...
Amen! But it takes me longer than an afternoon.
FWIW, I also like LR 4th (or Bessel) and will use whatever electrical function it takes to get that acoustic slope. For me, it's often 3rd order electrical.
I wish I could be so certain about what sounds best when auditioning! In my homebrew software the aim has been to make it as real time as possible. I can calculate and load new filters on-the-fly, and modifying delay, crossover point or filter slope is done by changing a single value so I really can do an A/B test very easily. I also have facilities to instantaneously switch baffle step compensation in and out, and amplitude & phase corrections based on measurement. The stochastic interleave function is next on the list.
But I can't tell you what sounds 'best'. I can tell you which grossly maladjusted settings sound bad, and I can hear subtle differences on white noise, but there is a broad range of settings that make little audible difference, or do not clearly show up as better or worse than other settings. Or are matched in significance by changes in listening position. I suspect that moving objects around in the room would also be significant. But then, how do I know that a 'good' setting isn't just a one-off for the particular types of music I've tried it on? All I can say is that the broad range of settings which don't change the sound particularly, in fact sound extremely good.
The one moment when I've gone "Wow!" has been when applying what I thought was erroneous phase modification. I was convinced my code was applying twice as much phase correction as required (based on mid and tweeter individual measurements), but having the time to sit down with it today, I see that it isn't... How wonderful it would be if I was hearing the effects of correct phase compensation through the crossover region, and that in those circumstances I could tell that a shallower crossover sounded better than a very steep one...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.