3-Way digital crossover: what are the options?

Hello folks!

I started working on the crossover in Xsim for my ATC SCM25a "clones" (or my closest approximation thereof), and realized that making a 3-way passive 4th order Linkwitz-Riley crossover was going to be a huge undertaking. So after talking to some folks, I think DSP is probably the way to go considering the tweaking that will surely follow.

I am a live sound and studio person, and the hifi world is a little bit foreign to me. Of course I've heard of the MiniDSP 4x10 HD, which looks like a great, albeit costly, option. My first thought was why not use something like the DBX Driverack 260 or PA2? These can be found pretty easily for a great price, particularly the former. I've used these to crossover PA systems, but never in the studio. Would love to hear yalls thoughts.

Thanks!
 
You are probably already aware of this, but most 4th order crossovers have a 2nd or 3rd order electrical topology... For example it is very common for a midrange or midwoofer to have a single series inductor and parallel cap as the low pass filter, while the tweeter high pass has a series cap, parallel inductor, and another series cap... and yet the result is 4th order acoustical slopes.

But yes, you are correct that a well designed 3-way passive speaker (including crossover, baffle, bass alignment, and construction details) is a big undertaking. If it is your first design, there is a low probability of getting it right the first time. It would be analogous to winning a chess tournament after learning how to play chess the previous week... possible, yes, but not likely.

However, if you approach it as a learning experience, and are willing to make multiple versions and learn from each evolutionary iteration, you can eventually end up with something really nice.

A DSP filter can simplify things a bit, but it is not a magic wand. Some people think it is a short cut around much of the design work, but it actually is not. You will still need to intellegently select your drivers, design the baffle with software simulation, design the bass alignment, design a preliminary DSP filter to show feasibility, design the cabinets, make detailed measurements of your drivers as installed in your cabinets, make a refined simulation, make a refined DSP filter, make system level measurements, and finally, do listening evaluations to fine tune your DSP filter.

For many of us, this process is very rewarding... it is our idea of fun. If this appeals to you, welcome to the team...

Regarding the MiniDSP: A MiniDSP 2x4HD can be configured as a very capable 3 way crossover. You will need a 2x4HD for each speaker, as well as 3 amp channels. Many people like the IcePower amplifier boards. I use Hypex Fusion FA253 plate amplifiers for my active 3-way speakers. It combines a DSP filter with a 3 channel amp on a single plate. I recommend it highly. I can not comment on the other products you named.

j.
 
Thanks for your reply.

I have a blast with all my projects. No expectations other than to learn!

My driver selection was assembled based on these criteria: flat response, good off-axis dispersion, Fs at least one octave above the theoretical xover points, and cost. All of these drivers and data were gathered from the parts express website, so nothing expensive or esoteric going on there. Heck, I have not even completely settled on the drivers. I have modeled several enclosures in WinISD with a variety of LF drivers, and each is a small compromise in some respect. Nothing "wrong" with any of these models either, just different! Makes pulling the trigger a bit hair-pulling.

I think I'd like to steer clear of MiniDSP. The amps that I will be using are 19" rack mountable class AB amps and I would prefer if the crossover unit could share the rack, which is why I gravitated towards the 4x10. Also, I think balanced line level analog I/O is a must. XLR even better, which is another nudge towards DBX products... but still I have not heard of anyone using them for designing crossovers for monitors in a critical listening environment, but that certainly doesn't mean it hasn't been done.
 
DBX driverrack is good for pa and so. But for critical listening i would not use it as the ad/da parts are mediocore. Xcilica and a few other brands make dsp's that are worth it, but they cost a lot, a lot more than the MiniDSP.

MiniDSP itself also has that issue, but you got the digital version, that you can connect to a good ADC and DAC to avoid that issue.

In any way, DSP's are expensive, that is why a lot of people just use a computer (often a rasberry PI) with Camilla DSP software and external DAC's. Camilla DSP is an open source project, still a bit in dev state (but stable). It may be that your anti-virus does mark it as unsafe as it's not signed and official published.
 
If you are a live sound and studio person then you’ve got a head start; you know all about PEQs, shelf filters and high/low pass filters, and you used to doing real time adjustments and you have good (trained) ears.

For the non-muso crowd; the venu360/260/PA2 from dbx have a 8 PEQs on the inputs, and 8 PEQ on each of the 3 outputs, gains of up to +/- 12 dB, Q of 0.1 – 15.909, and shelf filters (called SLOPE) of 3 dB/Octave up to 14.295 dB/Octave. In addition it’s got low and high filters (XOVER) of Butterworth type 1st to 8th order and LR 1st to 8th order, polarity invert, limiters, and delays of in increments of 0.02ms. It’s also go a phantom powered mic input and a RTA.

A friend of mine offered to lend his own to me for loudspeaker development, so I had to look into it carefully to see what it offered.

The main downside is that the circuit design is a bit compromised and the DA and AD conversion are not state of the art; so the SNR is about 90dB or so. Which I think is totally fine of PA use.
For home hi-fi where the background noise is much lower, and power amps may have SNR approaching 100+dB, and I have full disposal of a PC, I decided to go with what I already know (miniDSP). But apart from that, the dbx units have an advantage in they are very intuitive to use eg can without a computer; from an iPad where you can grab and play with all the sliders and hear it in real time.

If you go ahead with this route there's a few limitations with it, but that's not the main challenge.

Your main challenge will be learning what hifijim was talking about- baffle design, eg. where to place your drivers on the baffle; cabinet design eg. woofer/bass alignments, taking accurate quasi-anechoic measurements, as well as advanced topics like trade-offs like on axis, listening window, dispersion or directivity, power response. And all these topics are important regardless whether you do DSP or ASP or passive crossovers. I believe it's best to start with a 2 way wide band speaker than can go down to 50Hz, before you go a 3 way speaker. But some speaker to dive in and go full throttle 3 or 4 way. But

You will still need to intelligently select your drivers, design the baffle with software simulation, design the bass alignment, design the cabinets, make detailed measurements of your drivers as installed in your cabinets, take measurements

If this sounds like fun, go ahead!
 
Last edited:
Hi, my 2c advice for the stage you seem to be at, is to get the lowest priced dsp that gives you the functionality you want, without a lot of concern for the quality of AD/DA conversions.
I'd put channel counts, filter type and counts, I/O connections, rack mount, etc etc over DAC quality...for now....

I think what will be learned in taking measurements, and setting up the xovers and EQs, will strongly dominate the sound quality achieved.
Once you feel you have that down, you might then consider if you need better AD/DA, and exactly how much more you're willing to pay for it. And whatever additional functionality, like FIR maybe, you want to add in.

I don't know the DBX units, but I've used dsps from the lowly dcx2496 to Linea Research ASC-48 which specs a bit better than Xilica units, to the xovers/EQs in x-32s, to a bank of miniDSP Open-DRCs with both their own DACs and as pure digital with Midas stagebox DACs, and finally Qsys open architecture processing.
I'd say the worst specs on that list sounds better than the best specs on the list, depending on how well my tuning was measured & implemented.

That's why i recommend just get what you know you need.... for now...
 
DBX Driverack 260 or PA2 will be fine. Audiable issues are most likely incorrect gain structure. E.G for the 260 max output is listed as +22dBu wheras most power amps will be at clip by 8dBu (at most) therefore for best SNR ~12dB attenuators (to leave some allowance for poor clip behavior on the 260) should be placed between the DSP and the power amp (can be made from just 3 resistors) (for the hypothetical 8dBu power amp). It appears to also have jumpers that are used to match the maximum output of the source to the input stage.
 
If you are good with computers, e.g shell scripting then a raspberry pi or PC based Linux systemusing ecasound or camillia dsp, you just need an HDMI extractor, possibly an existing cheap HDMI a/v receiver. Number and type of filters only limited by CPU.
 
Thanks for the replies and advice, everyone.


I think of the DSP options mentioned in this thread so far, I will go with the DBX stuff. Familiar (to me) interface, rack mountable, balanced I/O, and easy to find second-hand.


For those interested, here is the driver list for the build:


Low:
https://www.parts-express.com/pedocs/specs/295-430--dayton-audio-ds215-8-specifications.pdf


Mid:
https://www.parts-express.com/pedocs/specs/285-020--dayton-audio-rs52an-8-spec-sheet.pdf


Hi:
https://www.parts-express.com/pedocs/specs/275-040-dayton-audio-nd28f-6-specifications-46119.pdf


Also, here's some screenshots from sketchup and WinISD's predicated LF response:


Front of enclosure:
flush-baffle — ImgBB


Rear/interior of enclosure:
flush-baffle-rear — ImgBB


WinISD graph:
https://ibb.co/YWdN7BJ


I've been tweaking the cabinet volume and subsequently the depth of the cabinet, so that plot is tentative, though in the ballpark of what type of LF response I'm hoping to achieve. Port diameter is 3", and roughly 8" long. Also, I would appreciate input re: the cabinet design. Part of the reason I chose these drivers is because I was able to position them relatively close to one another on the baffle. I see a lot of folks rounding off corners for diffraction's sake, I suppose. I really like the "vintage hifi" look; black, veneer, and SQUARE, so I don't know if I'm willing to sacrifice aesthetics... yet.


Any advice welcome. Thank you all again.