Digital X-Over idea

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Digital crossover

Recently, I built an analog active X-over for my bi-amped speakers. I also built a cheap DAC for this system. Then the idea hit me, why not have the X-over done in the digital domain before the DAC, without any loss or noise? Sure this would require additional DACs and AMPs, but it would prabably be the best sound ever. You could probably continuously change the slope and freq. of the x-over, becuase it would not require different electrical components. That way you could try many different settings to find out wich one sounds best. You could also build in an EQ, and a test mic so that you could have it equilize itself for any given arrangment. Keep in mind that I have not done this, as it is only an idea. But what an idea it is! If anyone knows anytthing about DSP, or computer audio programming, or if such a device/program exists, please post. Or if you want to call me an idiot, please post that too!
Phase coherency anyone?

[Edited by Yoda on 07-29-2001 at 08:25 PM]
This is one of those ideas that keeps popping up--like the one to begin a triamp system in the RIAA eq for a cartridge, i.e. use the 500 Hz and 2100 Hz rollover points to carry right on through the rest of the system. Elegant in concept, but not quite as practical in the real world as the usual implementations.
However, that said, and keeping in mind that it isn't difficult to build a variable analog crossover, you might as well go for broke and program this thing for the various peaks and dips that your drivers perpetrate on your signal so as to get a flat frequency response +- .1 dB. Then use it to time align your drivers for you. Once you've done that, you might as well use the mic to eq the speakers to the room...
That last part's already been done, by the way.
The fly in the ointment is that if you start diddling with frequencies in the digital domain, you often run into problems with phase shifts and such. It's not that it can't be done, it's that there's almost always a price to pay for things.
You'll need to decide whether you're going to implement the program in software (i.e. commit your PC to your stereo and write a program in C or assembler, etc.), or firmware (write the code, then burn it onto a dedicated chip). I'm not aware of anyone out there who supplies a plug-'n-play chip that you can drop into a circuit board for what you'll need. By the time you drop enough money to buy a PC (for the software route), or a chip-burner and sundry other odds and ends (for firmware implementation) you'll be into enough money to buy a pretty nice analog crossover and have change left over. And if you DIY the analog crossover the savings will pretty much speak for themselves.
Anyway, the bottom line is that, thus far, most pieces that do extensive manipulations in the time and frequency domains have not been well received in the market because they don't sound all that good. Eventually someone will get it right, but I personally don't think I can hold my breath that long. Now, assuming that you've got a really cool algorithm in mind that would blow the other guys out of the water...go for it. Maybe you're the one who can make it all come together.

Use a PC add gain to top end

I am considering using a PC with commercial software for this task.

Digital output, external dac's etc.

Now, the top end has lower energy. You can therefore apply a digital gain after crossover, and reduce analog gain in amplifier to compensate.

Of course, it would be opportune to upsample as well prior to doing floating point calculations.

In-room frequency response corrections should be easy once everything is working. I am not sure whether the applications I am considering (Nuendo and others) have provisions for phase correction.

There's a 3-part article in AudioXpress by Richard Mains starting in May 2001 on this topic. Seems to be lots of background information, particularly in part 1. Mains designed a custom DSP board and the circuitry is rather complex. He says he may be selling these boards if there is enough interest. I hope he doesn't mind me revealing his email address (it does appear in the article after all):
A dedicated DSP is a much better choice for this than just a PC, which doesn't handle hard real-time software very well. Perhaps a DSP board with codec that plugs into a PC and allows downloading of code to it might be available; this would give you the advantage of a convenient development environment with the performance of a dedicated DSP. Although running high quality audio in and out of a PC backplane without picking up noise may be challenging.
Digital X-over


I too have been looking into building such a device. I have been paitently waiting for Texas Instruments to come out with their new audio DSP. It is flash programmable and has enough onboard memory and outputs to handle multiple tri-amped speakers in this type of application. It is also the fastest audio processor available and has 64 bit precision. (Sorry, I don't mean this to sound like an advertisment but I must tout my employer!) I believe that this will have enough processing power to accomplish all that you ask. I also plan on using the BB PCM1738 DAC for the outputs. If I can get something together, I will try to post it.

These guys ( make some pretty flash room correction gear, they also make a digital crossover for subs. I think you would need to know quite a bit about signal processing and DSP before you could design this kind of gear. Has anyone got experience with DSP? How hard is it to get started?

I didn't mention this in the earlier discussion about DSP based eq, but the PRO audio world is changing over to digital processing as fast as it can. One single space DSP box can replace a whole rack of dedicated outboard gear (crossovers, eq, delay, compressors/limiters, feedback eliminators, etc). Best of all you set it up from a laptop (and keep knobturners from messing up your system setup). This is better than security covers over the eq and crossovers.

DSP solutions are finally starting to work their way down into lower cost sound reinforcement systems. The A&H DR66 sells for around $1200. TOA and Shure DSP boxes go for around $2500 and will do pretty much everything a typical installed SR system could need. For concert rigs and more advanced installed systems MediaMatrix is the best there is (and is priced accordingly).

Cool Toys.

Phil Ouellette
The prof that the audio industry is moving to digital, it that it's beginning to appear studio monitors that incorporate DSP for Eq, phase and everything.
Dynaudio acoustics is releasing the Air series monitors with DSP. They even comunicate by Cat 5 cable between them and the base controller. Another cheaper entry in this trend is Alesis. I suspest everybody and they parrots will follow very soon.
I do not have the skills to build and program such a device, so I dedicated an old computer running comercial software to equalize and correct my system. It works ok., or at least better that without it.


Hi Petter,

Car stereo's use digital EQ because the special purpose IC's designed for car stereos basically give you digital EQ for free. I am not aware of car stereo's that use digital crossovers (not my area of interest to be honest).

A car is a very different acoustic environment than a home. The noise floor is much higher (wind, tire and engine noise) and the acoustic space is much smaller which locates the listener closer to the speakers. I think most people would be disappointed by the performance of their car stereo's if they tried to use them in their houses.

Switching power supplies are regularly used in car audio but I haven't heard anything about digital amplification. The 12 VDC supply in cars limits the maximum power an amp can provide so high power car amps use a DC to DC converter to create the supply voltages they need. This is also what manufactures are usually talking about when they claim they use MOSFET's in their car amps.

Phil Ouellette
Coincidence always gets you...

I just ordered one of these boards (not yet received, project not started)

the big one: EZ1CUSB-12. Not a DSP per se, but I've seen them used as such and a very powerful tool. My plan is to do exactly what you're talking about, Petter. Bring SPDIF (or other) in, do the jitterbug, crossover, eq, FFT, whatever. Its completely reprogammable to any of the users needs.

About DSP: the math isn't so bad, but there is a lot of math and a lot of math to choose from. You must choose wisely:D.

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.