Active crossovers in software

[wwh]

The recent discussions about active crossovers have inspired me to complete my long-planned software based active system.

Some background --

I had a minimalist system consisting of a Squeezebox 3 feeding a Crown D-45 for my Khorns. I had no specific complaints about the sound, although the room isn't ideal (7 foot ceiling) and I thought the upper midrange sounded "shouty."

I like the convenience of having my entire music library on a file server, but the Squeezecenter software never really handled classical music well (with tracks on an album grouped into a "work"). I also have a growing amount of hi-res material from HDTracks (and Mr. Mallette), and the Squeezebox is only capable of 24/48 playback.

So, what I've implemented so far --

- Replaced the Squeezebox with a PC dedicated to playback, resampling, and crossover filtering.

This is an older Athlon XP running Ubuntu Lucid and a custom front-end I wrote in Python that serves a web-based control interface (that can handle the music tagging *exactly* the way I want. :-) ) The front-end controls the Music Player Daemon (MPD) software, which provides decoding and gapless playback and resamples everything to 24/96. The resampled stream is fed to BruteFIR for filtering, time alignment, and 3dB of attenuation for the squawkers.

- Replaced the D-45 with an Outlaw 7075

No real perceived change here. Both do the job and sound somewhat "warm" to me as compared to the excessive brightness of the mid-fi receivers I tried previously.

- Custom fixed attenuators between the sound card and amp.

I experimentally determined the attenuation required so that when the sound card output is at 100%, it is louder than I'd ever want. This allows me to utilize most of the sound card's volume control range and avoid the complexities of a 6-channel analog pot.

Impressions --

The sound seems "tighter" in the lower frequencies. As mentioned in one of the other threads, this is probably due to better damping by directly connecting each driver to an amp.

Attenuating the squawkers has mitigated much of the shouty character, although maybe 3dB is too much?

Hugely tweakable in software alone -- time alignment, EQ, crossover slopes and frequencies, etc., are just a matter of editing some configuration files.

Further experiments --

BruteFIR supports up to 128 bands of EQ and can also perform room correction.

Upgraded / different sound output hardware. I'm using an Asus Xonar DS card (OK specs, relatively cheap), but there are many alternatives, including external USB or Firewire-connected devices (subject to software support in Linux).

Put the computer in a case that looks more at home among audio equipment, instead of the current big beige box.

Have others here tried something like this? What were your results and impressions?

[Chris A]

Have others here tried something like this?

I've often wondered why more folks (who are handy with open source) haven't done stuff like this in the past. This looks interesting - I'll have to tinker a bit if I can convert an older computer to Ubuntu (in my spare time...).

The "shouty midrange" is something that you might use Room EQ Wizard (REW) and a calibrated microphone to help you understand this issue a bit better. The Richard Heyser Khorn review from 1986 might help you pinpoint some issues. In the end, though, some of the midrange "shoutiness" is probably related to frequencies below ~1 KHz when the Khorn midrange horn loses vertical pattern control and about 7 KHz where the mid crosses to the tweeter. These aren't necessarily items that will show up in REW because they are related to the precedence effect, speaker cabinet diffraction issues (~7 KHz) and room acoustics.

The sound seems "tighter" in the lower frequencies. As mentioned in one of the other threads, this is probably due to better damping by directly connecting each driver to an amp.

You might also find that the other two drivers sound cleaner. Are you tri-amping or bi-amping?

Chris

[mark1101]

So if I understand this you have passive crossovers in your speakers and are using software to control specific segments of the frequency spectrum that are allocated to each driver via the passive croosover? I'm guessing passives because you mention only 1 amp. Can you explain what you time aligned?

[wwh]

Actually I use six amps (the Outlaw 7075 is just 7 amps in a box) -- one for each of the drivers. There are no passive components between the amps and the drivers except for protection capacitors for the tweeters (crossover point at 3kHz -- well below the 6kHz crossover that the software filtering provides).

BruteFIR takes a 2 channel input (left and right) and splits it into 6 channels -- one for each driver. Then the appropriate frequency filtering is applied to each channel. In the BruteFIR configuration, you can specify an amount of time delay for each of the output channels, so I set the woofers for no delay, squawkers for 6.8ms, and tweeters for 8.4ms. (I read these numbers somewhere here but haven't measured the Khorn horn lengths. Can anyone verify this?)

I'll have to look into Room EQ Wizard. For others that have set up room correction -- any advice regarding specific (affordable) calibrated microphones? Are there any out there that don't require phantom power (that I could just connect to a line-level input)?

- Bill

[mark1101]

Hey that is really cool stuff. I have to beleive there is going to be more software like that coming. Is this an off the shelf program that does all this speaker processing, and what sound card are you using? Sorry for the questions but this is the first PC based speaker processor I've heard of someone using.

As far as mics go, I'm using the Behringer ECM8000 RTA measurement mic. It does require phanton power but seems to work well and has held up well.

Edit. I see you are using an Asus sound card. So question answered. I am actually using the same.

[DrWho]

Wow, that's really cool. I've tried doing just what you're doing in tbe past, but I'm a Unix noob and can never seem to find the right drivers for my HW. I'll have to look into tbis again as that would be a very cost effective approach.

As far as measurement goes, I'd recommend the Dayton mic over the Behringer since it measures better even though they both cost $50. I'll have to look up tbe delays for the Khorn again since I think they might be slightly different. BTW, all calibrated mics are gonna require phantom power.

[Chris A]

BruteFIR takes a 2 channel input (left and right) and splits it into 6 channels -- one for each driver.

I would hazard a guess that Brutefir looks to be a potential disruptive technology for the active crossover markeplace. In addition, I also wonder if there are efforts to host movie audio codecs - this could disrupt the AV preamp/processor community. I know how much that I spent on a good AV processor - if this could be done using a multi-core-processor laptop - well...[] Digital processing all-the way until the last audio stage, and for considerably less $s than dedicated audio hardware - with instant executable upgradeability via 802.11 and open source. And hardware refreshes would cost considerably less.

I wonder if my Harmony One could control a laptop running the audio processing plus a video switch to handle the new HDMI 2.0 without having to spend major bucks to get a new AV processor? Very cool.

Chris

[mark1101]

It's cool because it is alternative technology and because it can be paired with lots of different soundcards and/or DACs offers lots of combinations. I see an advatage in that it can create a processor system compatible with home audio equipment. That's a big plus. I really do wonder how this software processor and a good outboard DAC or soundcard would stack up against some of the conventional pro audio speaker processors we are using now.

[seti]

Wow, that's really cool. I've tried doing just what you're doing in tbe past, but I'm a Unix noob and can never seem to find the right drivers for my HW. I'll have to look into tbis again as that would be a very cost effective approach.

As far as measurement goes, I'd recommend the Dayton mic over the Behringer since it measures better even though they both cost $50. I'll have to look up tbe delays for the Khorn again since I think they might be slightly different. BTW, all calibrated mics are gonna require phantom power.

This is still a common linux driver problem. Not all linux distributions are the same. I've found SuSE linux to include more hardware drivers than anyone else. SuSE actually funds alot of driver development later released open source style but they release it first. If you run into trouble their forums are superb. The guys there can be a big help. I would ask in advance on the forums about various hardware support.

[wwh]

Wow, that's really cool. I've tried doing just what you're doing in tbe past, but I'm a Unix noob and can never seem to find the right drivers for my HW. I'll have to look into tbis again as that would be a very cost effective approach.

As far as measurement goes, I'd recommend the Dayton mic over the Behringer since it measures better even though they both cost $50. I'll have to look up tbe delays for the Khorn again since I think they might be slightly different. BTW, all calibrated mics are gonna require phantom power.

This is still a common linux driver problem. Not all linux distributions are the same. I've found SuSE linux to include more hardware drivers than anyone else. SuSE actually funds alot of driver development later released open source style but they release it first. If you run into trouble their forums are superb. The guys there can be a big help. I would ask in advance on the forums about various hardware support.

Linux support for newer devices can be a problem. Generally, the older the hardware, the better the support. Also, even when your hardware is supported, ALSA (Advanced Linux Sound Architecture -- the standard hardware / software / API layer) can be a major pain to configure. Most of the modern distributions, for example, will "just work" for typical configurations with a single sound device. Once you introduce multiple devices -- such as on-board sound with a multichannel PCI card, with a USB audio device -- it can get messy.

For the patient, though, this shouldn't be a deterrent. In general, if the hardware is supported, the software can be sorted out. There are many excellent sources of support out there.

One approach I had originally considered for multichannel output is HDMI directly from the computer to a compatible receiver. HDMI can carry up to 8 channels of uncompressed 24 bit 192 kHz audio. However, many of the HDMI output devices on current computers either don't support uncompressed LPCM audio over HDMI (only lossy compressed), or ALSA support doesn't exist.

ALSA-compatible devices are listed in their hardware compatibility matrix, here:

ALSA hardware compatibility matrix

- Bill

[mark1101]

Whatever happened to good old MS? Don't they have something less labor intensive that can do this stuff?

[wwh]

I have come across several Windows-based tools out there that implement FIR and IIR filters, some of which are integrated with friendly interfaces to configure the filters.

One is foo_dsp_xover, which is implemented as a foobar2000 plugin. I've never tried it, but it looks interesting:

http://xover.sourceforge.net/

There is a large amount of information about PC-based digital room correction here (for Windows and Linux), some of which involves crossovers as well:

http://www.duffroomcorrection.com/wiki/Main_Page

- Bill

[seti]

WWH, Yup I agree Sometimes it is easy and seemless and then there are times it isn't.

Whatever happened to good old MS? Don't they have something less labor intensive that can do this stuff?

If I could get it to work easily under linux I would go that route. You can custom build astem that has a much smaller footprint and can do more with less machine. I also like to tinker so linux is fun in that regard. MS is fine just make sure it is behind a firewall has antivirus protection or isn't on a network.