mefistofelez said:
Well, I can tell you, but you won't believe me
I'll eventually post more details about what I'm doing, but I didn't want to post a bunch of stuff before I had anything running just in case it didn't work.
I'm actually currently using TangBand 852's - little 2" drivers. They're only good down to 500 or maybe 400Hz, but they make it up to 2k in a 60x60 square conical flare. My eventual plan is to use the 2" Aurasound Whisper drivers which seem to be pretty much the best small drivers around.
I haven't really done any measurement of the system to ensure that the xover is properly aligned, but even so it's sounding pretty decent just tuned by ear. I doubt that my approach will work particularly well in a passive implementation - it definately needs eq and quite possibly some digital delay to make it work.
mefistofelez said:
Yeah, it's a tough set of tradeoffs. I really wanted to push the upper xover point as high as I could, which is how I ended up with the tiny 2" drivers - it's the only way you can get the mid entry holes close enough to the throat without having path-length cancellation kill you. They don't have enough excursion to handle the low end, though (although the Aurasound 2" are a bit better on that front). I was (and maybe still am) thinking of a 3-way Unity though, so I'd use 6-7" midbass drivers to bridge up to 4/500. Unfortunately, getting any response down below about 100 or maybe 80 with any efficiency is really tough, so you'd be forced to use real woofers, not just a mono sub in that case. Makes for a big Unity as well, and I'm not sure I can use that in my room.
Useful Spectrum Analyzer
I've been playing around with my setup a bit in order to measure how well I did with drc and am finding the jaaa spectrum analyzer easy to use and very important. The first thing that happened was I discovered that my crossover point is NOT 2600Hz, it's 5200Hz!!!!
This tool plugs nicely into Jack so it can look at the outputs throughout the entire chain of your system including a microphone recording your system response. I've already looked at the output of xmms and high and low channels of brutefir for example.
You can find out all you need here:
http://users.skynet.be/solaris/linuxaudio/
Like all Linux tools, you will need to fuss with installation but I must be getting good at this because it took a half hour between download and discovering my crossover mistake.
Still working on documenting my system in case you're still interested.
-Robert
I've been playing around with my setup a bit in order to measure how well I did with drc and am finding the jaaa spectrum analyzer easy to use and very important. The first thing that happened was I discovered that my crossover point is NOT 2600Hz, it's 5200Hz!!!!
This tool plugs nicely into Jack so it can look at the outputs throughout the entire chain of your system including a microphone recording your system response. I've already looked at the output of xmms and high and low channels of brutefir for example.
You can find out all you need here:
http://users.skynet.be/solaris/linuxaudio/
Like all Linux tools, you will need to fuss with installation but I must be getting good at this because it took a half hour between download and discovering my crossover mistake.
Still working on documenting my system in case you're still interested.
-Robert
Testing, testing...
I'm trying to test my system so I can optimize drc. As mentioned earlier, my initial drc iteration was "successful" and the sound is improved but I know there are problems and I want to "see" them.
I'm wondering if any of you can recommend some free software audio testing tools that interface with Jack?
A signal generator attached to Jack would help quite a bit, especially if it could sweep frequency and generate white and pink noise. Then, it would be simple to measure steady-state frequency response.
A pulse/impulse generator coupled with an oscilloscope would greatly help test the alignment of speaker drivers and measure group delay etc.
I'm hoping I don't have to generate test signals with Octave math software and then play the signals manually and hooking up hardware testing equipment isn't practical.
-Robert
I'm trying to test my system so I can optimize drc. As mentioned earlier, my initial drc iteration was "successful" and the sound is improved but I know there are problems and I want to "see" them.
I'm wondering if any of you can recommend some free software audio testing tools that interface with Jack?
A signal generator attached to Jack would help quite a bit, especially if it could sweep frequency and generate white and pink noise. Then, it would be simple to measure steady-state frequency response.
A pulse/impulse generator coupled with an oscilloscope would greatly help test the alignment of speaker drivers and measure group delay etc.
I'm hoping I don't have to generate test signals with Octave math software and then play the signals manually and hooking up hardware testing equipment isn't practical.
-Robert
Public humiliation
Humbly reporting the up-to-the-minute developments, I have discovered that the excellent jaaa tool reported a few posts back HAS a signal generator built in, I was just too stupid to find it.
I was running it in too small of a window and didn't realize the signal generator buttons were sitting off the bottom of my window.
It provides nice noise and sine signals. I'm off to the races for frequency domain.
Time domain coming up soon.
-Robert
Humbly reporting the up-to-the-minute developments, I have discovered that the excellent jaaa tool reported a few posts back HAS a signal generator built in, I was just too stupid to find it.
I was running it in too small of a window and didn't realize the signal generator buttons were sitting off the bottom of my window.
It provides nice noise and sine signals. I'm off to the races for frequency domain.
Time domain coming up soon.
-Robert
If this is off-topic here, please feel to ignore the question.
However, since a lot of you are already in the implementation stage, I would like to ask about a sound card that would allow me to send the digital data produced by the PC based filter over digital links to an amplifier (e.g., S/PDIF of Panasonic XR-10/25/45/50).
Unfortunatelly, I am not smart enough to correctly interprete cards' specification. As an example, here is Hammerfall DIGI9636:
2 x ADAT digital I/O, based on RME's reliable Bitclock PLL
1 x SPDIF digital I/O, based on RME's reliable DIGI96 technology
1 x ADAT Sync In (9-pin D-type) for sample accurate transfers
1 x Breakout cable for coaxial SPDIF operation
Zero wait state PCI Busmaster interface with additional burst FIFO (130 MB/s transfer rate in both directions)
New hardware design: 36 mono channels in block mode, organized in 32 bit ASIO double buffers
ASIO poured in hardware! The result in speed and performance is 0 (zero!)% CPU load when using ALL 36 channels!
130 MB/s transfer rate for record and playback results in only 7% PCI bus load when using all 36 channels
New Enhanced Zero Latency Monitoring
S/MUX poured in hardware: 8 channels 96 kHz/24 bit for record and playback
If someone could help, or point me to the right reading material, I would greatly appreciate it.
Thank you,
M
However, since a lot of you are already in the implementation stage, I would like to ask about a sound card that would allow me to send the digital data produced by the PC based filter over digital links to an amplifier (e.g., S/PDIF of Panasonic XR-10/25/45/50).
Unfortunatelly, I am not smart enough to correctly interprete cards' specification. As an example, here is Hammerfall DIGI9636:
2 x ADAT digital I/O, based on RME's reliable Bitclock PLL
1 x SPDIF digital I/O, based on RME's reliable DIGI96 technology
1 x ADAT Sync In (9-pin D-type) for sample accurate transfers
1 x Breakout cable for coaxial SPDIF operation
Zero wait state PCI Busmaster interface with additional burst FIFO (130 MB/s transfer rate in both directions)
New hardware design: 36 mono channels in block mode, organized in 32 bit ASIO double buffers
ASIO poured in hardware! The result in speed and performance is 0 (zero!)% CPU load when using ALL 36 channels!
130 MB/s transfer rate for record and playback results in only 7% PCI bus load when using all 36 channels
New Enhanced Zero Latency Monitoring
S/MUX poured in hardware: 8 channels 96 kHz/24 bit for record and playback
If someone could help, or point me to the right reading material, I would greatly appreciate it.
Thank you,
M
I just read this entire thread in one sitting; made the fingers I use for scrolling quite sore
Kudos to MWP. Although it seems like to him the usb interface on his DAC-3 is understandably one of the less glamorous parts of the system compared to removing chips from an m-audio card and pulling off I2S signals, it is the closest thing I have found to a DIY USB audio adapter. (I'm a poor student just wanting decent quality from my laptop on the cheap, and searching for a usb adapter brought me to this thread)
My current prediction is that the PC will eventually become the center of high end home and DIY audio systems, and the current reasons people use computers for music don't even enter into my thinking. (i.e. though I do use a mac, it is not a "digital hub" rant)
I will explain myself in a new thread, still in the digital section. This thread seems to deal primarily with projects that could be built with existing IC's and software; my idea is different enough that it is probably better in another thread.
Thanks for the great read you guys! I'll be on the lookout for IC's solve the kinds of problems we have, and also look forward to free ports of filtering software. BTW, matlab is pretty much the standard for filter design, at least here at the University of Illinois, Urbana-Champaign. I haven't seen any other software packages that compare; DSP researchers here all use MATLAB/simulink to prototype algorithms.
Drew Wagner
Kudos to MWP. Although it seems like to him the usb interface on his DAC-3 is understandably one of the less glamorous parts of the system compared to removing chips from an m-audio card and pulling off I2S signals, it is the closest thing I have found to a DIY USB audio adapter. (I'm a poor student just wanting decent quality from my laptop on the cheap, and searching for a usb adapter brought me to this thread)
My current prediction is that the PC will eventually become the center of high end home and DIY audio systems, and the current reasons people use computers for music don't even enter into my thinking. (i.e. though I do use a mac, it is not a "digital hub" rant)
I will explain myself in a new thread, still in the digital section. This thread seems to deal primarily with projects that could be built with existing IC's and software; my idea is different enough that it is probably better in another thread.
Thanks for the great read you guys! I'll be on the lookout for IC's solve the kinds of problems we have, and also look forward to free ports of filtering software. BTW, matlab is pretty much the standard for filter design, at least here at the University of Illinois, Urbana-Champaign. I haven't seen any other software packages that compare; DSP researchers here all use MATLAB/simulink to prototype algorithms.
Drew Wagner
A few weeks ago, I asked whether anyone had tried compiliing BruteFIR under OS X. I tried it. It didn't work. I bought a couple of books on C and sniffed around the Apple mailing lists and osxaudio.com. I "fixed" a couple of the things that kept BruteFIR from compiling on the Mac, but haven't gotten it running yet, but I haven't given up on the idea yet, either. I don't think it would be difficult for someone who knew C.
BruteFIR uses FFTW. I've read that the FFT routines in vDSP (a maths library built into OS X) are up to 5 times faster than FFTW on the G5. There are already several plug-ins (Audio Units) for delay, paramatric EQ, compressors, expanders, and a multi-channel convolution engine (Space Designer in Logic Pro) etc. - all optimized for the G5. So, all that's needed is an Audio Unit to implement filters designed in MATLAB. I believe MATLAB will export either C header files or the coefficients for any filters designed in MATLAB. So, I'm thinking the best solution is to write an Audio Unit that takes the C header file from MATLAB and runs it against the vDSP FFT routine. I'm interested in writing this AU, but, as you can imagine, not knowing how to program has slowed me down. But, I've been assured through the Apple mailing list that this is possible, and that it shouldn't be too difficult to implement.
I also found some more information on minimum phase filters -- another question I asked a few weeks ago, in this "book", towards the bottom is a section entitled "Minimum Phase Digital Filters":
http://ccrma.stanford.edu/~jos/filters/
Lots of other good information there, and in a book I bought called "Digital Signal Processing: A Practical Guide for Engineers and Scientists" by Steven W. Smith.
BruteFIR uses FFTW. I've read that the FFT routines in vDSP (a maths library built into OS X) are up to 5 times faster than FFTW on the G5. There are already several plug-ins (Audio Units) for delay, paramatric EQ, compressors, expanders, and a multi-channel convolution engine (Space Designer in Logic Pro) etc. - all optimized for the G5. So, all that's needed is an Audio Unit to implement filters designed in MATLAB. I believe MATLAB will export either C header files or the coefficients for any filters designed in MATLAB. So, I'm thinking the best solution is to write an Audio Unit that takes the C header file from MATLAB and runs it against the vDSP FFT routine. I'm interested in writing this AU, but, as you can imagine, not knowing how to program has slowed me down. But, I've been assured through the Apple mailing list that this is possible, and that it shouldn't be too difficult to implement.
I also found some more information on minimum phase filters -- another question I asked a few weeks ago, in this "book", towards the bottom is a section entitled "Minimum Phase Digital Filters":
http://ccrma.stanford.edu/~jos/filters/
Lots of other good information there, and in a book I bought called "Digital Signal Processing: A Practical Guide for Engineers and Scientists" by Steven W. Smith.
Programming and dsp
I too wish for tools that don't exist (for me it's wrt audio measurement tools for Linux/Jack) and so I've been mulling over bootstrapping my programming and dsp skills. This would be a satisfying direction for a hobby but it's a long road and would get in the way of other things in life that seem more important.
dc -
Anyway, the point I need to make is that BruteFIR is just one small part of the computer based dsp problem. BruteFIR is great and deserves porting to OS X and everywhere else, but you will need "digital room correction" and measurement tools if you're really serious.
I noticed Fuzzmeasure, a new MLS measurement program has come out for OS X and would be a valuable tool for measurement of your audio system. But you are definitely going to want to make correction filters for nonideal speaker and room responses. You may end up porting Denis Sbragion's DRC also
-Robert
I too wish for tools that don't exist (for me it's wrt audio measurement tools for Linux/Jack) and so I've been mulling over bootstrapping my programming and dsp skills. This would be a satisfying direction for a hobby but it's a long road and would get in the way of other things in life that seem more important.
dc -
Anyway, the point I need to make is that BruteFIR is just one small part of the computer based dsp problem. BruteFIR is great and deserves porting to OS X and everywhere else, but you will need "digital room correction" and measurement tools if you're really serious.
I noticed Fuzzmeasure, a new MLS measurement program has come out for OS X and would be a valuable tool for measurement of your audio system. But you are definitely going to want to make correction filters for nonideal speaker and room responses. You may end up porting Denis Sbragion's DRC also
-Robert
Robert--
Thanks for the note. Even though I neglected to mention it in my last post, the point of having the multi-channel convolution engine is to use DRC. I'll use ETF (www.etfacoustic.com) on a PC for measurements. It spits out an impulse response file that can be used with DRC. I don't think that there will be any need to port DRC -- I can run DRC on a PC, take the inverse impulse response it creates and feed it to Space Designer in Logic Pro 7 (Space Designer is the name for Logic's multi-channel convolver).
My desire to move to the Mac stems from the lack of driver support under Linux. I have a Lynx Two-B that I don't want to give up. There are no ALSA drivers for it, and I'm unsure of the quality / stability of the OSS drivers (or OSS itself, for that matter). Ideally, I'd like to get the AES-16 (also no Linux driver support) and an Aurora 16, but only after I've successfully finished the Audio Unit programming. I plan to use a PC as a music and DVD server, and the Mac as a "loudspeaker management" tool. The finished product will have all of the functionality of a DEQX unit with higher performance and much more control over the filter slopes, level of correction, etc. I think the DEQX is a great piece of hardware, but it gets expensive when you start looking at it for all channels of a home theater setup, or 4-way active speakers (my mains are 4-way, which would require two DEQX units). Plus, upgrading my proposed setup when newer. faster. better hardware becomes available will be a lot cheaper than buying all new DEQX units. I'm also thinking that new technologies like wave field synthesis and correction algorithms that create a wider sweet spot will be easier to implement on a computer than waiting for just the right piece of hardware to be marketed.
Someone posted on their success with AlmusVCU on this thread yesterday:
http://www.avsforum.com/avs-vb/showthread.php?s=&postid=4781547#post4781547
brad
Thanks for the note. Even though I neglected to mention it in my last post, the point of having the multi-channel convolution engine is to use DRC. I'll use ETF (www.etfacoustic.com) on a PC for measurements. It spits out an impulse response file that can be used with DRC. I don't think that there will be any need to port DRC -- I can run DRC on a PC, take the inverse impulse response it creates and feed it to Space Designer in Logic Pro 7 (Space Designer is the name for Logic's multi-channel convolver).
My desire to move to the Mac stems from the lack of driver support under Linux. I have a Lynx Two-B that I don't want to give up. There are no ALSA drivers for it, and I'm unsure of the quality / stability of the OSS drivers (or OSS itself, for that matter). Ideally, I'd like to get the AES-16 (also no Linux driver support) and an Aurora 16, but only after I've successfully finished the Audio Unit programming. I plan to use a PC as a music and DVD server, and the Mac as a "loudspeaker management" tool. The finished product will have all of the functionality of a DEQX unit with higher performance and much more control over the filter slopes, level of correction, etc. I think the DEQX is a great piece of hardware, but it gets expensive when you start looking at it for all channels of a home theater setup, or 4-way active speakers (my mains are 4-way, which would require two DEQX units). Plus, upgrading my proposed setup when newer. faster. better hardware becomes available will be a lot cheaper than buying all new DEQX units. I'm also thinking that new technologies like wave field synthesis and correction algorithms that create a wider sweet spot will be easier to implement on a computer than waiting for just the right piece of hardware to be marketed.
Someone posted on their success with AlmusVCU on this thread yesterday:
http://www.avsforum.com/avs-vb/showthread.php?s=&postid=4781547#post4781547
brad
FWIW, 4Front announced beta support for the AES16 in their latest set of drivers. When I tried the commercial OSS for my Delta 1010, I found them to work as advertised. With both Jack and BruteFIR having OSS modules, I think it would be well worth the experiment to try the OSS drivers - they have a free eval period.
AVS HTPC Forum
You should right away (if you haven't already) go over to AVS Forum and look in on their progress on using PC's as home theater and music systems. They're doing things we've only dreamed of. Just follow the link in dc's post a couple back.
Meanwhile, I've just rearranged my architecture a bit. You can read about it at the bottom of the thread in AVS
-Robert
You should right away (if you haven't already) go over to AVS Forum and look in on their progress on using PC's as home theater and music systems. They're doing things we've only dreamed of. Just follow the link in dc's post a couple back.
Meanwhile, I've just rearranged my architecture a bit. You can read about it at the bottom of the thread in AVS
-Robert
There's been some progress on this thread:
http://www.avsforum.com/avs-vb/showthread.php?s=&postid=5054723#post5054723
http://www.avsforum.com/avs-vb/showthread.php?s=&postid=5054723#post5054723
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.