Using PC's for adaptive equalization - diyAudio
Go Back   Home > Forums > Source & Line > Digital Source

Digital Source Digital Players and Recorders: CD , SACD , Tape, Memory Card, etc.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 9th December 2004, 07:29 AM   #1
diyAudio Member
 
Join Date: Nov 2004
Location: Urbana, IL
Send a message via AIM to drewm1980
Default Using PC's for adaptive equalization

There has been some discussion on this forum regarding what role, if any PC's will play in audio systems in the future. I believe that PCs will likely play a huge role in DIY systems in the future. This will have little to do with library convenience and music archival and everything to do with sound quality. I believe that adaptive signal processing will change the face of home audio systems forever; it just has not caught on yet.

Consider the following vision/prognostication:

As a DIYer, you rip the crossovers out of a few old speakers, which collectively have about 7 mismatched speaker cones. You build seven cheap gainclones, one for each transducer. You plug each gainclone into a board that has a IEEE1394 receiver chip and as many DAC's as you decided to solder onto your PCB (which someone else designed and can hold may DACs though you don't have to populate it all the way) and one ADC.

You buy a cheap microphone of a popular model and plug it into ADC. You plug the CODEC board into your iMac G5 or whatever PC you already have laying around and hang the mic from the ceiling about where your head would be.

You install open source software that has been hacked together collaboratively by DIYers all over the world, grad students at engineering universities as course projects, etc.

You tell this software a few things, like which speakers correspond to which signals in your digital music source and which model microphone you are using (other DIYers have posted models for the frequency response of several common mics).

If you're on a mac, you hit a big pulsing blue button labeled "Calibrate" and if you're on Linux you run a shell script with a name like "sp34kr_c4llbr8_2349x293.ruby" with 13 mostly documented parameters and walk away

The system automatically determines the frequency response (Amplitude AND phase) of each speaker/amp/dac, how much of each frequency each speaker/amp/dac can handle without non-linear distortion.

The system then plays a recording of Robin Williams "Good Morning Vietnam!!!" or something and you know it is save to come back into the room.

The software has determined an appropriate FIR (this will surely all be linear processing at least at first) filter for the output from each source signal to the listener, and you now have a fully calibrated and equalized "reference" system.

Since the overhead of having to tune each transducer and each amplifier carefully by hand has been removed, each transducer and each amp can be designed to work excellently and whatever portion of the frequency range it is best for.

Even the frequency response of your room has been compensated for, reducing the need to futz with speaker and furniture placement.

Of course, everyone has their own tastes. Once the computer has found its own baseline for your system, you can tinker with the filters all you want, re-equalizing certain recordings differently, etc. Since other DIYers out there have already built similar "reference" systems, you can share the filters you have come up with others.

The hardest part of this building this system will be designing and tuning the auto-calibration software, and fortunately that is the part of the system that can be created by specialists and shared with everyone for free.

Although I had the idea of closing the audio feedback loop independently a few years ago, I would be very surprised if I was the first, and several large american corporations have probably already patented it at the wonderful USPTO. I found one product implementing these ideas that uses custom hardware and is intended for a different market, but it should provide some indication of the technical feasibility of these ideas.

See the sabine REAL-Q2 Real-Time Adaptive Equalizer

http://www.sabine.com/newsite/realq.html

P.S. Active Room Equalization has also been discussed in another thread; this idea is different primarily in the 1 amp per cone and the use of PC's as the digital signal processor and the source rather than proprietary outboard hardware and algorithms. i.e. the Behringer Ultracurve Pro 8024 that was discussed in the other thread.

Anyone else out there share this vision?
  Reply With Quote
Old 9th December 2004, 08:25 AM   #3
diyAudio Member
 
Bill Fitzpatrick's Avatar
 
Join Date: Jun 2001
Location: Eugene, OR
I certainly don't share that vision.

Why would anyone want to build a system around 7 junk speakers?

I disagree with another point as well. You start messing around with EQ to correct for the room and you've destroyed the first arrival signal - the important one.
  Reply With Quote
Old 10th December 2004, 12:29 AM   #4
diyAudio Member
 
Join Date: Nov 2004
Location: Urbana, IL
Send a message via AIM to drewm1980
> I certainly don't share that vision.
>
> Why would anyone want to build a system around 7 junk speakers?

Thanks for the links wimms, and sorry Bob, I'll be more clear on what I meant. You wouldn't use junk speakers... When I said old I had in my head "old but high quality", and the cones are mismatched because they came from different sources, not a matched set. i.e. maybe a cone was damaged beyond repair but its sister is still good. The odd number was to point out the extra flexibility auto-calibration could buy you in flexibility of parts usage.

The idea I was trying to get across is that the overhead, primarily in design and tuning, for adding extra speaker cones would be greatly reduced. The idea is to get the best sound for the money, but using more, less expensive cones, and using your PC in place of commercial DSP based hardware.

The cones don't have to have a wide or uniform frequency response; you just have to have at least one cone per source channel that is good at reproducing any given frequency range. Thus, less expensive cones. Likewise, the amp doesn't have to have a uniform response either. Perhaps it would be possible to even do non-linear processing to cancel out the non-linear effect of a transistor that swings far from its designed bias point. The use of more, cheaper redundant elements could pay off in sound quality per dollar. (or maybe it never will, I don't see why not)

One of the reasons we naturally find simple designs to be more effective and reliable is that the engineer (you don't have to have a degree. if you can successfully build a stereo from parts you are an engineer in my eyes)

> I disagree with another point as well. You start messing around with EQ > to correct for the room and you've destroyed the first arrival signal - > the important one.
I haven't studied frequency response (if it is even reasonably linear) of echoing cavities yet. (I've only seen waveguides and resonant cavities in an E+M context.)

Why is the first arrival signal is (necessarily) destroyed? That is certainly a possibility with a poor implementation of the algorithm, but I don't see why it is a necessity for all such algorithms. Assuming an inevitable and noticeable effect on the first arrival, it may still be worth it:

If your playing space has an awful echo, the effect of the echo and problems with equalization will be worth correcting even if this does have a detrimental effect on the first arrival. If you have a great listening space, the equalization might not be worth it at all. One of the sound systems I want to improve the most is the PA we use at our weekly swing dances. We dance in the top floor of a church with a stone floor and a vaulted ceiling.

http://www.cu-swing.org/gallery/weekly/DSC04234_b

Our fender PA is on the table to the right. There is a fairly large powered subwoofer on the floor outside the frame. This is obviously an extreme case. Again, perhaps the average DIYer's listening space is good enough that equalization would never catch on.

Heck, I would think that even if you never actually use the equalization for playback, I would still expect it to be an excellent tool for designing and tuning your traditional passive crossovers and amps. i.e. if you assume your amps to be perfect, or you have a good model for its imperfections, you can use it with each cone separately, and the software will suggest an optimal location for the crossover point(s) for each speaker. There are a bunch of calibration schemes you could use... eventually the community would probably settle on standard routines/guidelines.

Having all of your crossover / equalization being done in a central, easily re-programmable place also gives you a lot of freedom to devise creative new methods of minimizing distortion.

Suppose you have two transducers that are both efficient in one frequency range, but complimentary distortion characteristics with respect to volume. However, one has low distortion only at high volumes while the other has low distortion only at lower volumes. Perhaps you could intelligently, perhaps even, distribute the power between the two transducers.

Maybe one cone has really high fidelity, but distorts at high volumes. Cones with neighboring frequency responses could help pick up the slack when the volumes are high even though they are less efficient power-wise in the mid frequencies.
  Reply With Quote
Old 10th December 2004, 12:55 AM   #5
diyAudio Member
 
Bill Fitzpatrick's Avatar
 
Join Date: Jun 2001
Location: Eugene, OR
I think you're grasping at straws here. Of course, what do I know?
  Reply With Quote
Old 10th December 2004, 01:09 AM   #6
paulb is offline paulb  Canada
diyAudio Member
 
Join Date: Jun 2001
Location: Calgary
Ignore the Luddites. I think your vision is worth pursuing. Mercator and I have been discussing a similar idea, although not as ambitious as yours. I like the idea of multi-amping and active crossovers. I like the idea of crossovers in software, particularly given that sources will all become digital (or will be ripped into being so). DSP (even if done on a PC) produces much better filters. I like the idea of PCs because they're cheap. Add in the autoequalization and it sounds pretty good.
Of course, it begs the question, how do you know what you like? I think your preferred settings would vary with source material. Maybe we need controls like soundstage and brightness and transparency instead of bass and treble.
I don't understand Bill's concern with first arrival signal - why would it be any different?
  Reply With Quote
Old 10th December 2004, 01:31 AM   #7
diyAudio Member
 
Join Date: Nov 2004
Location: Urbana, IL
Send a message via AIM to drewm1980
Quote:
Originally posted by Bill Fitzpatrick
I think you're grasping at straws here. Of course, what do I know?
Judging from your bio, probably a lot more than me about practical matters and the current state of technology. I'm still being steeped in academia, which means I'm full of wacky ideas about technology that would take years to be accepted (or become economical), in the unlikely chance that they get accepted at all.

That said, I think this is a pretty solid idea; the requisite technology is already available, though I don't know if it is economical (in time and money) for DIYers to implement something like this yet.
  Reply With Quote
Old 10th December 2004, 01:41 AM   #8
diyAudio Member
 
Bill Fitzpatrick's Avatar
 
Join Date: Jun 2001
Location: Eugene, OR
Quote:
Originally posted by paulb
Ignore the Luddites. . .

I don't understand Bill's concern with first arrival signal - why would it be any different?
If I was a Luddite I wouldn't have a computer now would I?

If you correct for rooms problems by equalizing you are altering the spectral balance of the first arrival signal. How could that statement be unclear?
  Reply With Quote
Old 10th December 2004, 01:59 AM   #9
diyAudio Member
 
Join Date: Nov 2004
Location: New York
Two cents from an audiophile.

I believe crossovers is really hard to tame. Adding the difference in units (tweeter, woofer.. etc), the sound from multi way speakers almost always feel altered to my ears. No matter what you do, it just doesn't sound smooth near/at the crossover point; at least not like full range units.

I think once the signal is altered you can't change it back, no matter how you do it. Even if the graph shows up correct, my ears would tell me otherwise.

I think the technology is not advance enough to fool the brain yet.
  Reply With Quote
Old 10th December 2004, 02:17 AM   #10
diyAudio Member
 
Join Date: Nov 2004
Location: Urbana, IL
Send a message via AIM to drewm1980
Quote:
Originally posted by paulb
[B]Ignore the Luddites. [\B]
We stand on the shoulders of giants.

I think your vision is worth pursuing. Mercator and I have been discussing a similar idea, although not as ambitious as yours. I like the idea of multi-amping and active crossovers. I like the idea of crossovers in software, particularly given that sources will all become digital (or will be ripped into being so). DSP (even if done on a PC) produces much better filters. I like the idea of PCs because they're cheap. Add in the autoequalization and it sounds pretty good.
Of course, it begs the question, how do you know what you like? I think your preferred settings would vary with source material. Maybe we need controls like soundstage and brightness and transparency instead of bass and treble.
I don't understand Bill's concern with first arrival signal - why would it be any different?
With a PC as both your source and an active element in the audio loop, it is easy to do things like having settings that are persistent for the recording. EQ settings can get stored in ID3 tags and shared between users a-la CDDB. Only thing that prevents that now is that so few people have callibrated systems, eq settings have no meaning. Using a PC to close the feedback loop allows everyone to have a callibrated system.

I haven't spent a lot of time having deep discussions about subjective/artistic qualities as soundstage, brightness, transparency, so I don't ~really understand what people mean by them. I have vague notions, but have not lived enough to be comfortable discussing them, let alone comment on and if/how you could improve them independently through DSP. I know what I like when I hear it, but I don't have a language for it yet.

We always start from engineering critera of good, and then add the human / artistic element. Using a PC for closed loop DSP only provides a new framework that seems to me to make a lot engineering sense, and allows a whole new playground for people to express themselves artistically through creative design. This time the design just happens to be mostly on a computer where most of what you can think up is possible.

Have to go design differential amplifiers... (this time it's homework)
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
78 equalization? Justinasia Analogue Source 12 22nd September 2012 02:02 AM
Equalization Vaughan Car Audio 6 27th August 2007 04:33 PM
Adaptive class A jane Tubes / Valves 4 19th March 2007 01:35 AM
adaptive bias theory Leolabs Solid State 24 17th October 2006 05:02 AM
active equalization help downward_dog Digital Source 0 2nd June 2005 11:19 PM


New To Site? Need Help?

All times are GMT. The time now is 09:36 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2