Since some years I have been working on a DSP plugins to correct loudspeakers. I have now completed a full project based on full range driver (Markaudio Alpair 7.3) & VST plugin for Windows OS.
As the result is more than encouraging I have decided to share this experience and created a dedicated web site. It explains the concepts and how I have implemented corrections with the DSP plugin (magnitude/phase/frequency intermodulation) . A demo version of the plugin can be downloaded for testing.
This concept can be adapted to any full range loudspeaker to improve greatly their performance.
https://sites.google.com/site/dspdpath/home
As the result is more than encouraging I have decided to share this experience and created a dedicated web site. It explains the concepts and how I have implemented corrections with the DSP plugin (magnitude/phase/frequency intermodulation) . A demo version of the plugin can be downloaded for testing.
This concept can be adapted to any full range loudspeaker to improve greatly their performance.
https://sites.google.com/site/dspdpath/home
Pro studio grade plugins comparison
Hi TT,
I agree, full range with good Eq is far superior to multi way with crossovers.
I found it easier to use off the shelf VST plugins as they are free or very low cost.
Have you compared your Eq with the free ones in JRiver?
Also both Fab Filter and Blue Cat Audio have full studio mastering grade, linear phase plugins available at low cost...Have you tried them?
Much of the music we listen to is mastered on equipment using Fab Filter plugins....
Thanks in advance
Derek.
Hi TT,
I agree, full range with good Eq is far superior to multi way with crossovers.
I found it easier to use off the shelf VST plugins as they are free or very low cost.
Have you compared your Eq with the free ones in JRiver?
Also both Fab Filter and Blue Cat Audio have full studio mastering grade, linear phase plugins available at low cost...Have you tried them?
Much of the music we listen to is mastered on equipment using Fab Filter plugins....
Thanks in advance
Derek.
Hi Derek,
The equalizers you are refeering to are focused on level equalization. My objective was to include also group delay & frequency intermodulation corrections (to improve bass response & limit "doppler" effect)
Group delay equalization can be done with a convolver (a tool like "rephase" can be used to design the associated FIR filter)
Doppler correction is really specific and I don't know any plugin with such feature
Regards,
Thierry
The equalizers you are refeering to are focused on level equalization. My objective was to include also group delay & frequency intermodulation corrections (to improve bass response & limit "doppler" effect)
Group delay equalization can be done with a convolver (a tool like "rephase" can be used to design the associated FIR filter)
Doppler correction is really specific and I don't know any plugin with such feature
Regards,
Thierry
Hi!
I like to go on the DSP path, first with a fullrange and later with a multiway system. But I get a bit confused by the enormity off info on DSP. Do you folks know a good place to start?
The speaker(s) in question are Philips ad9710/m8 and Philips Ad12100/m8.
Grt,
Raymond
I like to go on the DSP path, first with a fullrange and later with a multiway system. But I get a bit confused by the enormity off info on DSP. Do you folks know a good place to start?
The speaker(s) in question are Philips ad9710/m8 and Philips Ad12100/m8.
Grt,
Raymond
Thierry, I could see from what of your measurements that the doppler correction works. I wonder about how obvious is it when listening? I mean, applying only the doppler correction, not the EQ. A subtle or a big improvement?
Very nice work. I noticed you spent a lot of effort on developing your own model for the in room response from the driver, port, and baffle. All of these are very well modeled in other programs like AkAbak. Is there a need to model it in your program or could one not just use the results from another package?
The IM correction is interesting and does indeed appear to work but the example is for two frequency sinusoid. Music is complex and if decomposed into frequency components will have at least 32 (as in the case of the popular mp3 implementation). How does the DSP correct for 32 simultaneous sine tones? As you say, the way to handle this is to introduce a delay proportional to cone position - does this automatically satisfy all frequencies that are being affected by the IM distortion?
The IM correction is interesting and does indeed appear to work but the example is for two frequency sinusoid. Music is complex and if decomposed into frequency components will have at least 32 (as in the case of the popular mp3 implementation). How does the DSP correct for 32 simultaneous sine tones? As you say, the way to handle this is to introduce a delay proportional to cone position - does this automatically satisfy all frequencies that are being affected by the IM distortion?
Member
Joined 2009
Paid Member
Sounds very interesting.
fyi - I'm not convinced that IM products are from Doppler. There are other non-linear processes that can produce IM. Doppler needs a moving source. As both the high and low frequencies are generated by movement of the cone with respect to a stationary magnet there should be no Doppler. However, high frequencies produced by cone break-up modes, common in full range drivers, is movement of the cone with respect to the average cone surface which is moving and hence can produce Doppler.
fyi - I'm not convinced that IM products are from Doppler. There are other non-linear processes that can produce IM. Doppler needs a moving source. As both the high and low frequencies are generated by movement of the cone with respect to a stationary magnet there should be no Doppler. However, high frequencies produced by cone break-up modes, common in full range drivers, is movement of the cone with respect to the average cone surface which is moving and hence can produce Doppler.
Hi,
To answers some questions:
- the IM distorsion reduction is for sure of second order vs other corrections. Using this plugin (w/o level correction http://fresponse.free.fr/dPath_wo_level_corr.dll and listening with good headphones you can simulate the effect of IM by playing with the "doppler gain" volume
- The algorithm works in time domain and just compensate for the sound travel difference linked to cone position. Sound content will only affect the cone displacement. This is why IM distorsion if more critical for music with a combination of bass & treble. I found a good test with Tin Pan Alley by Steevie Ray Vaughan (subjectively the precision of guitar attacks are affected by the bass sounds when played on a full range driver)
- I have develop my own models because I was not able to find a product with all features, especially precise simulation of box diffraction
To answers some questions:
- the IM distorsion reduction is for sure of second order vs other corrections. Using this plugin (w/o level correction http://fresponse.free.fr/dPath_wo_level_corr.dll and listening with good headphones you can simulate the effect of IM by playing with the "doppler gain" volume
- The algorithm works in time domain and just compensate for the sound travel difference linked to cone position. Sound content will only affect the cone displacement. This is why IM distorsion if more critical for music with a combination of bass & treble. I found a good test with Tin Pan Alley by Steevie Ray Vaughan (subjectively the precision of guitar attacks are affected by the bass sounds when played on a full range driver)
- I have develop my own models because I was not able to find a product with all features, especially precise simulation of box diffraction
Isn't IM really more due to beat frequency generation - a non-linear process? I think if the cone movement is perfectly linear then this effect would not be present. It is due to non ideal, non-linear effects where the IM product arises from.
I won't pretend to know exactly why IMD happens but I always assumed it is simply because you have dramatically different frequencies being produced by the same surface. For example, you have a transducer moving at the proper rate to deliver 50Hz, but then you also ask it to produce 10kHz. Somehow in my mind, this now becomes a transducer producing 10kHz which is moving (Doppler) both forward and backward, to a much greater extent, at a rate of 50Hz. This seems like it would apply equally to both linear and non-linear cone movements.
My tiny brain also cannot imagine how software could change this. 🙂
My tiny brain also cannot imagine how software could change this. 🙂
Cone based IM is just the tip of the ice berg...
Hi Cogitec,
I a agree with your comments. In addition I think the problem is even worse because the cone / voice coil / suspension is all ocilating like a mass on a spring. Contrary to what driver manufacturers try to distract us with ie the cone surface may be pistonic within a given band width up to certain SPLlevels, BUT....The entire assembly is bouncing around like a mass on a spring, the cone never does a clean " Start, Stop, Return to Centre"...It over shoots and undershoots trying to loose the Kinetic energy....
Cheers
Derek.
Hi Cogitec,
I a agree with your comments. In addition I think the problem is even worse because the cone / voice coil / suspension is all ocilating like a mass on a spring. Contrary to what driver manufacturers try to distract us with ie the cone surface may be pistonic within a given band width up to certain SPLlevels, BUT....The entire assembly is bouncing around like a mass on a spring, the cone never does a clean " Start, Stop, Return to Centre"...It over shoots and undershoots trying to loose the Kinetic energy....
Cheers
Derek.
Rambling explanation...!
If anyone has too much time on their hands you might want to read a longer version of my "mass on a spring" rant"....Enjoy
One last point....
The amp only controls the voice coil when there is current in the coil ie when a signal is being sent..." So " bang" there is the drum strike...One single impluse to compress the air, then zero signal...Decay back to ambient, no restoring force, the cone is not pulled back and stoped, it bounces around until energy decay stops it.
The idea that the amplifier damping factor / back EMF issue stops the mass on a spring behaviour is flawed...The effect is tiny, its related to the efficiency of the driver ( or lack of) ie most cone / dome drivers are less than 1% efficient. Less than 1% of the centering or restoring to centre force come from the amp....99% of the kinetic energy is burned of as voice coil heat and suspension friction losses.
To test this switch off your amp and gently(!) push in the cone, feel the resistance. Now repeat with the amp switched on....No difference!
Cheers
Derek.
If anyone has too much time on their hands you might want to read a longer version of my "mass on a spring" rant"....Enjoy
One last point....
The amp only controls the voice coil when there is current in the coil ie when a signal is being sent..." So " bang" there is the drum strike...One single impluse to compress the air, then zero signal...Decay back to ambient, no restoring force, the cone is not pulled back and stoped, it bounces around until energy decay stops it.
The idea that the amplifier damping factor / back EMF issue stops the mass on a spring behaviour is flawed...The effect is tiny, its related to the efficiency of the driver ( or lack of) ie most cone / dome drivers are less than 1% efficient. Less than 1% of the centering or restoring to centre force come from the amp....99% of the kinetic energy is burned of as voice coil heat and suspension friction losses.
To test this switch off your amp and gently(!) push in the cone, feel the resistance. Now repeat with the amp switched on....No difference!
Cheers
Derek.
The whole mass on a spring is true but you forget that there is a huge effect of damping from both the suspension and the enclosure which really depends on how well the impedance matching between the cone and air it moves. Testing by pushing with your finger is way out of bandwidth - try quickly tapping because the cone and enclosure are tuned and work in their bandwidth of say 50 Hz to 20 kHz. The TS parameters and enclosure such as a BLH or BR all work together and this all comes out when the impulse response of the speaker is measured. It is the ability to have a clean impulse response that makes a speaker able to reproduce transients. There are under damped and over damped suspensions and enclosures - hence the Q uses to describe boxes. I am just saying that you have to take the whole system and it is not so simple as a mass on a spring. More like mass on a spring connected to a damper that is dynamically controlled via fluid-dynamics and acoustical response combined with a forcing input from a voice coil actuator sometimes mounted on an aluminum former that has eddy current braking.
Try pushing cone of a servo drive sub - it will resist. 🙂
Try pushing cone of a servo drive sub - it will resist. 🙂
Last edited:
X,
Right, however impulse response testing is limited by the fact that only single frequencies (or graduated "sweeps") are tested.
I'd like to see an impulse response test of a single transducer simultaneously producing 50Hz and 10kHz. Each will affect each other via IMD. The 10kHz signal is being produced by a surface that is moving considerably to and fro at 50Hz - I cannot imagine that there is zero effect on the 10kHz signal (but perhaps I am wrong). Vise versa is also true, it seems. Test the impulse response of 50Hz alone, then compare it to the 50Hz impulse response while the surface is also tasked with producing 10kHz. My gut says the 50Hz response in these two scenarios will not be identical.
Now take the more complex signal of even a simple three-piece jazz ensemble. With all that going on, it is a wonder it even sounds like music.
Take it to the extreme, and the result is the common assertion that full-range drivers simply fall apart on very complex music, particularly when played at >medium SPL.
Right, however impulse response testing is limited by the fact that only single frequencies (or graduated "sweeps") are tested.
I'd like to see an impulse response test of a single transducer simultaneously producing 50Hz and 10kHz. Each will affect each other via IMD. The 10kHz signal is being produced by a surface that is moving considerably to and fro at 50Hz - I cannot imagine that there is zero effect on the 10kHz signal (but perhaps I am wrong). Vise versa is also true, it seems. Test the impulse response of 50Hz alone, then compare it to the 50Hz impulse response while the surface is also tasked with producing 10kHz. My gut says the 50Hz response in these two scenarios will not be identical.
Now take the more complex signal of even a simple three-piece jazz ensemble. With all that going on, it is a wonder it even sounds like music.
Take it to the extreme, and the result is the common assertion that full-range drivers simply fall apart on very complex music, particularly when played at >medium SPL.
Hi,
You can see a measurement of combined 60Hz & 3000 Hz on my alpair7s
https://sites.google.com/site/dspdpath/concepts/implementation
The main emerging frequencies around the central 3000 Hz are coming from "doppler"
As you say this is normal as the position of the acoustical source is not fixed. Considering an acoustical pressure source with function of time P0(t), the acoustic pressure at a distance x of this source is P1(t)=P0(t-x/c0) where c0 is the speed of sound.
If the source is fixed x is a constant and a the waveform of the source will be preserved at the listening position. If x is not constant it is easy to see that there will be a modulation as the P1 function will become P0(t-x(t)/c0) where x(t) is not constant
IMD correction with digital processing is basically to reverse this formula using an estimation of x(t) i.e. cone position
You can see a measurement of combined 60Hz & 3000 Hz on my alpair7s
https://sites.google.com/site/dspdpath/concepts/implementation
The main emerging frequencies around the central 3000 Hz are coming from "doppler"
As you say this is normal as the position of the acoustical source is not fixed. Considering an acoustical pressure source with function of time P0(t), the acoustic pressure at a distance x of this source is P1(t)=P0(t-x/c0) where c0 is the speed of sound.
If the source is fixed x is a constant and a the waveform of the source will be preserved at the listening position. If x is not constant it is easy to see that there will be a modulation as the P1 function will become P0(t-x(t)/c0) where x(t) is not constant
IMD correction with digital processing is basically to reverse this formula using an estimation of x(t) i.e. cone position
First of all, this is just phase modulation, not necessarily a Doppler effect, as stated on the westhost pages. Secondly, how do you apply this correction? Is it a correction for the peak displacement at a single modulation frequency?
I was waiting for someone to jump on the Doppler thing. No this is not Doppler effect. These are the side bands that you get when you mix two frequencies. As long as the frequencies are mixed, and they have to be mixed if they are produced from the same diaphragm, you will get side bands.
Bob
Bob
The old Philips motion feedback speakers were impressive. A pity they disappeared from the scene.
Take a closer look at a wave form and you'll get the idea of how this can work...
An externally hosted image should be here but it was not working when we last tested it.
It would be interesting to see an original wave form with two simultaneous tones and the same one compensated by this algorithm.
After that we can look at a real music signal 😉.
Last edited:
- Status
- Not open for further replies.