There has been a lot written about predistortion over years but I have yet to a see a practical working implementation when it comes to speakers. The problem is that the speakers characteristics change over time and with different power levels etc. There are just too many variables to make a reliable inverse model. In any event some sort of sensor is still needed to measure the speaker such as an accelerometer or precision microphone so you may as well go the extra distance and employ feedback of some kind
Looking at it another way. Would you apply this methodology to an audio amplifier when it would be simpler or more prudent to use negative feedback ?
regards
david
Last edited:
A simple sensor makes a lot of sense, it just wasn't an idea I had considered! can't help but wonder if an optical sensor of some kind could be developed... I know that laser pickups have been successfully developed for use on guitar...
My thought process is that the way we are heading in consumer hifi is to all in one speaker systems:
wireless receiver>>DSP (EQ, Digital xo, distortion reduction [feedback from optical sensor?])>>DAC>>digital amplification>>drive units
starting to wonder if the DAC needs to move though
My thought process is that the way we are heading in consumer hifi is to all in one speaker systems:
wireless receiver>>DSP (EQ, Digital xo, distortion reduction [feedback from optical sensor?])>>DAC>>digital amplification>>drive units
starting to wonder if the DAC needs to move though
I have evaluated optical sensors for the use of feedback, but a 20 dollar accelerometer was the most sensible way to go.
I looked at this sensor;
Extreme dynamic laser sensor with anti speckle feature | Micro-Epsilon America
The use of such sensor brings in more trouble than that of mounting a accelerometer to your woofer.
I looked at this sensor;
Extreme dynamic laser sensor with anti speckle feature | Micro-Epsilon America
The use of such sensor brings in more trouble than that of mounting a accelerometer to your woofer.
I think we can agree on that
The ACH01 is a great little device for this purpose. David, have you implemented the gain stage from the spec sheet?
The measurements of the speaker that are made by the software are inclusive of the additional weight of the accelerometer so everything is accounted for
These days most woofers have cones that weigh between 100-500 grams depending on the size of the driver so an extra few grams of mass from the accelerometer is barely going to make any difference from an efficiency standpoint. It may lower the resonant frequency by a whisker but that is still not an issue in this design as any changes are accounted for regards
david
Last edited:
If you are talking about using a secondary voice coil wound around the main drive voice coil as a motion sensing device then this is plagued with problems. When I was debugging the accelerometer I needed something to compare it with. I used one of the voice coils from a dual voice coil speaker and drove the other coil with an amplifier. Noting that the sensing coil should output a signal proportional to velocity so it was just a matter of integrating the accelerometer signal to equate the two. Things checked out right at low frequencies but at higher frequencies above 100Hz the amplitude of the signal from the second voice coil was constant all of the time and then it struck me that I was essentially measuring a transformer and not a velocity sensing device. Because the sense coil and drive coil are tightly coupled the velocity signal is swamped by the drive signals coupling across at higher frequencies
Not only that, because the sensing coil and drive coil sit in the same gap, the sensing coil experiences the same flux modulation as does the main drive coil. The modulation in inductance and flux as the voice coil moves in and out of the gap also causes distortion of the very signal you are trying to measure. Even if you use constant current drive to avoid the inductance it doesn't help because the inductance rears its ugly head because of mutual inductance between the coils !! You don't have this problem with an accelerometer because it only responds to motion and not magnetic coupling
All in all this type of sensing arrangement is almost useless for a high quality servo system which probably explains why Infinity, Genesis, Velodyne, Phillips never used it. I think Infinity only ever used it once back in 1968 in their first Servo Statik speaker but all servo speakers after that used accelerometers
Even the original Phillips article on motional feedback speakers published way back in 1967 more or less came to the same conclusions. Those dudes at Phillips knew all about it then I'm not saying that accelerometers don't have their own set of problems but you can build out the problems if you know what issues you are dealing with
regards
david
Last edited:
Bang for the buck you still can't beat the ACH-01-03. It comes with an integral 40 inch flexible shielded coax lead which is needed in this application. I couldn't even find anyone that sold that special type of flexible lead.
Unfortunately it has gone back up to $40 for some unknown reason
regards
david
Given that most gross distortion products are consistent from driver to driver (not perfectly matched, but close), I wonder if one could use this on a single driver, but wire the amplifier output to a number of series/parallel identical drivers. Naturally it wouldn't as effective, but could be a good way to go. Another interesting option would be to mount it on a coupling ring, from former to former, in a clamshell isobaric pair. Not only would the coupling ensure proper distortion reduction from the isobaric mounting by eliminating the effect of the air compliance between drivers, the feedback loop would not have to work so hard against the even-order components.
The measurements of the speaker that are made by the software are inclusive of the additional weight of the accelerometer so everything is accounted for
regards
david
Used peizo's for accelerometer feedback back in the mid 80's for the same reasons you've given. Problem encountered was the mass of the element. Unlike cones of today I was using much lighter weight cones <50gms. Turned to cutting my own elements to keep mass down. The lighter the better. Today I'd like to do the same, but the Mms of the test driver is <14gms!
I gave up on the idea then due to the cost component. To do as close to what you've done would have been >$50k and a few percent of your results in the final analysis. Simply too ambitious for the day.
This was the antithesis of my last design some 20 years ago. A rather odd MMTMMM with a dual 10" isobaric sub floor coupled 44" tower. Using 5/ 5" fullranges crossing at 200 and 500 (mixed duty) and 3k at the top end between the center two mids tweeter handoff. OMG complex crossover.
I do have pics, but wifey, whom back then wasn't so pleased with my way (job related), did those in, between moves from one state to another. Only heard of the carnage from a friend a few years later.
Pics are MIA, hiding in one of hundreds of unmarked boxes full of stuffed toys and collectables in one of our storage units. Been 13 years since last seen and know they are there 
Crossovers survived tho! Yet another saving grace of external networks!!!
Those parts are... LOL... in yet another unmarked box in another ******* storage unit... need that box dagnabit! Where does it end?! 
Last edited:
was more thinking for application to midrange, a lighter accelerometer would be needed...
PCB Model 352B10
maybe.. haven't gone through the specs properly yet
PCB Model 352B10
maybe.. haven't gone through the specs properly yet
How the hell did you ever manage to build a crossover for all that?!!
Almost sounds like torture, I'd love to see them.
So did you wife take an axe to them or something?
I think I need to hide my speakers in the loft in that case...
As long as the drivers are very similar (ie same model size etc) you can use multiple drivers using one as the control woofer with attached accelerometer and the other drivers as slaves. They would share the same enclosure volume and be connected in series or parallel or a combination of parallel and series connection so that the same current flows through each driver. This is how it is done in the Infinity IRS,Beta,Gamma etc. There is no need to use a servo for each driver.
regards
david
Depending upon what box you put it in and what amp you hook it up to...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.