Experiments with the current drive

  • Thread starter Thread starter Deleted member 375592
  • Start date Start date
^Voltage drive does not provide damping above driver resonance, because of mass (not being cancelled by spring) makes (disturbance) force to accelerate the mass so any electrical damping that would counter it is at least 90deg late and thus not effective, perpendicular force or actually extra excitement rather than damping. On resonance electrical damping works, as force goes to velocity directly. See https://www.edn.com/loudspeaker-operation-the-superiority-of-current-drive-over-voltage-drive/ , jump to heading "The assumed control of cone motion"
 
  • Like
Reactions: Hydrogen Alex
But driver resonances aren't strictly confined to the bass. Smaller impedance glitches, indicative of a resonance, are often seen around a few hundred Hz to 1kHz. There's likely a cut-off, of sorts, but I expect it has a first order slope approaching 20dB per decade.

Attempting a thoughtful guess, I would expect resonant 'nodes' and 'antinodes' (referring to regions of maximum velocity vs regions of maximum pressure, for air resonances, like in a transmission line, or various spider or cone resonances) to basically switch places, or undergo some degree of inversion, when comparing the extremes of near-zero or near-infinite output impedance.

At one extreme, external excitation of the voice coil does not cause any current to flow, so it's like an open-ended wind instrument, or the sharp end of an antenna. At another extreme, a short-circuited voice coil produces maximum braking force, but contrary to misleading language falsely implying that "damping" is at a maximum, a different mode of resonances can occur. The coil becomes a 'node' with minimum velocity and maximum pressure, and reflects a lot of mechanical energy.

Notice how blocking the end of a flute-like instrument often does not reduce resonance, it merely changes the resonant frequency. And so, with speakers there can be some competing criteria, like minimum harmonic distortion, versus minimum resonance, and the latter is probably achieved most effectively with a classic, tube-like middling output impedance of a few ohms (for a speaker load of a few ohms).

~~~
Edit: a short note, resonances are often viewed on a time scale, and therefore use a linear amplitude scale, so there's only so much you can see before running out of resolution. The same applies for impedance.
 
Last edited:
^Voltage drive does not provide damping above driver resonance, because of mass (not being cancelled by spring) makes (disturbance) force to accelerate the mass so any electrical damping that would counter it is at least 90deg late and thus not effective, perpendicular force or actually extra excitement rather than damping.
Hhm, I'm of the opinion that the main reason why damping isn't effective (way) above resonance is that the back-EMF, the microphonic voltage (proportional to velocity) is falling 20dB/decade above resonance. It quickly gets so small against the static voltage (current times static VC impedance) that no effective feedback is established.

OTOH, damping via feedback is still established even when the counteracting force is injected 90 degree off. For positive feedback we'd need 120 degree or more and
and lot of back-EMF.

The distortion reduction under current drive mainly comes from removing the nonlinear and excursion-modulated static impedance, which is what I call the transfer impedance which transform voltage into current which in turn exerts the force on the cone.

Distortion in the back-EMF, while present, is only a problem when any considerable amount of feedback is at work in the first place, which means the back-EMF should be larger than the static voltage (this ratio is highest at resonance, where current and thus static voltage is small).
 
...
Many of the better softwares do a full Farina Log-Sweep where we can see the distortion in the current right away and know the problem zones.
Farina's Exponential Sine Sweep is a somewhat unprofessional attempt on the regularization of the approach originally developed for radars in 1950s. Of course, for an acoustician is it beyond regret but for a mathematician it is ... let's say underwhelming. Unfortunately, MatLab copied his way into Audio Toolbox. I modified it according to the appropriate Kernel-based System Identification, to be included into the next FSAF version, which I promised MathWorks to release by Nov 1st.

Yes, measuring current together with voltage and the acoustic output is the right way to go.
 
  • Like
Reactions: aslepekis and TNT
IME once you really exactly(!) clone the driver terminal voltage of voltage drive those difference tend to go away completely.
The cloning method is to measure the impulse response of the driver current under voltage drive and then use that IR as the EQ (FIR convolution kernel) for the current under current drive. It contains all the warts seen in the current**) under voltage drive, including the effect of internal standing waves etc.
The Purifi blog had an article awhile back (I lost the link) giving some good insights into memory effects in ferrite and magnetisable cores in general.

Retrofitting a FIR filter to a voltage amplifier would be ineffective against such pseudo-random glitches. Though, curiously, it might be feasible to develop an algorithm that anticipates when the next glitch is likely to be triggered, in the spirit of forward correction.
 
Hhm, I'm of the opinion that the main reason why damping isn't effective (way) above resonance is that the back-EMF, the microphonic voltage (proportional to velocity) is falling 20dB/decade above resonance. It quickly gets so small against the static voltage (current times static VC impedance) that no effective feedback is established.

OTOH, damping via feedback is still established even when the counteracting force is injected 90 degree off. For positive feedback we'd need 120 degree or more and
and lot of back-EMF.

The distortion reduction under current drive mainly comes from removing the nonlinear and excursion-modulated static impedance, which is what I call the transfer impedance which transform voltage into current which in turn exerts the force on the cone.

Distortion in the back-EMF, while present, is only a problem when any considerable amount of feedback is at work in the first place, which means the back-EMF should be larger than the static voltage (this ratio is highest at resonance, where current and thus static voltage is small).
Yeah the back-EMF that dampens electrically basically makes the impedance peak, reduced current with constant voltage input. Image capture is from the article, shows it nicely, while suggests some of it extends quite high.

Screenshot_2024-09-28-00-24-49-23_40deb401b9ffe8e1df2f1cc5ba480b12.jpg

You are right that the higher series impedance reduces effects on the static impedance. Impedance is literally relationship of voltage and current, and if impedance modulates it means relationship of voltage and current is being modulated.

The driver motor is a voltage source, the back-EMF, so voltage modulates due to any forces affecting the cone making it move making the back-EMF. But if same circuit has high series impedance outside the driver, the voltage modulation would not make much current in the circuit, so modulation in current is less than it could. Since any current in voice coil would affect force in the motor and appear into acoustic domain, reducing modulation in current would make less of it acoustic.
 
Last edited:
Retrofitting a FIR filter to a voltage amplifier would be ineffective against such pseudo-random glitches. Though, curiously, it might be feasible to develop an algorithm that anticipates when the next glitch is likely to be triggered, in the spirit of forward correction.
It eould be cool to have some learning algorithn do few test passes and learn a driver and then correct it 🙂 Until then we can manipulate impedance in series with a driver with passive network to get some attenuation on distortion harmonics, or manipulate electrical damping, and set the frequency response in the DSP.
 
Last edited:
  • Like
Reactions: Abhishmuk
It eould be cool to have some learning algorithn do few test passes and learn a driver and then correct it 🙂 Until then we can manipulate impedance in series with a driver with passive network to get some attenuation on distortion harmonics, or manipulate electrical damping, and set the frequency response in the DSP.
Then we would be looking for another hobby🙂
 
The Purifi blog had an article awhile back (I lost the link) giving some good insights into memory effects in ferrite and magnetisable cores in general.

Retrofitting a FIR filter to a voltage amplifier would be ineffective against such pseudo-random glitches. Though, curiously, it might be feasible to develop an algorithm that anticipates when the next glitch is likely to be triggered, in the spirit of forward correction.
That's why I wrote that the compensation only works for the linear part of the relationsship, which excludes distortion of any kind, including "spot noise" spuriae.
 
Their Convex speaker range is said to be the world's first commercially available current drive active speakers.
ADAM Audio F5 and F7 models, for which I designed the electroncis more than a decade ago, applied current drive on the tweeter and mixed impedance drive for the woofer (voltage drive around resonance and increasing impedance from there).
 
Has anyone ever measured an amplifier that works with mixed operation, i.e. with current and voltage control?
I built one, but my schematics from 5 years ago are too embarrassing to display.

Run a simulation with a couple of different speaker loads like 8 ohm and 16 ohm, and compare the output gain vs frequency for both. Try and figure out the output impedance.

Also, to convert your schematic to a 'filtered' version so bass has a lower output impedance than treble, replace the 330 ohm resistor with series RC, say 1k ohm + 22nF, and add a grounded resistor (eg. 1k) to the inverting terminal of the amplifier. Playing with the values is a bit weird, and you're quite likely to have some kind of shelving high-pass response, which then has to be EQ'd with a mirror image of that.

What you want is when switching the speaker load between 8 & 16, bass gain changes very little, but high frequencies undergo a decent shift.
 
@mikets42

I have two different tweeters that do very well at 96dB-100dB / m over their intended (and extended) passband. They are also reasonably affordable (<$100 ea)

How can we reconcile these current drive measurements to very good, albeit voltage driven, devices?

Here's a Farina log sweep equivalent to 106dB/1m, and the harmonics are low enough to wonder...

IMG_1848.png
 
Last edited:
ADAM Audio F5 and F7 models, for which I designed the electroncis more than a decade ago, applied current drive on the tweeter and mixed impedance drive for the woofer (voltage drive around resonance and increasing impedance from there).
I have (had, gave away) Adam F5. Current drive... where those fantasies are coming from, and what for? When I looked at the layout (once, I wanted to replace the chip with LM3886) I did not find anything non-standard.

Low-freq chip amp was going to protection on strong midrange passages where the woofer impedance dropped below 4Ohm. Adding a 1 Ohm resistor fixed the bug. This woofer was the first attempt of ADAM to design their own driver, and not very successful one, the sound was horrible regardless of the amplifier. It was actually the worst 5.25" driver I ever measured. The tweeter is a low-power AMT that is crossed way too low. It measures ok on sine sweep but the non-LTI residuals on music are plain disgusting. Such a small membrane can not dissipate the power and constantly heats and cools with low tau. I published the objective measurements on MathWorks community site a few years ago. Overall, IMHO, I can not distinguish between ADAM Audio marketing and a scam.

AFAIK, the current driving of tweeters, either planar, AMT or dome, does not bring any benefits. I tried about a dozen from different manufacturers (@tktran303 - including SB tweeters) - no luck.
 
Current drive... where those fantasies are coming from, and what for? When I looked at the layout (once, I wanted to replace the chip with LM3886) I did not find anything non-standard.
Obviously you're not an electronics expert and therefore missed the main give-away, the current sense resistors. The details are in the feedback network which you seem to have missed as well.
This woofer was the first attempt of ADAM to design their own driver
It wasn't the first driver. All of A series and AX series had custom designed woofers.
It was actually the worst 5.25" driver I ever measured.
Indeed this specific woofer wasn't spectacular but you know what? F5 and F7 were ADAM's entry level speaker and corners had to be cut, some of them very deliberately (you never make your entry level product as good as it gets, for obvious non-technical reasons).
low-freq chip amp was going to protection on strong midrange passages where the woofer impedance dropped below 4Ohm.
Would you have preferred the amp or woofer to break instead?
BTW, we never had any reliability problems or customer complaints with F series that were beyond what is typical (at least until I left the company in 2014. But my personal stock of F5 and F7 are still going strong). Again, this was an entry level budget offering not designed to take constant party level abuse 24/7.
The tweeter is a low-power AMT that is crossed way too low. It measures ok on sine sweep but the non-LTI residuals on music are plain disgusting. Such a small membrane can not dissipate the power and constantly heats and cools with low tau.
As mentioned, there were compromises to be made. Actually, one of the reasons current drive for the tweeter was the thermal handling and thermal protection, current drive keeps output constant even when the thing gets hot and resistance increases. And when it gets too hot and bulges so that traces start to touch each other, lowering resistance, current drive then reduces power accordingly and prevents burn-out.
AFAIK, the current driving of tweeters, either planar, AMT or dome, does not bring any benefits. I tried about a dozen from different manufacturers (@tktran303 - including SB tweeters) - no luck.
Other peoples experiences are different. So far I haven't seen any AMT that didn't profit from current drive. And most cone drivers also sound and measured better with high impedance drive above main resonance region.

Enough said from my part, please note I won't engage in any further discussion.
 
@KSTR

can you give the exact production time? Or other hint like a foto to recognize these models if sold?

Were these with eton drivers?
Eton drivers were only used in S and SX series.
There is still some info on F series on current ADAM website. https://www.adam-audio.com/en/f-series/f5/

BTW, F5 and F7 are still in the top ranks for lowest self-noise in the industry. That also was a design goal, making a very quiet speaker suitable for very low listening levels without the annoying hiss of most of the competitors at the time.
 
Obviously you're not an electronics expert and therefore missed the main give-away, the current sense resistors. The details are in the feedback network which you seem to have missed as well.
....
Other peoples experiences are different. So far I haven't seen any AMT that didn't profit from current drive. And most cone drivers also sound and measured better with high impedance drive above main resonance region.

Enough said from my part, please note I won't engage in any further discussion.
Eton drivers were only used in S and SX series.
As a scientist, I based my words on my own objective measurements, and the series of (very positive) reviews containing interviews with the ADAM's CEO https://prosound.ixbt.com/monitors/adam-monitors.shtml (i.e., his own words)

Here is X-ART on pure voltage drive, 80 dBSPL output (~10s mW), Sine sweep residuals:
0-fig6.png

and with a series 24 Ohm resistor, calibrated to the same loudness
24-fig6.png

The same long tail of harmonics reveals 1-order discontinuity which actually worsens.

Just my 2c - it would be great if people could substantiate their questionable claims with hard data (not with even more questionable claims) while refraining from making any unsubstantiated claims (however tempting it may feel in the heat of a moment) and personal attacks, in the first place.
 

Attachments

Last edited by a moderator: