Well as in all speaker tests, the driver distortion is always lower than when listening. When I bring up the level to a modest 90db spl at 3 feet/ 1 m, the distortion goes way up compared to testing very close to the cone. To hear the bass, because of our loudness curve, we have to play a lot louder than for a tone a few inches away. When you are 3 m away, to get the same spl level, a lot more power is needed and thd goes way up fast.
The practical solution is to use larger cone area drivers for bass. Keeping the cone movement to a more limited excursion.
THx-RNMarsh
PS. since spkers for a car has the driver closer to you than in a living room... like 1-3 feet max, the cone size can be reduced and still have cone movement small enough to have low distortion.
A 6.5 inch diam might be OK for near-field listening in the home.
Last edited:
If you look at IM plots and THD plots, IM distortion reduction is same as THD H3 reduction above 4kHz. IM FFT shows more spectral components, it is in fact odd harmonic reduction above 4kHz.
Electrically, everything in the current path (left side from the gyrator) is linearized. This is not true for the lumped components right from the gyrator. My guess is that for this tweeter the source of asymmetric distortion is the Kms, spider and surround stiffness. This cannot be corrected by the current drive, so it remains in the signal. Symmetrical distortion (enough above resonance) is corrected.
jneutron, do you know the relative permeability of a typical loudspeaker magnet? It's important for determining Bl(i) or reluctance force. Judging by some Le(x) measurements it's usually above 2.
After a few days I finally figured out the flux model. This allows me to model reluctance force, steel saturation, eddy currents, etc.
I have scrapped numerous schematics over more than a year and started over. If I knew of a better way, I would be doing it. If you really want to help, point out equations for Bl(v), Rms(v), Le(v) etc. These appear to be the distortions that don't fall into nonexistence as excursion decreases. All the distortions you mentioned respond to excursion which decreases with frequency, so if they were dominant then there would be no distortion above Fs.
After a few days I finally figured out the flux model. This allows me to model reluctance force, steel saturation, eddy currents, etc.
I have scrapped numerous schematics over more than a year and started over. If I knew of a better way, I would be doing it. If you really want to help, point out equations for Bl(v), Rms(v), Le(v) etc. These appear to be the distortions that don't fall into nonexistence as excursion decreases. All the distortions you mentioned respond to excursion which decreases with frequency, so if they were dominant then there would be no distortion above Fs.
One man's modest is another's being driven out of the room.Unless 90db is peaks on some of the highest creat factor music out there.
(Just a friendly chuckle about how different our listening needs are.)
What exact level you enjoy at your listening distance has nothing at all to do with what I told.-RNM
I have worked only with neo's, which are in the 1.05 range.
The path reluctance is dominated by neo reluctance. If a magnet has higher permeability, then it could be the gap dominating.
I'd have to google alinco or ferrite, never used them.
If the speaker is very non linear and you just connect the amp, the current draw will be dominated by the non linearity. If you put a thousand ohm resistor in series, the resistor will dominate, a thousand volts, one amp.
A current source by design, has an infinite resistance so the current dominates. A voltage source by definition has zero series resistance.
Jn
JN: I think a more fundamental explanation is needed for those who are not living in this stuff. It may also help us get clarity in what we are trying to accomplish with all of this driver fiddling.
Missing from the explanation I think is explaining that the driver motion is more proportional to the current than the voltage and the non-linearities from the nonuniform magnetic field affect the currents much less when the source is pure current rather than voltage.
With a voltage source driving a nonlinear load the non-lineararity manifests itself in the current waveform, it flips with a current source and the voltage waveform shows the nonlinearity. However the forces in the driver are current driven, not voltage (except electrostatic speakers. . .) so the motion should be more linear. However this does not help significant mechanical non-linearities from suspension or especially cone "breakup".
Speaker magnets are often called motors, and they behave similarly to electric motors. When you use your old hand drill, the speed stays fairly constant regardless of force (if it is a good one). This is because when the load is higher, the motor draws more current and gets more power from the input voltage. The reason this happens is that when the motor encounters no load, it produces it's own counter voltage which blocks the input voltage (commutation causes ripple, but don't let that confuse you about the basic relationship. The same applies for AC motors).
In speakers this is also true. The counter voltage (often called back-EMF or just EMF) is also not perfect. It contains extra nonlinearities that are often not present in the current-to-force function that is the primary mechanism of action. So the counter voltage adds these harmonics to the motor current, which are converted to force at the output.
With current drive, the counter voltage has no way to become a current (In an 8 ohm speaker, if the counter voltage is 1V, you will get 1V/8ohm less current. If your drive impedance is 1k, the loss in current will be much smaller, 1V/1008).
Of course now that the counter voltage is not regulating the motor, if you use current drive on your hand drill it's output force will be constant but it will go faster on light loads and slower on heavy loads. You would have difficulty controlling it.
In a speaker, the load is constant and more linear compared to a hand drill. The cone mass is the primary load above Fs and it never changes unlike the load on a hand drill. The box air load and spider are the main loads near Fs and they don't change much either unless you're driving the speaker hard. So what happens in current drive is you exchange load regulation for better motor force linearity knowing (ideally) that the regulation is less linear than the mass suspension system you are driving. Of course this is not always the case, speakers vary widely.
However the loss of regulation means the cone is not as well controlled across the frequency range. Speaker designers basically use coil resistance, inductance and eddy currents to adjust the speaker output to make it flat across the frequency range. With current drive all this tuning makes no difference, so you get very uneven frequency response, making most speakers useless without additional EQ designed for that speaker. At the same time the inductance and eddy currents would have been causing distortion, which is the compromise you have to make. Most current drive discussions are dominated by the desire for a solution to the EQ issue.
@RNMarsh: your post disappeared, however, I am posting a simulation of Zout of the amplifier that I use as a current output amplifier. Basically it is a VFB amplifier with Rf/Rg = 25k/1k. Output impedance reflects loopgain, or in other words openloop frequency response and Rf/Rg, not absolute values of Rf and Rg. Rf would be reflected if it was a CFB amp. But the output impedance reflects Rshunt as well (reference resistor)! The lower Rshunt, the lower Zout.
We can see that Zout is about 8.5kohm at 100Hz and 850ohm at 10kHz. Cut-off frequency about 1kHz.
We can see that Zout is about 8.5kohm at 100Hz and 850ohm at 10kHz. Cut-off frequency about 1kHz.
Attachments
Last edited:
This amp's purpose was to check effect of current drive to distortion. The output impedance curve is Rshunt*loopgain.
Re your question, as explained by 1audio, keantoken and as stated on my webpage
Current drive of speakers and speaker distortion
there is an option of pre-filtering the input signal by inverse characteristics to impedance plot. This can be precisely and effectively done in rePhase - that's what I experimented with 4 years ago.
I don't think it's too hard to make the impedance low at bass, maybe people just aren't sharing their experiments. The issue is that you can't make the impedance rise too fast, and this means it's a compromise. The area in between Fs and treble doesn't get as much of an effect as it would with pure current drive.
- Status
- Not open for further replies.