开云体育

Vlad(imir) is a Slav name. Most, but by no means all, Romanian names are of Latin origin. So, not only Russian.

Best wishes
John Woodgate OOO-Own Opinions Only
J M Woodgate and Associates 
Rayleigh, Essex UK

On 2018-11-28 08:19, Vlad wrote:

"I see you are using an ideal approach, so you could use a different
topology that is simpler, with the same effect."

Where you end up often depends on where you start. You are right, of course, but in my defence, I started from a physically realisable circuit and optimised it as a simulation tool. It also has the advantage that it should work with any SPICE.

Feel free to offer a more efficient alternative.

Regards,

Tony

In the pretend world of simulation, the same requirements apply
regarding noise and distortion measurements. I have uploaded a model of
an AES17-20k filter that meets all the requirements of the 20kHz brick
wall filter:

Insertion loss at 20kHz: < 0.1dB
Rejection at 22kHz: > 20dB
Rejection from 25kHz to 10MHz: >65dB

Earlier today I received a message from Mike. There is a now a new update
of LTspice available in which this minor bug has been squashed.

I am running LTspice version XVII. This is in the context of LTspice simulating the Servo Sound SL15, SL20, Korn & Macway KM30, KM40, KM50, and 3A Andante speakers.?

They are active speakers dating back from 1969 (Servo Sound SL15) to say 1980 (3A Andante).

They embed a 0.33 ohm shunt resistor enabling their built-in power amplifier to benefit from a supplementary feedback channel, taking into account the instantaneous current that's flowing into the speaker voice coil. Kind of refined negative output impedance drive.

Unfortunately, the following LTspice set of directives must be issued on the main schematic :

.param R1e 3.5? ? ? ? ?; ohm (coil copper resistance)

.param L1e 390?? ? ? ;? henry (main coil)

.param K1e 0.81? ? ? ?; Le1 Le2 coupling?

.param L2e 3.7n? ? ? ?; henry (eddy current coil)

.param R2e 68?? ? ? ?; ohm (eddy current resist.)

.param Bl 4.2? ? ? ? ? ? ; A*m or N/A (magnetic force)

.param Mms 5.5m? ? ;? kg (moving mass)

.param Rms 0.87? ? ?; ohm (resistive loss)

.param Cms 1.36m? ?; m/N (suspension compliance)

.param Sd 52 * 1e-4 ; m^2 (piston area)

.meas Fs avg 1/(2*pi*(sqrt(Cms*Mms))) ; Hz

.meas Vas avg rho*c*c*Sd*Sd*Cms*1e3 ; litres

.meas Qes avg (2*pi*Fs*Mms*R1e)/(Bl*Bl)

.meas Qms avg (1/Rms)*sqrt(Mms/Cms)

.meas Qts avg (Qms*Qes)/(Qms+Qes)

.meas Rend_pc avg 100*(4*pi*pi*Fs*Fs*Fs*Vas*1e-3)/(c*c*c*Qes)

.meas SPL_2.83v avg 112.1+10*log10(Rend_pc/100)+10*log10(8/R1e)

.param Po 101k ; Pa

.param rho 1.184 ; air density kg/m3

.param c 346 ; speed of sound in air m/s

This is eating a lot of real estate on the schematic.

Here are a few questions :

1 - The "Text Edit on the Schematic" tool allows to opt for the "(not visible)" modality through the "Justification" drop-down list. I tried this. There is a problem. Once you opt for the?"(not visible)" modality, the whole set of directives vanishes from the schematic. And there is no way to retrieve the set. Do I miss something ?

2 - How how to encapsulate the above directives into a file ??

3 - How to invoke the directives file on the main schematic ?

4 - How to ensure that the directives get read by the schematic, and by the sub-circuits ?

5 - How to automatically open and view the "SPICE Error Log" after each simulation run ?

6 -? How to define a piecewise linear (PWL) element, in the Frequency-Impedance plane :?

20 Hz, 3.5 ohm magnitude, 0.0 degree phase

50 Hz, 20.0 ohm magn, 0.0 degree phase

100 Hz, 9.0 ohm?magn, 0.0 degree phase

200 Hz, 5.0 ohm magn, 0.0 degree phase

500 Hz, 5.1 ohm magn, 0.0 degree phase

1 kHz, 5.2 ohm?magn, 0.0 degree phase

2 kHz, 6.0 ohm?magn, 0.0 degree phase

5 kHz, 9.0 ohm?magn, 0.0 degree phase

10 kHz, 13.0 ohm magn, 0.0 degree phase

20 kHz, 18.0 ohm?magn, 0.0 degree phase

This is for enabling LTspice to plot such real-world impedance in function of the frequency, as reference, for easing comparisons.?

Many thanks in advance

**

For your entertainment, I am adding a documentary section that you better read using the "read aloud" text-to-speech Chrome extension.

Why-oh-why fiddling in 2018, with active speakers dating back from the years 1970 and 1980??

Say there is a power amplifier, imposing a voltage on a electro-dynamic speaker.

How to analyze it ?

?

1 - There is a voltage that's generated by the current that's flowing into the voice coil copper resistance (Ohm's Law). It is easy to compute, and equals the instantaneous current times the copper resistance.

2 - There is a voltage generated by the current that's flowing into the voice coil inductance (Lenz's Law).?It is easy to compute, and equals the instantaneous current first derivative, times the coil inductance.

3- There is a motional voltage generated by the voice coil (Faraday's Law).

The motional voltage is the voltage that's generated by the speaker voice coil operating as a giant dynamic microphone.

The?motional voltage is proportional to the constant magnetic induction that's provided by the permanent magnet. We should say "magnetic flux". Indeed, the narrower the circular air gap is, the bigger the flux is. Unfortunately, the circular air gap must be quite large, for the voice coil to be able to slide back and forth inside it, without touching the walls.

Low power speakers of say 10 watts made in Western Europe, dating back from 40 years ago, featured a voice coil made of a small diameter copper wire, wound on a paper coil former. Consequently, the air gap could be very small. Consequently, the magnetic flux was suite large, despite using a relatively light permanent magnet.

The first China-made high power speakers of say 100 watts, featured heavy voice coil made of a big diameter copper wire, wound on a aluminium coil former. Consequently, the air gap had to be very large. Consequently, the magnetic flux was low, unless using a very heavy permanent magnet.

The motional voltage is also proportional to the voice coil copper wire length exposed to the magnetic flux.

The motional voltage is proportional to the voice coil instantaneous speed. This is most important, as the voice coil movement, hence speaker cone movement, is what's actually at the origin of the sound.

Knowing this, it suffices to deduct from the total speaker voltage, the above mentioned voice coil copper voltage (Ohm's Law), and the above mentioned voice coil inductance voltage (Lenz's law), for knowing the Faraday voltage (motional voltage), hence actual voice coil speed, hence actual speaker cone movement, hence the actual sound being delivered.

Starting from this, comes the supplementary feedback scheme idea, popularized by Servo Sound as soon as 1969.?

The electric signal that's physically representing the instantaneous voice coil speed, gets presented back to the power amplifier, on a auxiliary input, for allowing the power amplifier to compare it against the desired voice coil speed that's present on its main input.

The power amplifier is now able to correct any deviation. Such correction takes a couple of microseconds, which is less than what it takes for the sound wave to actually get airborne. The voice coil speed gets massaged by multiple corrections and re-corrections before the sound wave it emits, truly gets airborne.

One wonders why starting all audio equipment manufacturers have not jumped on such bandwagon.

There may be six reasons.

1. Marketing.
You can realize more sales, and better motivate your customers to renew their equipment, when marketing a power amplifier that's completely separate from the loudspeaker. Say a DC-coupled power amplifier. Or say a MOSFET power amplifier.

You can realize more sales, and better motivate your customers to renew their equipment, when marketing a speaker enclosure that's completely separate from the power amplifier. Say a two-way speaker. Say a acoustic suspension speaker. Say a 3-way bass-reflex speaker.??

And you can sell a range of speaker cables.

2. Credibility.

In the seventies and eighties, you could not appear as a creditable audio gear manufacturer in case you put a power amplifier inside a speaker cabinet. You make it difficult for audio techs to measure the distortion, slew rate and signal/noise ratio. You make it difficult for audio freaks to tell if your power amplifier allows to perceive the pianist's sockets color.

You look suspicious, like you need to hide something.

3. Electric safety and-or the WAF.

Active speakers require a AC power cable, plus the audio signal cable. In the early seventies, Servo Sound and Korn & Macway relied on multi-wire cabling, connectors and plugs (Painton/Jones, and Bulgin after a while), conveying the AC power and the audio signal. Unfortunately such arrangement failed some electric safety tests. The latest models from Korn & Macway had their AC cable separated from the audio cable. Comes another issue then. Imagine your wife discovering that you just installed speakers in the living room, requiring long AC cables, squatting two AC power outlets. Such is the WAF (Wife Acceptance Factor) issue.

4. Voice coil copper resistance, which is temperature-dependent.

For the voice coil speed feedback scheme to be so effective, so strong, that it wins over your finger trying to immobilizing it, the electronics circuitry that's elaborating the voice coil speed feedback needs to track the real-world voice coil copper DC resistance with great precision. Another way to see this is to consider how the voice coil speed feedback operates at the very low frequencies, where one can neglect the Lenz's Law voltage that's developing on the voice coil inductance. The aim of the game here, is to drive the speaker coil using a *negative* output resistance of say 3.0 ohm, in the context of a speaker featuring a physical voice coil resistance of 3.5 ohm. This way, the system behaves like the coil copper resistance is only 3.50 ohm minus 3.00 ohm = 0.50 ohm. Anybody can implement a power amplifier featuring a negative output resistance. It suffices to add a positive feedback, basing on the current that's flowing into the speaker. There you see the importance of the 0.33 ohm resistor used as shunt. Say you achieve a negative output impedance of 3.0 ohm. The virtual 0.50 ohm copper resistance that you achieve, is 7 times less than the physical copper resistance. Consequently, your finger will face a 7 times more counter-acting force, when it tries to immobilize the speaker voice coil. This looks excellent. The problem here, is that in case the circuitry makes a 5% error in assuming the voice copper DC resistance (temperature dependent), it will implement a 3.50 ohm minus 3.15 ohm = 0.35 ohm virtual copper resistance drive, instead of 0.50 ohm. Which is 30% off, causing almost a 3 dB error on the sound amplitude.

? ? ??

5. Voice coil inductance, exhibiting high iron core losses.

This is caused by eddy currents circulating inside the soft-iron coil core, explaining why the "catalog"? 230 ?H coil is is only 18 ohm minus 4 ohm = 14 ohm at 20 kHz, while it should be close to 29 ohm. You voice coil model coil requires a 3 dB per octave slope filter. Kind of "pink noise" filter, easy to implement using a few resistors and capacitors. Most people regard this as a obscure fix for some obscure issue, while in reality, this is is a fundamental asset for precisely modelling the non-standard voice coil impedance rise in function of the frequency.

6. The stability issue when trying to implement a "true and trivial" voice coil speed feedback.

Such feedback scheme is indeed relying on the audio current that's ingested by the speaker coil, sensed by the 0.33 ohm shunt. Unfortunately, past 5 kHz, the speaker coil becomes mostly inductive, as its impedance module is now 9.0 ohm instead of 3.5 ohm. This means that the ingested current, serving as feedback signal, is mostly lowpass. Fundamentally, this could be a show stopping argument. You are running for trouble, in case you design a closed-loop feedback system that's basing on a feedback signal that's a lowpass. The higher the frequency is, the less feedback signal you get, forcing the closed-loop system to deliver a high gain at a high frequency. The most recommended approach consists in analyzing the feedback signal in gain and phase, for tailoring the power amplifier gain and phase according to some "optimum control" theory enabling to optimally balance the speed and the precision, following your requirement. Such approach consists in tweaking the power amplifier open loop gain and phase, and tweaking the feedback signal filtering, for ending up with a compact circuitry. There should be no "pre-equalizer". All the equalizing should be implemented inside the closed-loop system. Unfortunately, the Servo-Sound and Korn & Macway circuitry feature a "pre-equalizer" that's boosting the 20 Hz frequency by 20 dB. Contrary to this, the 3A Andante circuitry is not relying on a "pre-equalizer".

7. No service manual, hence no official method for adjusting the "Motional Feedback balance".

The?Servo-Sound, Korn & Macway, and 3A Andante circuitry all feature a trim resistor, aiming at subtracting a certain percentage of the speaker voltage, from the voltage that's developing on the 0.33 ohm shunt resistor. There is no official information, saying how to adjust such trim resistor. Long time ago, I vaguely assumed that such trim resistor must be adjusted for the feedback voltage to become zero when applying some audio frequency at the system input, while at the same time, forcing the speaker cone to remain immobile. Surely this is not the proper way. There must be a better, more practical way.

8. The interdiction to elaborate a feedback signal, that's the motional voltage.

There is a 10 dB resonant peak 400 Hz in the sound pressure level, in case you adjust the trim resistor (see above) for the supplementary feedback signal to equate the motional voltage. This is the?most disturbing, and most disappointing feature. Remarkably, the 400 Hz frequency doesn't correlate with the main speaker box resonance that occurring at say 80 Hz. Something else is coming into play. This may relate to the issue described above, consisting in applying a lowpass signal as feedback.

9. The obligation to elaborate a feedback signal that's departing from the motional voltage. The feedback signal that's required for getting rid of the 10 dB resonant peak at 400 Hz, appears to contain a excessive amount of speaker voltage, compared to the amount of speaker current.? ?

10. A good way to assess the true motional feedback effect.

In case the bass frequencies are governed by a motional feedback, whatever the static pressure that you finger is imposing on the speaker cone, the system will counteract your finger, for trying to move the speaker cone, as if your finger was not there.

In case the bass frequencies are governed by a bass equalizer, the system will counteract your finger in a weak way, and as consequence, it will be easy for your finger, to reduce the cone movement.

Such behavior can be tested using LTspice. It suffices to put the AC voltage source, not at the system input, but in series with the force/speed section of the gyrator. Such AC voltage source implements your finger, and now you can impose a mechanical force of say 1 Newton, that's not only static, but also at any frequency you want.

Open the LTspice schematic that's representing the Servo Sound SL15 speaker.

Cut the "motional feedback" track entering the power amplifier. Run a?AC simulation, and you will see the sound pressure level that the 1 Newton is causing, from 20 Hz to 20 kHz, when the speaker cone is held in place by the power amplifier, just like most power amplifiers are, no special feedback, only relying on the conventional voltage feedback.

Restore the "motional feedback" track entering the power amplifier.?Run a?AC simulation, and you will see the sound pressure level that the 1 Newton is causing, from 20 Hz to 20 kHz, when the speaker cone is held in place by the power amplifier, assisted by the "motional feedback".

Say you see a difference in the lowest frequencies, between 50 Hz and 100 Hz.

There may be a sound level pressure decrease, by say 4 dB or so between 50 Hz and 100 Hz, compared to the "no motional feedback" plot. This is the true motional feedback. This means that between 50 Hz and 100 Hz, when enabling the "motional feedback", the speaker cone gets 1.6 times more controlled. The speaker movement gets 1.6 times less impacted by a suspension non-linearity, by a box size mismatch, or by a suspension compliance mismatch.

There may be a sound level pressure decrease, by say 12 dB or so between 50 Hz and 100 Hz, compared to the "no motional feedback" plot. This is the true motional feedback. This means that between 50 Hz and 100 Hz, when enabling the "motional feedback", the speaker cone gets 4.0 times more controlled. The speaker movement gets 4.0 times less impacted by a suspension non-linearity, by a box size mismatch, or by a suspension compliance mismatch.

When tuning a "motional feedback" system, one must pay attention to the true motional feedback, and consider as non-functional, or badly configured, a "motional feedback" system that's exhibiting a less than 10 dB true motional feedback when enabled (equivalent to the speaker cone getting a 3.2 times more controlled).? Unfortunately, most of the time, when you manage to tune a motional feedback system for reaching a true motional feedback of say 10 dB, after you revert to a conventional AC analysis with the AC source connected at the system input, you discover that the frequency response has worsened, exhibiting valleys and mountains. This is where the sport begins !?

11. A good way to assess the dumb equalizer effect.

Open the LTspice schematic that's representing the Servo Sound SL15 speaker.

Put the AC signal source as usual, at the system input. Make sure all other AC signal sources produce no signals.

Cut the "motional feedback" track entering the power amplifier. Run a conventional?AC simulation, and observe the frequency response from 20 Hz to 20 kHz.

Restore the "motional feedback" track entering the power amplifier.?Run a conventional?AC simulation, and observe the frequency response from 20 Hz to 20 kHz.?

You will see a huge difference in the lowest frequencies, say between 50 Hz and 100 Hz, up to 12 dB or so, ironing out the the deep bass frequency response, quite a impressive result.

Compare the 12 dB improvement seen here, against the true "motional feedback" contribution in dB. Say it was 8 dB.

Thus, from the 12 dB improvement seen in the deep bass frequency response, one could say that 4 dB may originate from a conventional albeit cryptic, dumb equalizer.

What I want to make clear here, as final conclusion, is that when one is cutting the "motional feedback" wire, and facing a 12 dB worsening of the deep bass frequency response as result, it is a false assumption to say that the "motional feedback" was 12 dB.

The only valid test for assessing the true motional feedback, consists in using the finger method, or possibly, the simulation method consisting in placing the audio source (as mechanical force) in series with the gyrator output, so one can measure how much the feedback system succeeds in overcoming such perturbation, compared to the same schematic, having its feedback wire cut.? ??

? ?

That's all for today.?

Have a nice day? ? ??

Regards

Steph