Balanced VLF Tx

[Qiaozhi]

How about cancelling my warnings #1 and #2 Mr. Q?
Warning #3 would very likely ban me and the people here will perhaps miss my freaky & mad contribution. Even PJ and his brothers would get sad.
Aziz

I'll take that as an apology then.
Freaky and mad contributions are acceptable, but personal abuse (and the disguising of profanity using special characters) are not.
Try to stick with the technical (freaky?) stuff and keep the ranting monologues to yourself (patent trolls included).
If you have a gripe about something, then make your point and move on. After that it just becomes boring and repetitive.

[ODM]

True, this should be a technical forum anyway

But it's good to be careful with unidentified magnetic cores. Ferrite cores are saturated easily by a relatively small H, and some nickel-iron alloys with even less.

It's not likely to happen that a power supply series choke is easily saturated by a few mA, but if it's some other choke for high RF or only AC it's better to be safe than sorry. Otherwise all the fine simulations will go to naught if we first spend days with LTSpice and then just sit on the bench with two soldering irons in hands!

[Davor]

Aziz, no current generator in your schema. It is necessary for the parallel tank circuit

I'm afraid not.
The most common solution for a complementary (H-bridge) cross-coupled oscillator as seen in literature is using a single current source, either in high or low side. A single current source introduces 2nd harmonic on each side of a coil that is not seen in the coil current because these cancel each other, but if you have some plans with this voltage, say extracting phases for the Rx, then you are missing the point.
Two current sources would be a disaster since they'd have to source/sink exactly the same current, which is impractical at best. Point is that these BJTs while in linear regime are in fact current sources, and only when they are pushed to saturation the coil is exposed to whatever lies beyond the emitters. So in full honesty you may say that a coil is in fact sourced with current source most of the time, and the resulting average impedance is well over the Q limiting one (in this case ~500ohm). So my vote is for resistors or chokes.

In fact you are right, pressing it too hard would reduce linear regime periods, the resulting average impedance and hence the tank Q. It would increase harmonics, but not the even ones, and the circuit power consumption. But you can design it to behave nicely.

Most interesting is the voltage limiting action of BJTs going into saturation - they take care voltage on each coil side overshoots the rails only by a little, and only in case it is pushed too hard. I've seen dozens of incredibly complex oscillators that employ various stabilisation schemes, and this one works by itself

[WM6]

It's not likely to happen that a power supply series choke is easily saturated by a few mA, but if it's some other choke for high RF or only AC it's better to be safe than sorry.

Hi ODM,

can you explain this a little more? Do you suggest to use high RF ferrite core?

..... pressing it too hard ...

Davor, what mean "pressing it too hard" in this case?

[Sergey_P]

the tank Q

mathematical dependence tank Q => http://www.md4u.ru/viewtopic.php?f=77&t=7299&start=125 (post: Пт: 14 сен 2012 19:03)

Maximum current in the tank circuit, Im ~ U/wL

[WM6]

mathematical dependence tank Q => http://www.md4u.ru/viewtopic.php?f=77&t=7299&start=125 (post: Пт: 14 сен 2012 19:03)

Maximum current in the tank circuit, Im ~ U/wL

Maybe this google-trans will help:

The thing is that even if you and I will draw these calculations, there is no good of it will not. This puzzles need not formulas, and an understanding of processes and their relationship ...
I'll try to draw on my fingers ...
There is the most complex scheme requires an analysis of its work for the influence of source resistance on quality of the oscillatory system.
The resistance of the circuit elements in the points ab is:
Z = (R1 + iwL) (R2 +1 / iwC) / (R1 + R2 + i (wL-1/wC))
If the reactance is much more active, we can write:
Z = (L / C) / (R + i (wL-1/wC)), R = R1 + R2 - resistance of the losses in the circuit
L / C = r ^ 2 (the square of the characteristic impedance of the circuit)
Z = p ^ 2 / (R + i (wL-1/wC))
It is better for the parallel circuit its composition expressed in terms of conductivity, there is clearly visible three components:
1 / Z = 1 / (p ^ 2 / R) + iwC-i/wL
Not hard to see that series resistances R1 + R2 = R are equivalent bypass circuit resistance Re = p ^ 2 / R
Similar influence on the circuit resists the excitation source, which results in total loss of the equivalent resistance in the circuit to the value Rekv = R1 + R2 + p ^ 2/Ri
Q of the circuit is:
Q = p / Rekv = p / (R1 + R2 + p ^ 2/Ri)
This relation shows that the Q of the circuit in the TX depends not only on its own merit naked loop, but also on the source resistance and the resistance is higher, the closer the Q of the circuit in the generator circuit to his own merit. For an ideal current source Ri = endless , but in our case it is not so, because we have only to all current generator operating in a limited range of voltages, which leads to limited values Ri. For small values of Q, is only a little current from the current shift key the pump for maximum undistorted signal is generated, as immediately change Ri, which will lead to change in the Q and all your calculations ... => ... The same thing will happen to the voltage source, the wire diameter, the resulting inductance, capacity ...
But that's not all, because except Q, there is another option for overriding the MD - the value of the generated field (A + vit)
Easier and more reliable: wind, measure, correct ... watching Q and A * turns ... but not from the light, and purposefully

In quasars - serial loop. There are no restrictions on the voltage on the coil resistance of the source is determined ... so contour elements can be considered with a certain error, selecting current and Q, choose a wire, etc.. Examples of calculation given above.
In a parallel circuit are constrained by the need to implement a harmonized regime for current, voltage, resistance, inductance ... Change one parameter in the process of manufacturing or service, invariably leads to changes in the others and the nature of these changes (sensitivity) is much higher than the degree of their influence, rather than in a sequential circuit .. where

[Aziz]

Hi guys,

I want to show you some interesting voltages and currents.
1. Half sine wave supply voltages V+ and V- and the differential voltage (V+) - (V-) of the oscillator.


Note: Supply voltage is going almost to zero! Look at the differential voltage! Nice. Nice oscillator having zero voltage supply phase. And having a maximum voltage supply phase. *LOL*
Here is the magic buried.

2. Inductor current and the ripple feedback voltage from the oscillator to Vcc and Vee (note, Vee is inverted)


The ripple voltage is approx. 0.5 mVpp. The inductor "injection'" current: Average DC current = 22 mA and oscillating AC current = 24 mApp.

Comeon, let's do more magic things to the circuit.
Aziz

[Aziz]

3. And here is the oscillator coil current and voltages:

Not bad.

Aziz

[Aziz]

One of the very important fact is, that V+ and V- should be half sine wave.
Don't smooth V+ and V- with capacitors. They have to be half sine waves and not DC voltages.

If you put some smoothing elco's to V+ to V- and V+ to GND and V- to GND look at the power consumption rise and at the transistors temperature or power consumption (they could even smoke).

Again:
Don't smooth V+ and V- with capacitors. They have to be half sine waves and not DC voltages.

(That would be a magic killer.)

Cheers,
Aziz

[WM6]

Great work Aziz, thanks.

Frequency here is about 80kHz. Why so high? Wouldn't be better (regarding depth) at say 3kHz?

[Davor]

Aziz, I'm about to do some magic to your Rx circuit to gain you somewhere about 7 times less noise voltage for almost free

Davor, what mean "pressing it too hard" in this case?

In my original circuit you can give emitter resistors more or less resistance, as long as they are equal. If they are lower, the saturation stage becomes more prominent, and the circuit draws more current. As a consequence, in a saturation the tank sees lower parallel impedance which lowers its Q.

A simplified example... If a coil has inductance, say 300uH, and resistance 0.5ohm, there exists a frequency for which this coil in a resonant tank has a Q=30 (hence omega*L=15). The same Q would be if I have an ideal coil in a tank (Rcoil=0), but a resistance in parallel of a tank is 15*30=450ohm. So, when equivalent parallel impedance, which is a complex and nonlinear function of those previously mentioned emitter resistors, drops in it's average value to 450ohm, my tank will have 1/2 of it's maximum Q, because both series resistance of a coil and parallel impedance of the oscillator are loading my tank.
With emitter resistors too high, peak to peak voltage of the oscillation is lower than the rails, but when they are reduced oscillation sticks to the rails. Lowering further would be pressing it hard, and it would yield only more current consumption and harmonics. So there is an optimum choice for emitter resistors.

I did not analyse Aziz' circuit yet.

[Aziz]

Forgot the collector currents. Sorry, but I have lost the previous time base. Anyway.

4. Collector currents of the transistors (base/emitter current isn't important at the moment)


Aziz

[Aziz]

Great work Aziz, thanks.

Frequency here is about 80kHz. Why so high? Wouldn't be better (regarding depth) at say 3kHz?

No, I have a resonance frequency of approx. 12 kHz. LTX, parallel coil capacitor, the left and right capacitor in the oscillator are defining the resonance frequency.

Aziz

[WM6]

No, I have a resonance frequency of approx. 12 kHz.

O yes, thanks. I take by mistake your sim. Vtp2 period as approx. 10us instead 100us (more precisely 85us).

[Aziz]

Let's calculate the resonance frequency:
(You know, I love math!)

Thompson resonance frequency formula:
f0 = 1/(2*pi*sqrt(L*C))

L = 300µH (Transmit TX coils inductance)
C = Total capacitance, well, let's collect all the capacitors now.

Cp = 470nF (my reference value)

Cp: parallel capacitance over the TX coil.
Cv: series capacitance of voltage divider:
1/Cv = 1/220pF + 1/1000pF => Cv = 180 pF

right part: Cp/2 + Cv = 470nF/2 + 180 pF = 235.18 nF
The voltage divider contribution can be negligible as you can see it.

left part: Cp/2 = 235 nF

Series left and right part:
1/Cs = 1/235nF + 1/235.18nF => Cs = 117.545 nF
Cs is added to the parallel capacitance Cp
C = Cp + Cs = 470nF + 117.545 nF = 587.545nF
Ok, C is the total resonant tank capacitance.


f0 = 1/(2*3.1415*sqrt(300µH*587.545nF)) = 11,987.8 Hz
Yep, that is the correct value.

Aziz