BIPOLAR ALTERNATIVE TO H BRIDGE - GENERATING SINE WAVEFORMS - HALF / SINGLE & FULL

[Aziz]

Hi all,

this TX circuit is awesome.

I like it. Thanks Moodz for sharing it.

[Tinkerer]

Thank you, Carl and Teleno, for the help with simulating the ground. I am afraid that your suggestions are beyond my capabilities.

In the meantime, in the simulation, I mimicked the losses of coil energy due to the ground by adding losses in the RX. This causes the slope of the TX current to increase.
The losses in the ground also reduce the peak TX current. Changing the height of the coil above the ground increases or decreases the losses.

With a traditional PI, at each cycle the coil is totally discharged and fully recharged.
With the Square wave current TX, the losses in the targets and the ground, accumulate.

The only solution is the replace the losses at each cycle.

Here is the effect of the TX slope shown on the RX. (Real circuit) The red trace is a target.
Sampling this RX signal, AC coupling to the integrator works, but reduces the amount of gain possible in the preamp. It also complicates the Ground Balance a lot.

[Teleno]

Not sure if LT-Spice supports cores.

Here are some LTSpice examples. They look intimidating.

https://ltwiki.org/index.php?title=The_Chan_model

https://electronics.stackexchange.com/questions/243927/transformer-with-hysteresis-in-ltspice

Models that should work on all SPICE based programs.

http://www.intusoft.com/articles/satcore.pdf

[Carl-NC]

Does anybody have a good suggestion how to simulate the ground?

The best way I can think of is to add a core material to the TX coil. Easy-Spice lets you define a saturable core for inductors and transformers that I assume produces a B-H curve. That's the key to simulating anything in the ferrous quadrant as the B-H curve gap determines the loss angle. You can't do it with just an RL model. Not sure if LT-Spice supports cores.

Another possibility is to use an analog behavioral model. However, it still needs a way to sense the TX magnetic field and produce a ground magnetic response, so it would need to look something like this:


f(x) can be a LaPlace transform or a simple analog behavioral block, I think both are supported by LT-Spice. This model can also support iron targets and even non-ferrous targets.

[Tinkerer]

This is the full circuit.
In this simulation I do not use an induction balanced RX. Just a separate concentric RX coil of about half the diameter of the TX coil.
Moodz_bipolar_Square_v_TINKERER.rar In the simulation we see the slope of the TX square wave current (green) and one target with a TC of 50us, (red) and a target with a TC of 100us (blue). We see that the target do not fully decay during the 100us cycle time, therefore the remaining eddy currents subtract from the amplitude of the eddy currents of the next cycle.
To get more amplitude we should extend the cycle to 2 target TC's.
It is also obvious that a compensation circuit is needed to reduce the slope of the TX current.

I am still working on a simulation of the ground.

KingJL, thank you for the feedback.

[KingJL]

So what happened here?
We see a distinctive slope in the TX that is not there on the last picture. In the last picture the RX was not coupled (K1 L1 L2 0.3). Now that the TX and RX coils are coupled, the slope in the TX appears. This shows us that the energy absorbed by the RX coil and the damping of the RX coil cause the slope in the TX coil current.
As the energy is lost during the Flyback, the coil current increases to return to the constant current level...

Tinkerer, Not sure as I cannot see the complete circuit, but I suspect it at least partially due to your TX/RX relationship and coupling? If you are trying to simulate a concentric RX coupling, you seem to be missing the bucking coil in the TX circuit which would oppose the primary TX/RX coil coupling. I have found that trying to simulate a concentric coil and target arrangement in LtSpice is extremely difficult as it tends to generate an error about an impossible transformer configuration.

[Tinkerer]

Targets
L2 is the RX coil is its simplest form, concentric about half the diameter of the TX coil
R3 is the damping of the RX coil
R12 limits the current to the input of the preamp
D1,D2, protect the input of the preamp

L3,L4,L5,L6,L7, represent targets with TC's from Ius to 100us.
A nickel, (US$ 0.05) is a common target with a TC of 10us
A very small gold nugget could have a TC of 1us
The (transformer) coupling of the TX coil L1 with the targets is exaggerated to make it easier to see.
The picture shows the current wave form of L1 in red
The eddy currents in the targets L3,L4 in green and blue

So what happened here?
We see a distinctive slope in the TX that is not there on the last picture. In the last picture the RX was not coupled (K1 L1 L2 0.3). Now that the TX and RX coils are coupled, the slope in the TX appears. This shows us that the energy absorbed by the RX coil and the damping of the RX coil cause the slope in the TX coil current.
As the energy is lost during the Flyback, the coil current increases to return to the constant current level.

Next I will try to show the problem with the ground. Does anybody have a good suggestion how to simulate the ground? I have never done it yet, but I will try to find a solution unless somebody is willing to help?

[Tinkerer]

Tinkerer, The boards are offered for free... just need to take care of shipping cost. For most shipping destinations outside the US, it may be much cheaper to order directly from JLCPCB (5 boards for $2 - $3.33 plus their shipping). US shipping internationally is expensive. I can get something from Hong Kong and have it shipped for as little as $3.93... but to send the same thing to Hong Kong it costs 10 times that. Must be missing something!!

If you order from JLCPCB, I recommend to use the gerbers in post #16 of this thread.

Thank you King, this is most generous of you. Yes, here also it is much cheaper to order directly from China. I will use your Gerbers.

[KingJL]

As I am looking closer at the Moodz circuit, I find it very interesting and suspect it might be even better than my own bipolar square current design. I would like to buy your boards. Can I pay with Paypal?

Tinkerer, The boards are offered for free... just need to take care of shipping cost. For most shipping destinations outside the US, it may be much cheaper to order directly from JLCPCB (5 boards for $2 - $3.33 plus their shipping). US shipping internationally is expensive. I can get something from Hong Kong and have it shipped for as little as $3.93... but to send the same thing to Hong Kong it costs 10 times that. Must be missing something!!

If you order from JLCPCB, I recommend to use the gerbers in post #16 of this thread.

[Tinkerer]

If anyone is interested, I have 2 extra bare PCB's of the original Moodz alt2h-bridge that I have been using for about 4 months. This version does not have the half-sign capacitor or damping network on board... have to add externally for experimenting with half-sign.


After the PCB is populated:


Ref post #11 of this thread for schematic.

As I am looking closer at the Moodz circuit, I find it very interesting and suspect it might be even better than my own bipolar square current design. I would like to buy your boards. Can I pay with Paypal?

[Dean Sarelius]

A separate RX coil is almost always better. Even if it's not induction balanced you can better optimize the number of turns instead of being stuck with whatever the TX coil has. And if it's induction balanced you can usually sample earlier.

Thanks Carl - I was hoping to keep it simple but now I fear will have to tread down the rabit hole of dual coil and phase analysis...cheers

[Carl-NC]

A separate RX coil is almost always better. Even if it's not induction balanced you can better optimize the number of turns instead of being stuck with whatever the TX coil has. And if it's induction balanced you can usually sample earlier.

[Dean Sarelius]

Ummm, no. A linear ramp in the TX current causes a constant induced EMF in the target, which creates an exponentially rising eddy current. The signal at the RX coil is a derivative of the rising exponential, which is a spike followed by a falling exponential. In a standard PI the TX turn-on creates a negative RX exponential which, if it has not decayed to zero, will subtract from the flyback's positive RX exponential and reduce sensitivity.

Dean, your question specifically refers to the TX pulse width of a monopolar PI in a thread that's all about bipolar PI, mainly constant current & half-sine. Mixing the different technologies can be confusing as the results can often look radically different. Here is a thread that discusses the target responses to a whole bunch of TX waveforms:

Optimizing Target Responses

I think post #12 might answer your question. Feel free to post further questions about monopolar pulsing to that thread to minimize confusion here.

Thanks Carl - I will check the link you provided. Out of curiosity with the dual coil arrangement does anyone feel that the sensitivity is reduced in the RX coil compared to sampling a mono...? Although I am yet to experiment in a dual coil design I was assuming there will be a lack of target sensitivity by implimenting a seperate RX coil and the phase data availabe from the RX would only be available from targets within relatively close range to the RX coil..?

[Carl-NC]

That makes sense, I wrongly assumed that a current was induced when in fact it's the EMF that creates the current.

Yes, correct. It's a common mistake because we often read that a changing magnetic field induces a current, but it's really an EMF.

Then we have a problem because stabilizing the ramp right at the beginning is impossible, but at least the transient is orders of magnitude smaller than in a classic configuration.

Yes, you rely on the really fast di/dt to give you the results you want and hope any residual slow di/dt is slow enough not to mess things up.

[Teleno]

Ummm, no. A linear ramp in the TX current causes a constant induced EMF in the target, which creates an exponentially rising eddy current. The signal at the RX coil is a derivative of the rising exponential, which is a spike followed by a falling exponential. In a standard PI the TX turn-on creates a negative RX exponential which, if it has not decayed to zero, will subtract from the flyback's positive RX exponential and reduce sensitivity.

That makes sense, I wrongly assumed that a current was induced when in fact it's the EMF that creates the current.

Then we have a problem because stabilizing the ramp right at the beginning is impossible, but at least the transient is orders of magnitude smaller than in a classic configuration.