BIPOLAR ALTERNATIVE TO H BRIDGE

[green]

Back to playing with the unipolar circuit. Added a 7V Tx supply so Tx signal would stay the same as battery discharged. Interested in what others see when using ferrite as the target with their detectors. I have been using a toroid core about 25mm diameter. I have tried other cores similar size and bags(around 10 cores)with smaller cores with similar results. Thinking ferrite target should be X signal with very little if any R signal.

Tx_328uH 4.2 ohms, Rx_133mm fig8

Haven't given up on bipolar yet but so far unipolar is winning for me.

[lucifer]

Tried some MOSFETS with lower Coss. Was thinking it might help. Not. Circuit works but can't get Tx constant current(A/sec)after reversal as low as I would like. Still not sure how low it should be. Maybe my coil is causing some of the problem. Back to playing with unipolar and standard bipolar.

The key in those circuits is the supply. It has to be a constant current source. Then both the standard H-bridge and moodz circuits work. I've been playing for a while with a standard H-bridge powered by a constant current constant voltage module. Power consumption was OK. However the current pulses were not completely flat-topped but I was using a high inductance coil and didn't measure the A/sec rate. Seems that a lower inductance coil is better. Still have the circuit laying somewhere round here.

[Riss]

So they don't gnaw me dark thoughts anymore (that there is no way to get it clean DC ) , I took 6 pieces IRF630 and the soldering iron to see how the circuit works on the oscilloscope ? The key phrase in this thread is What's missing? 06-03-2021, 09:47 AM #4 , Carl -NC . My coil is about 1.5 millihenry / 4,5 ohm , i added 4.7 nanofarads cap. in parallel . Frequency - about 1 kHz . Power supply - 2.5 volts . I really liked how it worked - there was little difference between the starting current and its magnitude before the reversal . Of course something is missing - this is not PERPETUUM MOBILE - a simple compensation circuit must be added to compensate active losses in L and C ( when the current through the coil reverses ) . Current generator ideas are unreasonable -heating energy became more expensive .

[moodz]

The key in those circuits is the supply. It has to be a constant current source. Then both the standard H-bridge and moodz circuits work. I've been playing for a while with a standard H-bridge powered by a constant current constant voltage module. Power consumption was OK. However the current pulses were not completely flat-topped but I was using a high inductance coil and didn't measure the A/sec rate. Seems that a lower inductance coil is better. Still have the circuit laying somewhere round here.

The key is NOT the supply. A constant current source has a high source impedance and cant be used in this application. The detector coil ( and associated magnetic field ) is waving around in the air anyway so the motion ( acceleration / deacceleration ) is more of an issue.

moodz.

[lucifer]

The key is NOT the supply. A constant current source has a high source impedance and cant be used in this application. The detector coil ( and associated magnetic field ) is waving around in the air anyway so the motion ( acceleration / deacceleration ) is more of an issue.

moodz.

Some refer to this as a current source inverter. For example see attached extract from an old book on power electronics. Another similar concept with some improvements is presented in the second PDF paper attached.

[moodz]

Some refer to this as a current source inverter. For example see attached extract from an old book on power electronics. Another similar concept with some improvements is presented in the second PDF paper attached.

I had a look at your references ... The h bridge is fine for non inductive loads but I think you are not going to get CC pulses in an inductive load ??.
The second reference PDF is interesting for allowing the pulses to be spaced by off periods however the pulse tops show a clear tilt ( ie not constant ) so this scheme is not better than what I proposed. ( and the pulse rise times will be impacted by storage of energy in the switched cap )

The main aim of my proposal is to generate pulses at relatively high current int the TX coil with minimal supply requirements ( eg 1 or 2 volts ). The main criteria is that the rise times of the pulses ( di/dt ) is much greater than the dwell time ( di/dt).
DSP in the RX coil takes care of resolving target information. You can achieve CC operation with a compensation network ( in 100us the current changes by just over 1 ma ) However because the coil of a metal detector is in random motion over the target a fixed field ( ie true CC ) is a moot point.
A well known company spent alot of time and money fine tuning the CC control of their high end CC pulse metal detector product ... and guess what ... it made no difference to the performance when they did effectively achieve that goal. ( short answer = no improvement ).

..just my opinion ... I respect your view there is no right or wrong ... whatever works :-)

Have a look at this patent ....

https://patents.google.com/patent/US4157579A/en

The guy uses a DC current source ... not to be confused with a CC source.

[moodz]

Here is a patent to consider :-)

https://patents.google.com/patent/US20140232408A1/en

[lucifer]

I had a look at your references ... The h bridge is fine for non inductive loads but I think you are not going to get CC pulses in an inductive load ??.

Well.. it's been quite a while since I played briefly with those circuits. I've tested the H-bridge circuit with a capacitor across the load (coil) and it works but pulses are not quite flat-topped.

The second reference PDF is interesting for allowing the pulses to be spaced by off periods however the pulse tops show a clear tilt ( ie not constant ) so this scheme is not better than what I proposed. ( and the pulse rise times will be impacted by storage of energy in the switched cap )

Yes, I suppose this circuit won't produce perfect flat-top current pulses too. The capacitor however is charged by the coil flyback voltage and then this voltage is applied to the coil for a faster rise time. Authors claim they got faster current rise times this way.

The main aim of my proposal is to generate pulses at relatively high current int the TX coil with minimal supply requirements ( eg 1 or 2 volts ). The main criteria is that the rise times of the pulses ( di/dt ) is much greater than the dwell time ( di/dt).
DSP in the RX coil takes care of resolving target information. You can achieve CC operation with a compensation network ( in 100us the current changes by just over 1 ma ) However because the coil of a metal detector is in random motion over the target a fixed field ( ie true CC ) is a moot point.
A well known company spent alot of time and money fine tuning the CC control of their high end CC pulse metal detector product ... and guess what ... it made no difference to the performance when they did effectively achieve that goal. ( short answer = no improvement ).

Agree, I suppose you never get a perfect waveform, even with compensation circuits. If you refer to ML, I've seen their patents. Even a small tilt in current shouldn't be much of a problem. To me the coil itself and RX circuit are more important.

..just my opinion ... I respect your view there is no right or wrong ... whatever works :-)

I just want to share information and my humble experience, don't mean to offend anybody or tell what's right or wrong

Have a look at this patent ....

https://patents.google.com/patent/US4157579A/en

The guy uses a DC current source ... not to be confused with a CC source.

Thanks, I'll look at that.

[Teleno]

Here's my attempt to solve the ramp problem. I've added a current limiter on each side of the bridge. When M3, M4, M5 and M8 are selected for both low Cgs and Rdson the system doesn't ring and the variation in the top current is about 100uA for a 1A pulse.

[Riss]

Here's my attempt to solve the ramp problem. I've added a current limiter on each side of the bridge. When M3, M4, M5 and M8 are selected for both low Cgs and Rdson the system doesn't ring and the variation in the top current is about 100uA for a 1A pulse.

[ATTACH]temp_50921_1673892536251_942[/ATTACH]

Ingeniously simple solution , but immediately after the current is reversed , the current starts to rise , cause - losses in the coil during the reversal of the current . After some time your circuit starts working , but the current will not be constant - a horizontal line . If a computer simulation helped you with this decision - this program is for update . I used a thought experiment , if I'm wrong - I'll update myself by hitting myself on the head …

[Teleno]

After some time your circuit starts working , but the current will not be constant - a horizontal line . If a computer simulation helped you with this decision - this program is for update . I used a thought experiment , if I'm wrong - I'll update myself by hitting myself on the head …

Nothing is constant in the Universe, there are no absolutely horizontal lines but I challenge you to get a closer approximation especially when the Tx coil is coupled to a target.

[Tinkerer]

With real circuits I found that there is a certain amount of tolerance in the TX square wave slope.
However, the slope needs to be compensated.
The way we compensate it is measuring the losses at every cycle and adjust the TX current through microprocessor control.

More and more the PI becomes digital.

[moodz]

.... a proposed solution to the ramp problem.

1.2 amp swing / 750 volt flyback / 2.6 watt consumption from 1.5 volts

moodz

[Altra]

.... a proposed solution to the ramp problem.

1.2 amp swing / 750 volt flyback / 2.6 watt consumption from 1.5 volts

moodz

Wow that looks interesting. Can't wait to see a test circuit in action.

What is the Tx frequency?

Also following the Awesome PI thread.

Thanks for posting your work.

Mark

[moodz]

Thanks Mark ...

attached is the sim file to play with.

The TX frequency is 200 microseconds ( 5 Khz ) ... with 5 microsecond sampling at start and end of current pulse to measure "tilt or ramp error".

The target switches in and out at a 10 millisecond rate to test loop recovery.

With correct timing adjustment will support other TX frequencies.

moodz / Paul.