OCD and feedback on one transformer?

[thaumatichthys]

On the UD2 and some other drivers, I see that the OCD and feedback use separate current transformers, but would it be possible to have both functions on one transformer?

Excuse my terrible diagram, but here on the UD2.7 schematic, would it be reasonable to connect a diode rectifier, RC filter, and then a voltage divider and the OCD comparator across R1?

If this is possible, is there a reason that this isn't more popular?

[Mike]

It's certainly possible, I do it with my freewheeling driver here: https://highvoltageforum.net/index.php?topic=2054.msg15605;topicseen#msg15605.

As to why it's not more common, I'm not sure. Lots of people are using drivers they found (esp. the UD2 and derivatives) or only modifying them slightly. I was thinking of doing the same, mostly because I wanted to make minimal changes initially (less risk in doing something well proven), till Dave showed me his design which I immediately saw the benefit of.

[Anders Mikkelsen]

Diode rectifier is not viable, since one end of the current transformer is ground referenced. A full wave rectifier would short out the transformer on one of the polarities with that connection.

You can potentially just do a positive cycle OCD, i.e. a single threshold that would only catch the positive half cycle. The disadvantage is that the OCD needs to be set lower, as you are basically blind to negative overcurrents.

One option is to have a precision rectifier circuit, to get the absolute value of the ground-referenced signals, but these are a bit of work to get to perform well at higher frequencies.

Personally, I would use two comparators, one for positive and one for negative overcurrents, i.e. a window comparator. If you use open-collector parts, their outputs can simply be wired together withut further logic. If you have a single supply, the CT signal needs to be terminated to a virtual midpoint ground, to not exceed the common mode input range of the comparators.

[TMaxElectronics]

Yes that is certainly possible, though it requires some slightly different circuitry to the one you suggest. I recommend you have a look at the ZCD circuitry in the UD3.x, that only requires one CT



Or Page 4 of the PDF if you want to have a look at everything: https://github.com/TMaxElectronics/UD3.1/blob/master/Schematic%2BBoard.PDF

It also allows you to actually measure and display the peak current in your primary for every pulse with just a peak hold circuit in the PSOC

[davekni]

You can potentially just do a positive cycle OCD, i.e. a single threshold that would only catch the positive half cycle. The disadvantage is that the OCD needs to be set lower, as you are basically blind to negative overcurrents.

UD2.7 registers OCD, so allows shut-down only once per full cycle.  Since primary current is reasonably symmetric, only the one half-cycle before clocking is needed.  UD2.7 is already effectively looking at positive current only (unless for some reason negative half-cycle is large and following positive half-cycle is lower current).

[Anders Mikkelsen]

You can potentially just do a positive cycle OCD, i.e. a single threshold that would only catch the positive half cycle. The disadvantage is that the OCD needs to be set lower, as you are basically blind to negative overcurrents.

UD2.7 registers OCD, so allows shut-down only once per full cycle.  Since primary current is reasonably symmetric, only the one half-cycle before clocking is needed.  UD2.7 is already effectively looking at positive current only (unless for some reason negative half-cycle is large and following positive half-cycle is lower current).

That's a very good point, and it greatly simplifies using a single CT.

Yes that is certainly possible, though it requires some slightly different circuitry to the one you suggest. I recommend you have a look at the ZCD circuitry in the UD3.x, that only requires one CT



Or Page 4 of the PDF if you want to have a look at everything: https://github.com/TMaxElectronics/UD3.1/blob/master/Schematic%2BBoard.PDF

It also allows you to actually measure and display the peak current in your primary for every pulse with just a peak hold circuit in the PSOC

I tried this exact scheme for an induction heater, and I struggled with the phase shift being all over the place at currents low enough to not forward-bias the diodes. So the phase shift through the circuit is highly dependent on both amplitude and frequency, only being close to zero for currents large enough that the voltage dropped across the magnetizing current is large enough to forward bias the rectifier diodes. I bet it wouldn't be too bad in a DRSSTC circuit where the tank gets a solid kick at the start of the burst, but trying to make it work with a PLL was just a disaster.

The note mentions avoiding too many turns on the CT to avoid saturation, but more turns (with a given CT output voltage) should lead to less flux density in the core.

[TMaxElectronics]

, and I struggled with the phase shift being all over the place at currents low enough to not forward-bias the diodes

Are you sure about that? A current so low that the diodes don't start conducting would need to be on the order of the input leakage of the detector, I highly doubt that your feedback current would be so low.
Those diodes help you avoid phase shift by resulting in immediate >0.8V swing once the current changes direction, even if that current is very small. The only delay would come from the current discharging the junction capacitances of the diodes which is tiny.

Simulating the circuit suggests barely any phase shift throughout pretty much the entire useful range of input current. Even at 1mA peak (which is 0.08% of full range...) the phase shift is only 14°, dropping to 5° at just 5mA (all calculated agains the maximum FRes of 500kHz). Even with the 1:1000 scaling of my large coil that would only be 1A and 5A respectively on a resonant circuit designed for ~1.200A peak.

Especially in an induction heater where frequencies tend to be somewhat lower I don't see how that could cause problems.

Here's the sim at 1mA input current: