In the presence of high EMI, normal sensors pick up too much noise which covers up the true waveform on the oscilloscope. Thus Hall effect sensors are not helpful. Resistive voltage drop sensors yield a low signal to noise ratio !
If the converter current does not contain much DC, a ferrite ring core (15mm outer diameter) with 20 turns secondary and a two ohm load does marvelous things:
0.1V/A. I used a lot similar designs.
- Hall sensors have to be screened from dv/dt
- Typical shunts may have 10-100nH inductance, so they make the derivative of the current.
- If current flows in the mass of power electronic circuits a lot of HF voltage is present in circuits (ghost signals).
Thanks Alex, your observation of false signals matches with mine. But most of the cases there is dc component in current since we try to monitor the switching device current, so any further suggestions ?
If the DC component is pulsed, for instance max 85% duty ratio, and limited time length, one can put a (schottky) diode in the secondary of the core, it even works better with amorphous cores. It results in rextremely clean signals.
It acts like somewhat like a fly back. It results in very clean signals too, but needs calcilations to check if no saturation is reached.
With shunts one can try common mode filtering with ferrite and a better lay-out, a low pass filter. With hall: test the sensitivity to dv/dt.
Abstract:This document presents a method to optimize integrated
LDMOS (Lateral Double-Diffused MOSFET) transistorsfor use in very high frequency (VHF, 30-300 MHz) dc-dc converters.A transistor model valid at VHF switching frequencies isdeveloped. Device parameters are related to layout geometry andthe resulting layout vs. loss tradeoffs are illustrated. A methodof finding an optimal layout for a given converter applicationis developed and experimentally verified in a 50 MHz converter,resulting in a 54% reduction in power loss over a hand-optimized
device. It is further demonstrated that hot-carrier limits on devicesafe operating area may be relaxed under soft switching, yieldingsignificant further loss reduction. A device fabricated with 3 umgate length in 20-V design rules is validated at 35-V, offering
reduced parasitic resistance and capacitance as compared to the5.5 um device. Compared to the original design, loss is up to75% lower in the example application
Optimization of Integrated Transistors for Very High Frequency dc-dc Converters
Anthony D. Sagneri,, David I. Anderson, David J. Perreault
IEEE Transactions on Power Electronics, Vol. 28, No. 7, pp. 3614-3626, July 2013
in case your converter has enough space available, you can try including a Pearson probe instead of shunts, Hall or Rogowski.
Though Alex is right regarding basic shunts, there are - expensive - coaxial measurement shunts with parasitic incuctance far below 1nH if you're willing to spend the money.
The coaxial shunts are indeed better, but still a common mode current can appearas they are linked to the circuit and mass currents flow trough the circuit, so still a common mode current appears, generating inductive drops elsewhere in the circuit. So it is sometimes better to use a normal shunt with 20nH and 10mohm giving a time constant of 2µs and to compensate that time constant with a 2µs low pass close to the place where you need it. However, all of these methods are invasive in the circuit.
If no precision is needed but just to compare a converter with another identical one, the following trick can be used. This is using a small SMD inductor (wound I-core) to sense the near field. The smd inductor should be capacitively screened (coax) and loaded with a damping resistor (to avoid resonances with the capacitance of the probe coax). It mainly senses di/dt, but I tried it and one can distinguish also resonances in the drain current and also in the gate current. The advantage is that it is non-contact (no mass current) and non invasive. One still can recognise the duty ratio and even some aspects of parasitic resonances in the supply, or just compare signatures in a qualitative way. However I don't think of using it yet for a reliable current-feedback.