Automotive Start-Stop … Keep the lights ON!

Mike Donnelly's picture

You’ve probably noticed when you start your car that some of the optional electronics turn off, such as the radio and some LED lighting circuits. That’s because the high current draw of the starter motor pulls the system voltage far below the normal 12+ volt operating level. Critical circuits are designed to operate at these low voltages … if the ignition circuit didn’t work, you would be cranking for a very long time! But many functions are not designed for that low-voltage condition. After all, it’s a once-per-trip occurrence, and drivers expect the drop-out behavior.

But for new vehicles with “Start-Stop” operating mode, the engine (ICE) is turned off frequently during idle periods to improve fuel economy. These same low-voltage drop-outs must be avoided during re-start. The radio must play on, even in heavy traffic!

The ON Semiconductor® NCV8876 is designed for this exact purpose. It is a current-mode boost converter controller, and it includes many integrated functions to reduce the complexity of the external circuit. A typical application circuit is shown in the SystemVision Cloud schematic below:

The simulation results show the output to the fixed load resistor during battery drop-out and recovery. In this application, the output voltage (dark blue waveform) is maintained above 6.4V during the drop-out transient, and regulates at 6.8V during sustained low voltage (4V) battery operation (orange waveform).

Note: This schematic is “Live”, you can pan and zoom (like Google Maps), and you can move the waveform probes around and look at signals on any of the nets (i.e. “wires”) or within any of the components. You can also double click on components to see their parameter values. If you choose "View in", you'll be able to save a copy and then change and re-simulate the circuit as you please.

This circuit also includes a soft-saturation inductor component from Coilcraft®. The load current for the XAL4030-332 inductor in this application is 4A during nominal 12V operation. But during boost operation, the current reaches 6.6A peak (light blue waveform). This could saturate a typical inductor if it were sized for the nominal load, resulting in a collapse of the effective inductance. But notice that the actual instantaneous inductance (green waveform) only drops to 2.2uH, for this nominal 3.3uH part.

If that 33% drop in inductance doesn’t sound small, the next example shows what can happen if a typical “non-soft-saturation” inductor carries current beyond its operating limits, even momentarily.  Inductance collapse leads to large current spikes that can potentially damage other power electronics components. Fortunately in this case, the “hiccup-mode” overcurrent protection capabilities of the NCV8876 limit the magnitude of the current spikes to a level below the ratings of those other components.

The simulation results show that in this case, the boost output voltage (dark blue waveform) falls to approximately 5V instead of the proper boost set-point of 6.8V, when the battery voltage (orange waveform) drops to 6V. Because the NCV59302 Linear Regulator has a very low drop-out voltage, the current through the 2.3 Ohm load resistor is maintained at just under 2.2A throughout the entire operation.

While this load current level is within the rating limits of the inductor (2.3A RMS Max.), it is approaching the saturation inductance “cliff” for this non-soft-saturating part. So when the NCV8876 activates boost control, it switches on the Power MOSFET briefly, effectively grounding the low side of the inductor through the small current sense resistance (0.03 Ohms). If the nominal 3.3uH inductance value were maintained, this would result in a slow current build-up:

     V/L = di/dt = 6V/3.3uH = 1.8A/us

But in fact the inductance collapses with further current increase, causing the di/dt value to become very large. The inductor current spikes to well over 10A (light blue waveform), and the corresponding inductance crashes to a small fraction of its nominal value (green waveform).

The current spikes could have gone much higher, possibly resulting in damage to the NVGS3130 Power MOSFET, which has a rated pulse current maximum of 19A. If you zoom in on the magenta waveform, you'll see the Ids current rising, slowly at first while the inductance is intact, but then rapidly rising to over 11A as the inductance collapses, all in just 100ns! Fortunately the gate voltage (brown waveform) is cut off to prevent further current rise. This is thanks to the overcurrent protection feature of the NCV8876. When an overcurrent condition is detected, the device immediately goes into “hiccup-mode”, in which the gate drive is turned off and remains off for a count of 1024 clock cycles. After the mandatory hiccup period, the NCV8876 reattempts boost operation, but with continued overcurrent monitoring. Note that the current spikes are repeated with just over a 2ms period, because the clock is programmed to just over 2 us period (450 kHz switching frequency) in this design.

Click here to learn more about the capabilities of the ON Semiconductor NCV8876.

Click here to learn more about soft-saturation inductor technology from Coilcraft.

The author would like to thank Robert Davis, Automotive Power Supply System Architect at ON Semiconductor, for his contributions to both creating the high-fidelity model of the NCV8876, as well as providing key design insight for Automotive Start-Stop Applications. The author would also like to thank Chris Hare, Technical Marketing Engineer at Coilcraft, for his help in developing inductor models with accurate saturation behavior.


Norm's picture


I've modeled DC-DC converter chips for semiconductor companies using commercial versions of Spice. I admire these simulators but I always wished I could develop the models using VHDL-AMS instead of equivalent circuitry. This NCV8876 model exemplifies the benefits of VHDL-AMS.

Consider for example, the programmable clock generator. The Spice model is constructed using B-sources to derive the clock period, a VCO part to convert that period into a frequency base, a timing capacitor to generate a blanking pulse and logic to pull all of this together. To complicate the modeling effort there is a clock jitter phenomenon that derives from the simulator's transient timestep control.

Compare this to the Programmable Frequency Oscillator Process in the NCV8876 model. A simple mathematical expression defines the clock period an if-then-else construct generates the pulse train with blanking. The model is defined outside the scope of the simulation algorithm so there can never be a clock jitter problem.

In short, the SystemVision model is simpler, more comprehensible and more dependable. As usual, nice work Mike!


Please log in to comment.Log in