Computing stuff tied to the physical world

An improved MPS circuit

Now that we’ve seen what the voltage monitor does, and how startup can still fail under certain conditions, it’s time to try and resolve everything. So let’s make a few changes:

  1. instead of two LEDs, we’ll use a special TLV431 “shunt regulator” IC
  2. let’s avoid the RC delay by omitting the capacitor on the NPN base
  3. and finally, we add the 1 MΩ positive feedback to get a better “snap action”

Here is the new circuit:

Screen Shot 2015 02 24 at 11 23 22

The TLV431 is an interesting device: it’s essentially an “adjustable zener”, where the chip will do whatever it can (by conducting more or less) to accomplish one goal: maintaining a fixed voltage of 1.24V between its “reference” pin and its “anode”. For your convenience:

Screen Shot 2015 02 24 at 17 00 50

So as long as it can, the TLV431 will try to keep 1.24V across R2. This corresponds to a certain current, and since R1 carries the same current, it will have ≈ 0.6V across it, for a total zener-like effect of 1.82V across the TLV431. At all times, if there is enough voltage.

What this also means is that R3 + R4 will have whatever additional voltage there is from the power supply. Another way to put it is that R3 + R4 will have Vres – 1.82V across them.

The red LED now plays a completely different role: it just limits the voltage across R3 + R4 to about 1.8V. As a result, Vref cannot rise above about 3.6V, which is also the limit of what we should put on the LPC810 and the wireless radio (yet to be attached). Very convenient.

Diode D3 drops the voltage by about 0.6V, and NPN transistor Q1 starts to switch on when its base rises to about 0.6V. This happens a bit sooner than expected perhaps, due to the idle current of the TLV431 itself.

Lastly, we have the 1 MΩ resistor feedback, which will start pulling the base up the moment Vdd rises, i.e. when the transistor starts to switch on, it’ll enter a runaway mode where its base rises further, causing it to switch on even more, etc – until fully on, i.e. “in saturation”.

You can see all of this happening nicely in the following oscilloscope capture:


The blue line (CH2) is the transistor’s output voltage on its collector. Around 1.8V it starts dropping sharply with respect to the supply’s Vres. By the time Vres reaches 2.2V, the whole circuit tips over and switches on very rapidly, due to the feedback resistor.

Meanwhile, the collector voltage drops below ground level! This is due to sudden discharge of the P-MOSFET’s gate charge (in combination with the Miller effect). It’s actually quite a welcome side effect, as it helps turn the P-MOSFET on even faster.

So the TLV431 helps create a really well-defined switching point, and the feedback resistor helps the circuit to behave in a very definitive and extremely quick manner. Note also that there are no RC-based delays anymore – this circuit is now purely DC-level based.

Now let’s find out how it behaves with slower supply changes, both up and down…

[Back to article index]