I have been investigating a report from a user of two deployment ematches igniting instantly when a 3S lipo battery (12.6V) was connected to a Blue Raven altimeter via a screw switch. After looking into possible causes of this event, which did not happen when using the same hardware before or after, I was reminded of a potential failure mode for MOSFETs like those used on the Blue Raven.
The failure mode is called Miller turn-on, or parasitic turn on, or the Miller Effect.
The Blue Raven and other altimeters use a silicon-based solid state switch called an N-channel MOSFET.
Here is the data sheet for the one used by the Blue Raven and the previous Raven4 altimeters:
https://www.vishay.com/docs/63233/si5442du.pdf
You can think of the drain of the MOSFET, labeled "D" as the input, where positive voltage is applied. In a common configuration, the terminal "S" is connected to ground. When the gate, labeled "G" is grounded to the same voltage as the source, then the switch stays off and current can't flow from drain to source. But when the gate's voltage is raised high enough above S, then the switch turns on and current can flow from D to S through a small resistance. The gate-to-source voltage at which the current starts to flow is called the threshold voltage. The threshold voltage for the MOSFETs used in the Raven4 and the Blue Raven is 0.4V-0.9V, depending on temperature and part-to-part variation. This is low enough the the altimeter's microcontroller can fully turn on the FET with a 3.6V (Raven4) or 3.0V (Blue Raven) signal applied directly to the gate. One end of the ematch is connected to the battery + and the other end is connected to the Drain terminal of the MOSFET that's on the altimeter. When it's time to fire the charge, the microcontroller applies voltage to the gate, the MOSFET turns on, and one side of the ematch is connected to ground, allowing current to flow through the ematch, which heats up the pyrogen past the ignition temperature.
A more complicated version of the MOSFET schematic is shown below, along with an external resistor Rg that is connected to the MOSFET on the board:
Every physical transistor has some very small unintentional capacitors (parallel line symbols) that connect the three terminals. These capacitors are like tiny batteries that can supply absorb or supply current when a voltage is applied to them. They have the effect of preventing fast changes to the voltage across them.
When a battery is connected to the Drain (top) of the MOSFET, the drain terminal goes up to the battery voltage, but the capacitance between the gate and the drain is trying to keep those two terminals at the same voltage, so it also raises the gate voltage. Meanwhile, the capacitance between gate and source is trying to keep those two terminals at the same voltage, so it's pulling the gate down. If the drain voltage is raised instantaneously, then the gate voltage will get raised to a level in between the drain and the source, according to the ratio of the Cgd and Cgs capacitors. Meanwhile, the resistor Rg is pulling the gate voltage down to the source. In the steady-state case, Rg will keep the switch turned off. But for the case of instantaneous application of voltage to the drain, it takes a little bit of time before the charge stored in Cgs can get discharged through Rg. During this time, the gate of the transistor might pulled high enough to go over the MOSFET threshold voltage, and the MOSFET will turn on until the Rg has enough time to pull the gate back down again.
In the case of the MOSFETs used in the Raven4 and the Blue Raven, the Cgd is 115 pF, (Cgd + Cgs) is 1700 pF, and so the gate could get pulled up to about 115/1700, or 7%, of the drain-to-source voltage. When a 1S 4V lipo battery is used, the gate is only pulled up to 0.27V, below the threshold and the switch stays off. But if a 3S lipo battery is used, 7% of 12V is about 0.81V, which can be above the threshold voltage of the MOSFET.
To see if this could really happen on a Raven4 or Blue Raven, and if so how much, I have been doing testing of the MOSFETs on a Blue Raven.
The battery at the bottom is a 550 mAhr, 3S lipo battery, similar to the one that used during the accidental ematch firing. I have one end of an unfired ematch connected to the main terminal of the Blue Raven, and the other end is exposed and can be contacted to the battery wire to instantly apply the battery voltage through the ematch to the output terminal of the Blue Raven. I have test points attached to the main channel MOSFET gate, drain, and source. The Blue Raven has a 25 mOhm shunt between the source and ground, for monitoring the firing current. The test points are monitored at 50 million samples per second by a Saleae Logic 8 pro test equipment.
Here is the result from testing the configuration that most resembles the event that started this investigation:
When the battery is connected to the drain terminal (yellow trace), the gate voltage (red) goes up over 1V, and the source voltage (orange) also goes up, indicating that the output switch did briefly turn on. How briefly? Here's a zoom in of that event:
The source voltage is measuring the voltage across the current-measuring resistor. The resistor plus the rest of the path adds up to about 35 mOhm, so the peak current was about 6.7 Amps. But it only lasted for a fraction of a microsecond. Ematches typically take hundreds of microseconds to fire, so this one turn-on event probably only heated up the bridge wire around 1% of what's necessary to cause it to fire. In fact, in my test setup where I connected the battery multiple times in multiple ways, I was not able to cause the ematch to fire, despite attempting to make the contact bounce or have multiple contacts. This should not be too surprising, as there are thousands of Raven4 and Blue Raven altimeters in service, and this report was the first time I could recall hearing that an ematch had fired on power-up. The original poster also had used his same hardware setup multiple times for ground tests and flight before he had his accidental firing, and the he did not experience an accidental firing again when he re-tested later. My best guess as to what happened in his case is that the screw switch he was using had a particulary noisy closing, and made dozens or hundreds of contacts during the fraction of a second in which the switch was being closed, enough of them to heat up the bridgewires in both charges to the ignition temperature. The OP is kind enough to ship most of his av-bay configuration back to me, and I'll re-test with his screw switch and wiring when I receive it, to see if I can reproduce a flurry of turn-on events that would be required to ignite the ematch.
Next up: How risky are other configurations?
The failure mode is called Miller turn-on, or parasitic turn on, or the Miller Effect.
The Blue Raven and other altimeters use a silicon-based solid state switch called an N-channel MOSFET.
Here is the data sheet for the one used by the Blue Raven and the previous Raven4 altimeters:
https://www.vishay.com/docs/63233/si5442du.pdf
You can think of the drain of the MOSFET, labeled "D" as the input, where positive voltage is applied. In a common configuration, the terminal "S" is connected to ground. When the gate, labeled "G" is grounded to the same voltage as the source, then the switch stays off and current can't flow from drain to source. But when the gate's voltage is raised high enough above S, then the switch turns on and current can flow from D to S through a small resistance. The gate-to-source voltage at which the current starts to flow is called the threshold voltage. The threshold voltage for the MOSFETs used in the Raven4 and the Blue Raven is 0.4V-0.9V, depending on temperature and part-to-part variation. This is low enough the the altimeter's microcontroller can fully turn on the FET with a 3.6V (Raven4) or 3.0V (Blue Raven) signal applied directly to the gate. One end of the ematch is connected to the battery + and the other end is connected to the Drain terminal of the MOSFET that's on the altimeter. When it's time to fire the charge, the microcontroller applies voltage to the gate, the MOSFET turns on, and one side of the ematch is connected to ground, allowing current to flow through the ematch, which heats up the pyrogen past the ignition temperature.
A more complicated version of the MOSFET schematic is shown below, along with an external resistor Rg that is connected to the MOSFET on the board:
Every physical transistor has some very small unintentional capacitors (parallel line symbols) that connect the three terminals. These capacitors are like tiny batteries that can supply absorb or supply current when a voltage is applied to them. They have the effect of preventing fast changes to the voltage across them.
When a battery is connected to the Drain (top) of the MOSFET, the drain terminal goes up to the battery voltage, but the capacitance between the gate and the drain is trying to keep those two terminals at the same voltage, so it also raises the gate voltage. Meanwhile, the capacitance between gate and source is trying to keep those two terminals at the same voltage, so it's pulling the gate down. If the drain voltage is raised instantaneously, then the gate voltage will get raised to a level in between the drain and the source, according to the ratio of the Cgd and Cgs capacitors. Meanwhile, the resistor Rg is pulling the gate voltage down to the source. In the steady-state case, Rg will keep the switch turned off. But for the case of instantaneous application of voltage to the drain, it takes a little bit of time before the charge stored in Cgs can get discharged through Rg. During this time, the gate of the transistor might pulled high enough to go over the MOSFET threshold voltage, and the MOSFET will turn on until the Rg has enough time to pull the gate back down again.
In the case of the MOSFETs used in the Raven4 and the Blue Raven, the Cgd is 115 pF, (Cgd + Cgs) is 1700 pF, and so the gate could get pulled up to about 115/1700, or 7%, of the drain-to-source voltage. When a 1S 4V lipo battery is used, the gate is only pulled up to 0.27V, below the threshold and the switch stays off. But if a 3S lipo battery is used, 7% of 12V is about 0.81V, which can be above the threshold voltage of the MOSFET.
To see if this could really happen on a Raven4 or Blue Raven, and if so how much, I have been doing testing of the MOSFETs on a Blue Raven.
The battery at the bottom is a 550 mAhr, 3S lipo battery, similar to the one that used during the accidental ematch firing. I have one end of an unfired ematch connected to the main terminal of the Blue Raven, and the other end is exposed and can be contacted to the battery wire to instantly apply the battery voltage through the ematch to the output terminal of the Blue Raven. I have test points attached to the main channel MOSFET gate, drain, and source. The Blue Raven has a 25 mOhm shunt between the source and ground, for monitoring the firing current. The test points are monitored at 50 million samples per second by a Saleae Logic 8 pro test equipment.
Here is the result from testing the configuration that most resembles the event that started this investigation:
When the battery is connected to the drain terminal (yellow trace), the gate voltage (red) goes up over 1V, and the source voltage (orange) also goes up, indicating that the output switch did briefly turn on. How briefly? Here's a zoom in of that event:
The source voltage is measuring the voltage across the current-measuring resistor. The resistor plus the rest of the path adds up to about 35 mOhm, so the peak current was about 6.7 Amps. But it only lasted for a fraction of a microsecond. Ematches typically take hundreds of microseconds to fire, so this one turn-on event probably only heated up the bridge wire around 1% of what's necessary to cause it to fire. In fact, in my test setup where I connected the battery multiple times in multiple ways, I was not able to cause the ematch to fire, despite attempting to make the contact bounce or have multiple contacts. This should not be too surprising, as there are thousands of Raven4 and Blue Raven altimeters in service, and this report was the first time I could recall hearing that an ematch had fired on power-up. The original poster also had used his same hardware setup multiple times for ground tests and flight before he had his accidental firing, and the he did not experience an accidental firing again when he re-tested later. My best guess as to what happened in his case is that the screw switch he was using had a particulary noisy closing, and made dozens or hundreds of contacts during the fraction of a second in which the switch was being closed, enough of them to heat up the bridgewires in both charges to the ignition temperature. The OP is kind enough to ship most of his av-bay configuration back to me, and I'll re-test with his screw switch and wiring when I receive it, to see if I can reproduce a flurry of turn-on events that would be required to ignite the ematch.
Next up: How risky are other configurations?
Last edited: