Different responses for different MOSFETS in flyback driver circuit

[icecats]

Hi,
I am building a high voltage flyback driver circuit for an electrostatic sprayer concept that I am experimenting with. I successfully constructed the circuit with an IRLZ44N logic level MOSFET and was able to drive it directly with a 5v Arduino PWM output. However, inspection showed that the drain-to-source voltage can be in excess of hundreds of volts which exceeds the IRLZ44N's V_ds_max voltage of 60v.
Therefore, I decided to implement a MOSFET with a higher V_ds_max. No logic-level MOSFET's were available at the ratings I am looking at, so I decided to use a MOSFET driver.
Here's a description of my circuit to accompany the attached schematic:
- the circuit is powered by a 3.7v 18650 battery
- A TPS61040-based 12v boost converter (Adafruit) boosts the battery voltage to 12v
- The 12v is fed into the gate driver along with a couple of decoupling caps to provide local high current source
- The gate driver accepts a PWM control signal from a 3.3v GPIO on an Arduino/STM32
- The gate driver switches the MOSFET which interacts with a flyback transformer in a typical flyback driver arrangement.

I acquired a plethora of MOSFET's to test with. I found that some MOSFET's worked in the circuit and resulted in an arc, while others failed to produce an arc.
- Worked: IRLZ44N, IRFZ44N, IRF510N, IRF520N, IRF530N, IRF540N, IRF640N, IRF3205
- Did not work: SIHU2N80AE-GES, IRF840HPBF, IRF840BPBF, IRF740N, IRF840N

What I've tried:
- The MOSFET's that did not work tended to have higher V_ds_max ratings, but I didn't spot any other significant differences in the datasheet.
- I verified that the gate driver was producing a clean pulse above R_ds_on voltage for both working and not-working MOSFET's.
- I measured the current draw from the battery and noted that working MOSFET's drew about 1.6A while not-working MOSFET's drew 0.8-1A
- I scoped the MOSFET drain pin and noticed a significant difference in the output waveforms.

What characteristic am I missing that determines if a MOSFET will function as expected to drive the flyback circuit?

[T3sl4co1l]

What is the transformer?  ~20kHz seems slow, but maybe it's fast, no idea.

Note that you may have a wide selection of counterfeit transistors.  You didn't mention from where you obtained them.  Your seemingly erroneous measurement of the IRLZ44N seems to support this.  But you need to eliminate the scope and probe, and circuit wiring/layout, as possible culprits first (I'm assuming these are both random and poorly defined things).

3.7V isn't much input, and you will need quite a large turns ratio (implied: low primary turns count) to get anything sparky out of it.  This seems to prioritize high current over high voltage.  What did you notice about the Rds(on) parameter of your selection of MOSFETs?

You missed a ground connection to the battery, but as the circuit otherwise works I'm assuming that's just a schematic typo.

Tim

[temperance]

I successfully constructed the circuit with an IRLZ44N logic level MOSFET and was able to drive it directly with a 5v Arduino PWM

To drive a MOSFET full into saturation, you will need to raise the gate source voltage high enough. For example, the IRF840 requires a gate drive voltage well above 5 V. This can be seen in the data sheet, fig. 6

https://www.vishay.com/docs/91070/91070.pdf

The sIHU3N80 requires an even higher gate source voltage. See Fig. 7 in this data sheet:

https://www.vishay.com/docs/92343/sihu2n80ae.pdf

The IRLZ44N works because Vgsth is around 3 V. (data sheet fig.6)

To drive a MOSFET at a certain speed requires a certain current. The switching speed can be calculated from the charge required around the miller plateau voltage. Taking the IRLZ44N as an example we get Qge Q @ Vgsth= 17.5 nC. If the drive would be able to supply 1 A current to the gate, the MOSFET would switch in about 17.5 ns. Of course the micro controller cannot supply such currents. The current the micro can supply will around 50 mA @ Vgsth which translates into 350 ns. Whether 350 ns is fine, depends on the switching frequency, drain source voltage and the switched current and the allowed power dissipation.

[icecats]

What is the transformer?  ~20kHz seems slow, but maybe it's fast, no idea.

Note that you may have a wide selection of counterfeit transistors.  You didn't mention from where you obtained them.  Your seemingly erroneous measurement of the IRLZ44N seems to support this.  But you need to eliminate the scope and probe, and circuit wiring/layout, as possible culprits first (I'm assuming these are both random and poorly defined things).

3.7V isn't much input, and you will need quite a large turns ratio (implied: low primary turns count) to get anything sparky out of it.  This seems to prioritize high current over high voltage.  What did you notice about the Rds(on) parameter of your selection of MOSFETs?

You missed a ground connection to the battery, but as the circuit otherwise works I'm assuming that's just a schematic typo.

Tim

Thanks for the reply! My flyback transformer is pulled from one of those ubiquitous Amazon "bbq arc lighters" so specs are a bit hazy. I have found a similar-looking model on Alibaba and have reason to believe that the turns ratio is 90:1.
I did a characterization of Rds(on) resistance (see attached). Whether or not the circuit "works" appears to be correlated with this. It seems like the higher resistance MOSFET's are limiting the current to the device.
Now I guess I need to try to optimize the conflicting parameters of high V_ds and low Rds(on). I think that I need to:
1) Determine what minimum voltage is necessary on the primary side of the transformer to make the flyback work.
2) Clamp the MOSFET ds voltage at this minimum (not sure how to do this).
3) Choose a MOSFET with a higher rating than the clamp AND with a low Rds(on) resistance per my characterization.

[werediver]

What is the spark gap width in your set-up? Air takes 1–3 kV/mm to break down. When you're saying "not working" it could still work with a smaller spark gap.

Also, I'd suggest bypassing your power source (the battery) with at least 100 uF electrolytic capacitor right near the transformer primary and the MOSFET. That way you'll get a lower internal resistance of the power source for when the current demand is highest.

(and C4 of 2.2 F must be a mistake; is it 2.2 uF?)

Just this weekend I put together a dead simple spark generator using a small HV flyback transformer and an IRLZ44N driven directly by a 3.3 V GPIO without any snubbers or clamps. It's barely working with a spark gap of about 1–2 mm, the discharge sound is much quieter than I expected (much quieter than a mouse click), but technically it does.

I could improve the performance of my set-up by adding an RCD clamp and possibly a snubber network, after which I can drive the MOSFET harder, especially into the off state (which will increase the flyback voltage spike, but also the output voltage).

The simplest approach to improve MOSFET driving would be to add a diode + PNP BJT to speed the turn-off of the MOSFET (still driving directly by a GPIO). An overall better solution would be to use a proper driver (or a discrete buffer with voltage translation) and a higher gate voltage (10–12 V).