# How to drive a high side MOSFET with an optocoupler correctly?

I am trying to create an absolutely minimal motor driver, to create a <5$low power odrive. But since I am only a software developer, I do not have a 100% understanding of analog components, and need some help for calculating the right values. I am following this web page, which describes the calculations: https://learnabout-electronics.org/Semiconductors/opto_52.php I have also read this answer on this site: How to drive a MOSFET with an optocoupler? My schematics looks like this: For simplicity, I am using only OnSemi components. The goal is to drive a 24V around 2A BLDC motor at around 30kHz PWM. ECH8660 is a dual complementary MOSFET, 30V, 4.5A, 59mohm, -30V, -4.5A, 59mohm. Since KiCAD did not have the necessary symbol, I had to create one by copying parts from other symbols. My first question: The P Channel MOSFET is mirrored, so the Source looks up, and both MOSFETS have their body diode pointing upwards. Is this symbol correct? FOD817A is an optocoupler with Continuous Collector Current 50mA max. It has the following CTR: My second question: If the MCU's datasheet lists max 20mA output per output pin, then is driving by 10mA an acceptable choice, or should I go lower if it is possible? My third question: The FOD817A datasheet lists information for other versions, and actually I wanted to use FOD817C: CTR 200–400% (instead of FOD817A: CTR 80–160% which I guess is on the graph), but I cannot find the CTR graph for that. How can I deduce that information? Using the above linked Stack Exchange question, I have both a pull-up and pull-down resistor (R2 and R3), which should pull down the MOSFET gate to +12V (it has an absolute maximum Gate-to-Source Voltage of +-20V). Fourth question: Is -12V a conservative drive voltage, or should I use different resistors? (Having identical R2 and R3 would make the BOM simpler.) The above linked tutorial web page has two examples, both driving a PC817 at 5V/4mA. In the first example, the output power supply is 5V, and it calculates 4.6mA output (at 115% CTR). But in the second example, the output power supply is 12V and it says that: As the CTR of the PC817 is around 115% the phototransistor can supply about 9mA as the supply to the phototransistor output is now taken from the 12V motor supply. Fifth question: Should I follow this advice and calculate with 10mA * 125% * (24V / 5V) = 60mA instead? Or should I use 12V only in the calculation, because the two resistors halve the voltage at the optocoupler? (Obviously I want to charge/discharge the gate the fastest possible.) Sixth question: How to calculate what kind of D1/D2 I need for handling the back EMF? Seventh, final question: Is it possible to eliminate D1/D2 and just use the body diode? (I know that it is only a parasitic crap diode.) I ask this, because I see the following prices (for 1K+ units at OnSemi site): • ECH8660:$0.2071 @ 3K (seems to be cheaper than two separate N and P channel MOSFETs)
• FOD817A: $0.092 @ 2K • MBR130T1G:$0.0648 @ 3K (1.0 A, 30 V, Schottky Power Rectifier)

So two diodes + dual MOSFET costs 0.3367$while a ECH8690-TL-H costs only$0.2913 @ 3K. (It is 60V, 4.7A, 55mOhm, -60V, -3.5A, 94mOhm, protection diode built in.)

Thanks, any help is greatly appreciated!

• Diodes are already as part of MOSFET. Why you add external? – user263983 Apr 12 at 14:50
• NoiseEHC - Welcome :-) You said: "I think I am not allowed to link to all the datasheets, because I am a newbie here" If you aren't able to link to the datasheets (I don't remember the limits) then add the links in a comment and one of us will add them into the question for you. (FYI putting 7 different questions in one topic is quite extreme & may be downvoted. Such "whole project" questions tend not to do well, as you are effectively asking someone to write one answer - remember you can only accept one - that answers all those points better than other answers. That's a big "ask".) – SamGibson Apr 12 at 14:52
• For a s/w developer you have defined your problem better than anyone I’ve seen on this site. the main problems are Crossover Shoot Through deadtime to prevent shorting Pch+Nch across V+ to gnd and then the load reactance L/R = Tau needs to be defined that causes dynamic losses during switching. It’s a complex problem and better that you included all the questions. But learning the tricks of H bridges or half bridges comes from reverse engingeering commercial solutions that have solved these questions. – Tony Stewart EE75 Apr 12 at 14:52
• Please define load L, DCR with power levels to be computed in watts so energy in Joules to be switched from start to steady state. also acceleration demand affects surge start current to be controlled with low side PWM that generates a high side Boost voltage for preferred dual Nch half bridge. I assume non- reversible. – Tony Stewart EE75 Apr 12 at 14:57
• At 30kHz you have only 33usec per cycle so your on time might be 6-27usec for 20-80% control. Those optoisolators might take 30usec to switch off. Even at a much lower PWM rate your shoot-through will be problematic without a lot of added dead time. Suggest you run some simulations or buy a few dozen parts and be prepared to burn them out. – Spehro Pefhany Apr 12 at 14:58

The P Channel MOSFET is mirrored, so the Source looks up, and both MOSFETS have their body diode pointing upwards. Is this symbol correct?

Yes.

If the MCU's datasheet lists max 20mA output per output pin, then is driving by 10mA an acceptable choice, or should I go lower if it is possible?

It will probably be fine. Just make sure the total current on all I/O pins doesn't exceed the maximum permitted.

The FOD817A datasheet lists information for other versions, and actually I wanted to use FOD817C: CTR 200–400% (instead of FOD817A: CTR 80–160% which I guess is on the graph), but I cannot find the CTR graph for that. How can I deduce that information?

The relative characteristics should be similar, so just multiply CTR by 200/80.

Is -12V a conservative drive voltage, or should I use different resistors? (Having identical R2 and R3 would make the BOM simpler.)

12 V is plenty. The highest specified Gate drive voltage is 10 V.

Should I follow this advice and calculate with 10mA * 125% * (24V / 5V) = 60mA instead? Or should I use 12V only in the calculation, because the two resistors halve the voltage at the optocoupler? (Obviously I want to charge/discharge the gate the fastest possible.)

Not sure what that calculation refers to, but the optocoupler can't handle 60 mA.

Anyway it doesn't matter which calculation you use - it won't charge/discharge the Gate 'the fastest possible', and will be woefully inadequate for accurate PWM at 30 kHz. The optocoupler's inherent delay is 3μs, 9% of the PWM period. The pullup resistance and Gate capacitance will make it worse. With 1k you could be looking at a turn-off delay of 20μs or more.

How to calculate what kind of D1/D2 I need for handling the back EMF?

They should be rated for an 'average forward rectified' current equal to or higher than motor current / 3 if using just high or just low side PWM, or motor current / 6 if alternating high and low side PWM.

Is it possible to eliminate D1/D2 and just use the body diode? (I know that it is only a parasitic crap diode.)

Yes. You should then factor in the body diode loss when calculating MOSFET power dissipation. At lower PWM ratios This could be much higher than the loss due to RDSon.

You may want to consider changing the optoisolator for another type of galvanic isolation technique such as capacitive or inductive which have a higher bandwidth.

Texas Instruments has a great series of videos on the subject of galvanic isolation which discusses inductive, capacitive and optical isolation techniques.

For example, here's the data sheet for the Silicon Labs SI823Hx which is an isolated gate driver in one chip for less than \$5.