63 parts.

4 fets
2 bulk caps 63V 3300uF

5 v1 highside drive
5 v2 highside
4 v1 lowside
4 v2 lowside

5 Vdd signal contitioning

2 v1 pulse maker IR2184
2 v1 pulse maker

18 Vfb signal conditioning

3 Heat sink 10 NTC signal conditioning

3 underneath Nano
1 nano

3 15V -> 5V linear reg
some 0.1" pin headers
2 heat sink fan FET and gate pull down resistor

I use HY5608 FETs and when fed 50V, only need to switch about 40A  so I
think these FETs will be just fine.

idle current was 0.2A DC at 54V input for 239V AC output.
The toroid makes a huge difference in idle power and I am running
this one well below any hint of saturation.

I'm really quite happy with it.
Obviously I could build it with thicker V1 and V2 output cables but I
just could not be arsed, really.

It runs the picoverter code, which features low DC voltage shutdown
with auto restart, fan control for the heat sink and heat sink over temp shutdown.

The semiconductors are:
HY5608 80V 360A 1.5 mOhm Rds(on)
IR2186 gate drive IC
FOD3182 gate drive IC
LM358 dual opamp
LM7805 linear reg
any old logic level FET to drive the fan

it worked first time I applied power!


Edited 2026-01-25 15:50 by poida

---

wronger than a phone book full of wrong phone numbers

it works well, only tested it to 1330W AC so far.

for the 1330W AC output (0.99 power factor), input is 54.7V at 29.1A
1590W in, 1330W out, it's 83.5 % efficient

Rather thin wire is used in the primary circuit, it's 4mm2 I think.
this wire is the primary winding and I need 3 meters of it.
I probably could pick up a few more % if thick stuff was used.
the DC resistance of the primary is 48 mOhm.
(I determined that by putting 5.8A DC through it and reading 281mV across it)

here is a thermal image of it running under this load.



the hottest part is where the low and high side FETs are soldered
sort of where "VS1" and "VS2" are located in the above image.

Nice little board.  Is it 1 oz copper?

Also, curious if you have observed false MOSFET gate turn-on transients like the ones in your past reports on larger inverters.

it is 1oz copper, if it shows promise I will make a v.2 with 2 oz copper, error fixes, unused footprints removed and space for 3rd main cap.

I have 3 similar inverters to compare.
this one, the first version of it that uses TIP41/42 transistors in the totem pole
type gate drive and Wiseguy's inverter with it's professional design.

All 3 require about 1330 W input to drive the same load. Maybe Wiseguy's inverter
used 1300 W but the load varies over time so it's hard to get it accurate.

The next thing to examine is gate drive signal shapes and look for shoot through
at idle and under load.

I want also to play with more or less capacitor powering the drive side
of the FOD3182 chips. At the moment it is only 1uf ceramic
with ESR of about 3 Ohms.
In the past I have used 10uf electro (ESR of 10 Ohms) with a 0.1uF ceramic parallel.  

I varied gate drive voltage while running under load and changing from 12 to 15V
made no difference to input power.

Finally, I might need to put "some" resistance in the gate drive, at the moment
the FET gate is direct connected to FOD3182 output and I suspect there is
MHz oscillation. No doubt that is causing some shoot through.

That project looks pretty good,
Is it going to be usable or programmable for 24 volt as well?
Nice looking board, I like the fact that most components are through hole ones.
That makes it much easier for us fossils to solder
Pete

for sure it can be working for 24V.
it's always about the current so I would expect only 1kW from it at 24V

I am just seeing if this thing is useful and it's early days.

  poida said  

it is 1oz copper, if it shows promise I will make a v.2 with 2 oz copper, error fixes, unused footprints removed and space for 3rd main cap.

I have 3 similar inverters to compare.
this one, the first version of it that uses TIP41/42 transistors in the totem pole
type gate drive and Wiseguy's inverter with it's professional design.

All 3 require about 1330 W input to drive the same load. Maybe Wiseguy's inverter
used 1300 W but the load varies over time so it's hard to get it accurate.

The next thing to examine is gate drive signal shapes and look for shoot through
at idle and under load.

Cool.  Look forward to your findings.

  poida said  

I varied gate drive voltage while running under load and changing from 12 to 15V
made no difference to input power.

Not surprised since the gate caps are fully charged by 10V according to the datasheets.  I have tested some HYxxx MOSFETs at 10V and 12V and didn't see any significant difference.

  poida said  

Finally, I might need to put "some" resistance in the gate drive, at the moment
the FET gate is direct connected to FOD3182 output and I suspect there is
MHz oscillation. No doubt that is causing some shoot through.

That's a bit surprising.  Besides shoot through, the RFI/EMI can be pretty bad as well without any external gate resistance.

time to look at Gate voltages.

first is the switch on/off for the totem pole board, using a 10R current limit resistor and a 1N4148 diode for faster gate control when pulling it to ground (or source pin)
The drive current is supplied by a pair of TIP41 and TIP42 transistors.

test conditions are identical for both boards.
54V DC in, 200W out.
Yellow is low side gate voltage
Dark Blue is high side gate voltage



and



this is not too bad, not a lot of high frequency ringing.

Now we see the new board, with no gate current limiting resistor, nor the diode.
I connect the gate direct to the FOD 3182 output pin.




and




certainly we can see much faster switch on and off but there is now a lot
of Mhz or more ringing. This is seen in the Dark Blue trace when it goes from low voltage to about 15V, pulling the inverter output up to 54V.

I know that there should be a gate resistor according to just about all
power electrics design practice but I wanted to see what it looks like
without it.

One aim of this design is to have the smallest amount of "stuff" that makes up
the high side gate drive. I wanted to minimise the capacitance and inductance of this
by having just the drive side of the FOD3182, it's supply cap and 2 inches of thin PCB trace that follows the high side output as it switches from 0V DC to 54V DC at 20 kHz. I have got what I wanted.

time to see if a 4R7 gate resistor helps things. And after that I will add the 1N4148
for faster switch off.

things have changed after fitting the 10R resistor (I had no 4R7)

first, just the resistor:



and




the high frequency ringing is gone. as are the fast gate voltage slopes.
This is to be expected, less current into the capacitance of the FET gate will
mean slower rise and fall times.
And look at the second peaks. this is new to me.
I think the diodes will help with those.

now I also add 1N5819 diodes. These are Schottky type and switch fast, and
a voltage drop of about 0.5V



and




now the gate voltages (and therefore the switching performance of the FETs)
are looking good. I see very low voltage on the LOW side FET gate after
the high side FET switches on, pulling the inverter output up to DC supply voltage.
This is when shoot through is most likely.

there still is some high frequency voltage on the high side gate and this is when we
start thinking about ferrite beads.

here are these waveforms when under 1300W AC output load.

they look not bad and I see they show almost identical features
as seen in the above low power testing.

I see the low side gate easily getting high enough to cause some
shoot through.
maybe it's time to review what I found with ferrite beads since I see
some scope for improvement.

for reference, here is Wiseguy's gate voltages under the two tests
200 W and 1300 W AC output.

first is 200W:



and



I see the low side gate is being yanked up as the high side switches on.
not too much no doubt.

and for 1300 W output I saw this:



and


 
clearly the low side FET gate is being pulled high enough
to make it switch on while the high side FET is switching on.
this is shoot through.

overall the voltages seem better controlled than my 1 FET board as seen above.

I probed the gate voltages using the isolated probe and one non-isolated probe
which was on the low side FET's gate.

The impedance presented to the gate of the low side mosfet when switched off determines to some degree how high its gate voltage will be pulled up as the high side device turns on.

You could try lowering the gate resistor to 3.5r and see how it performs, and again with the schottky diode in parallel with it.

Another thing to try is placing a small cap 470pf directly across the gate\source of the low side device, this may help with shoot through.

Cheers
Mike

thanks for the tip
I do not have things like 3R5 resistors here at home.
it's probably worth a try tho.
I wanted to compare the old TIP41/42 thing with the new board so
I had to use 10R resistors.

I am curious about the idea of adding capacitance to the Gate.
when I add a cap to the gate pin, it still has to "operate" via the pin's lead
since this voltage jump is generated from the chip's internal capacitance that is coupled to the drain/source voltage difference.
I would guess there is a balance of this and that which could be found that might be just right.

I would love to get right in there and disable this coupling.
but it just is there.

That small cap (ceramic) is a very low impedance and can absorb some of the miller pulse getting back into the gate; too large and its going to stress out your driver chip...

AS an experiment: remove the gate drive resistor to one of the L0 mosfets, short the gate directly to its source pin via a similar resistor, this removes the inductance and any transmission line affect back to the driver IC.

Power it up, yeah the output wave form will be crap, but place your scope probe across the shorting resistor, see what gets reflected back by the miller effect as the H0 mosfet turns on; if the wave form is heaps better, then you know its circuit board layout dynamics affecting the gate drive, perhaps necessitating a better layout.

Aim to have the output of the mosfet driver 5-10mm or so from the gate pin, remember the gate circuit is a transmission line, the longer and thinner it is, the more time it takes for the driver transition to get to the gate, similarly a miller pulse getting into the gate from H0 mosfet turning on, takes longer to be negated by the impedance of the driver sitting at 0 volts.
Edited 2026-02-09 18:27 by Solar Mike