Another Inverter build

[Warpspeed]

Yes, definitely early/mid 1970's state of the art technology.
And it still works just as well now, as it did back then.

 

[tinyt]

What is the current waveform under no load?

I remember that I bought a PZEM-004T, so I un-boxed it and it included this current transformer. It has 1:1000 ratio and the coil has a resistance of 58 ohms. So assuming that I will be measuring about 1 ampere, the load resistor will be:
Load resistor current = 1 / 1000 = .001A.
Load resistance for 1 volt = 1 / .001 = 1000 ohms
Load resistor = 1000 -58 = 942 ohms.
I have two 470 ohms and in series it will be 940 ohms close enough.

The toroid transformer has an un-used 12vac winding so I scoped it together with the voltage at the 940 ohms resistor.
Picture 1 is the current transformer and scope probe setup.
Picture 2 is the scope screen capture. Green is at the un-used 12vac winding and yellow is at the current transformer load resistor. This is the first time I have observed the SPWM switching current, thanks for asking me to do this.
Picture 3 shows I paralleled a 1 uf capacitor across the current transformer load resistor to smooth out the SPWM.
Picture 4 is the scope screen capture showing the no load current without the SPWM switching.

 

[Hedges]

The filtered current waveform looks pretty good, mostly sinusoidal and shifted 90 degrees.

I've only looked at a transformer driven by my inverters or by the grid. Not between PWM and output transformer.
That switching current is in the transformer itself? Apparently so, paralleled leads was how you wired your new one. I had though a buck converter with inductor would drive the transformer. But I haven't delved into designs. I did put a transformer on the output of a VFD to do some experiments.

External resistor on CT? Avoid getting bitten.

 

[tinyt]

Update, modifications on the chin pcb:

Beefed up with solid copper wires and more solder the high current paths on the pcb.

Because I have extra capacitors in my junk box, I added two 10,000 uF capacitors to the existing six 3,300 uF. Total capacitance now changed from 19,800 uF to 39,800 uF. The intent is for better load current spike handling, that is what some say.

Also, I removed the 220vac output sense transformer and spwm filter capacitor, the existing connection to the spwm filter capacitor is thru a connector with thin wires, I think filtering will not be effective. I will be connecting the spwm filter capacitor directly to the 220vac output of the toroid. Outut sense will now come from the 12vac toroid winding, less dangerous voltage on the pcb.

 

[tinyt]

Update.
My order for the EGS002 modules came. I need to make some modifications to prevent possible blown mosfets due to its existing design.

IIRC the modification was developed by oztules and others in different forums a long time ago. It disables shutdown of the driver chips by the on board opamp. Also, pin 17 (FANCTR) is now re-purposed to be EG8010 SPWMEN which now serves as the module ON/OFF control.

 

[tinyt]

I checked the jumpers of the EGS002, and had to jumper JP1 (60Hz) and JP2 (3 second soft start). I did not touch existing JP7 (DT0 500ns dead time) and JP8 (DT1 1.5 us dead time). Other jumpers are not present.

 

[tinyt]

In the EG8010 spec sheet, VFB must be set to 3.0 volts to adjust pwm duty cycle for desired AC output. So, I connected the 12VAC winding of the toroid to the 12vac input (was 220vac) of the chin inverter. I then connected one of the 120vac winding of the toroid to a variac set to 120vac. Using a dmm I measured VFB and adjusted the trimpot for a reading of 3.0 vdc. The inverter needs no power for this procedure. See pictures.

 

[tinyt]

With the EGS002 still un-plugged, I used a variable dc power supply to set the DC power input low voltage detect (LVD) and high voltage detect (HVD) trimpots.

I set the power supply to 45vdc and adjusted LVD trimpot so that LVD T.P. changes from 0 volt to about 11 volts. Then I set the power supply to 58 volts and adjusted the HVD trimpot so that HVD T.P. changes from 0 volt to about 11 volts.

With connector pin labeled G and K connected together, the orange relay on the pcb will be energized when the battery voltage is between 45vdc and 58vdc. When energized, the relay will apply operating voltage to the EGS002. I have also attached revised schematic.

 

[tinyt]

I bought this 200A inductor for the spwm filter, I don't know how it will perform. I also made a wooden cradle for the toroid and base for the pcb. I hope the cradle will be able to carry the 42 lbs toroid.

 

[tinyt]

Here is initial assembly with 2 ohms resistor temporarily installed to prevent smoke, fireworks, and blown mosfets.

The AC output is configured for 240V split phase (120 - 0 - 120)

 

[tinyt]

I used a dual 30V 5A power supply for battery and with the spwmen switch at off, applied power to the unit.

Looks like the 12volts buck converter in this condition draws 53.5 x .03 = 1.6 watts.

 

[tinyt]

Moment of truth:

I attached an oscilloscope to the 10vac fan power winding of the toroid and turned on SPWMEN. The power supply showed 1.34A current draw, the 2 ohms resistor heats up fast. But no fire, no smoke, and no blown mosfet.

The scope shows sine wave

 

[tinyt]

I removed the 2 ohms resistor and joined the inductor to the toroid. Applied power again, no fire and no smoke.

The scope shows sine wave with slight wiggle near zero crossing.

The power supply showed 0.34A current draw. Idle power draw is 53.5 x 0.34 = 18.2 watts.
IIRC warpspeed recommends tuning the inductor-toroid-capacitor for resonance at 1.5 times the operating frequency or 60 x 1.5 = 90Hz.
I will work on reducing this idle power draw later.

Edit: I forgot to add the AC output voltage reading. It is now attached.

 

[Warpspeed]

I bought this 200A inductor for the spwm filter, I don't know how it will perform. I also made a wooden cradle for the toroid and base for the pcb. I hope the cradle will be able to carry the 42 lbs toroid.

Those bright blue Chinese Ferrosilicon Aluminium cores appear to be something quite new that I have never tested myself.
But they seem to have a very soft gradual saturation and a very high flux capacity. which should be ideal for our application.
Results over at The Back Shed with these new blue choke cores look very encouraging.

Now we definitely need the inductance of a series choke to reduce current spikes through the mosfet switching bridge.
The self capacitance of the high voltage secondary winding looks like a dead short at 20 Khz. We can greatly reduce these current spikes with a series choke, and the higher the inductance of the series choke the lower will be the high frequency ripple current and also the inverter idling current.

Inverter idling current is made up of the toroid magnetizing current, plus switching losses mainly due to these nasty high frequency switching spikes. The higher the choke inductance, the happier your switching mosfets will be. And potentially more reliable too.

Putting some impedance between the switching bridge and the toroid can have one unfortunate effect.
This impedance rises with frequency, which is exactly what we want.
Negligible series impedance at 50/60Hz and very high series impedance at the 20Khz switching frequency.

At a few hundred Hz or a very few Khz, the choke impedance can start to become significant, and if our toroid has a self resonance in that range, a harmonic of 50/60Hz can excite ringing in the toroid, and we can get the infamous wobbles, especially under no load.
These harmonics come from harmonic distortion.
The more perfect the pwm sine wave, the lower the harmonic distortion, and resultant harmonics.

Now people experimenting with chokes see these wobbles and assume its caused by a crappy choke.
Choke A has no wobbles, choke B wobbles.
So the logical conclusion choke A must be better.
That is the wrong conclusion. Choke B has a higher inductance and series impedance and is definitely the better choke.

So now having a nice big choke and low ripple current, and low idling current, what to do about the wobbles ?

The cure is to lower the self resonance of the toroid.
This also is not intuitive.
If we just slap a capacitor across the secondary we can certainly lower the self resonant frequency, but it then may be tuned to a lower harmonic of 50/60Hz and the wobbles gain in amplitude and look even worse.

But there is a magic solution !
If we tune the toroid to exactly x1.5 the inverter frequency (75Hz or 90Hz) the wobbles will vanish.
We can then run a big efficient choke with all of its advantages without the wobbles that would normally cause.

By tuning the toroid to x1.5 resonant energy cannot build up.
It certainly tries to, but in the following cycle, the resonant build up energy is out of phase, and this self damping action can be very effective.
Tuning needs to be about +/- 1Hz to be really effective and that takes some precision to get right.
Trial and error will not be good enough, you will need to do it properly with a digital frequency counter, a tunable oscillator, and some patience.

This all sounds like voodoo magic, but it does actually work, and its well worth the effort if you have the wobbles problem.

Under more heavy, and especially a reactive load, this toroid self resonance will be de tuned, but the load itself will damp out this ringing.
We need to tune it very precisely under no load, it will then be fine.

 

[Hedges]

The self capacitance of the high voltage secondary winding looks like a dead short at 20 Khz. We can greatly reduce these current spikes with a series choke, and the higher the inductance of the series choke the lower will be the high frequency ripple current and also the inverter idling current.

Parasitic capacitance will depend on winding style (which windings are on top of which, the AC voltage between them), also dielectric thickness.

For an 11 MHz RF transformer I designed, enamel magnet wire gave too low SRF. I used Kynar wire/wrap wire, and its 10 mil insulation reduced capacitance, raising resonant frequency.

 

[tinyt]

Update:
I disconnected the toroid transformer and replaced the VFB signal source coming from the toroid transformer with a separate 12vac transformer.

This way the EGS002 will continuously run without faulting and I can measure idle power draw less the toroid transformer.
It is 53.5 x 0.14 = 7.5 watts, see attached picture. Most of this is probably to drive 24 mosfets, more mosfets more drive power needed.

With the toroid transformer, it is 53.5 x 0.34 = 18.2 watts (See post #73). So to drive the toroid transformer (with inductor and capacitor) un-loaded: 18.2 - 7.5 = 10.7 watts.

I am happy with the numbers.

 

[Warpspeed]

Those are excellent numbers, a very nice transformer and choke.
Well done !

 

[tinyt]

Those are excellent numbers, a very nice transformer and choke.
Well done !

Thank you!
Now to develop an automatic power on pre-charge circuit.

 

[Warpspeed]

I just use a push button and series resistor in parallel with the dc circuit breaker. Very simple and far less to go wrong.
Its not a big deal to push the button, wait for the ac voltage to come up, typically in two to three seconds, then close the dc breaker.
If something in the inverter is blown, the ac does not come up, so remove finger from button pretty smartly !!
That way, you might avoid a very loud bang when closing the breaker.

Early on with previous prototype inverters, I did have several charging resistors fail open circuit after only a short time, due to thermal shock.

I now use one of those gold metal cased resistors rated for two hundred watts. That will happily run into a dead short for quite a long time without getting more than slightly warm.
My system is 100v and I used a 30 ohm 200 watt charging resistor, and a ten amp rated push button.
That has worked for me reliably for five years in my current inverter without a single issue.

Previously I had a lot of problems with fully automatic pre charge systems, and smaller more fragile resistors.
This is a spare replacement resistor I bought, which was never needed.

 

[tinyt]

Update:
Using my earlier work here as reference, I used 5 mosfets HY5608 instead of high current contactor. Each HY5608 has an RDSon of 1.5 milliohms. So, 5 in parallel will be .0015 / 5 = .0003 ohm. Power loss at 150A will be 150 x 150 x .0003 = 6.75 watts. I think it is tolerable.