3
\$\begingroup\$

I'm buiding a 60Hz sine wave inverter based on 19Khz PWM generated by a PIC16F883 (lookup table) using a half bridge with 4 x IRF2807 - switching a pair of low-side/high-side. I'm using my own designed MOSFET driver with discrete transistors. I'm feeding the bridge with 12V and using a 700VA transformer to step up the voltage from 12V to 127VAC.
I still have some small wave distortion specially with inductive loads such as a fan motor (in the picture, an example with a 60W/127VAC fan).

Question: Looking at the oscilloscope screen capture, do you guys think I should go further in reducing the deadtime to enhace the wave quality and still be safe to avoid shoot thru current? Or should I keep it and go for the final PCB, since the circuit is still on protoboard, so parasitc inductances may be present. Each vertical division means 10us in the PWM wave. I fixed the duty cyle just to capture the image.

Thank you!

enter image description here

enter image description here

\$\endgroup\$
4
  • \$\begingroup\$ Any reason you need less distortion? You're already way ahead of any of the quasi-sine inverters. Anything you're trying to power that would be sensitive to the distortion you're seeing? \$\endgroup\$
    – John D
    Commented Jul 1, 2016 at 3:03
  • 1
    \$\begingroup\$ The right amount of dead time is whatever delay keeps the upper and lower fets from turning on at the same time under worst case conditions (like full design current at some higher than normal temperature), if you've got a really strong gate driver, the fets usually switch faster so you can use a shorter dead time. Check how fast each fet is switching if you can, if the dead time's equal to or greater than the transition time then you should be ok I'd imagine. (some big IGBTs might need several uS, while some MOSFETs might only need tens of nS, having more isn't really going to hurt though) \$\endgroup\$
    – Sam
    Commented Jul 1, 2016 at 6:13
  • \$\begingroup\$ My idea is to build an inverter to power all the home outlets based on battery/solar power - So it would have to be generic and able to power any equipment, from refrigerator to a home theater. I'm starting the prototype small (700VA) but the idea is to move to something around 3000VA. The deatime at the PIC is 0.4us. I'm getting additional deatime due to the driver not being so fast when turning up. I'll review each part to make sure I make it faster. Thank you! \$\endgroup\$
    – Ron Groove
    Commented Jul 1, 2016 at 16:31
  • \$\begingroup\$ @ Ron Grove .You have a good scope .Keep your good scope .I think that your sinewave looks quite good .Your big 50Hz transformer could be helping filter out the switching frequency .You could experiment with an X cap of say i microfarad across the mains output.This could make the waveform better. \$\endgroup\$
    – Autistic
    Commented Jul 13, 2017 at 12:41

2 Answers 2

Reset to default
1
\$\begingroup\$

Try putting a (non-inductive) sense resistor in series with the inductive load and measuring the voltage across it. The voltage across this resistor will then be in phase with the current, which will be phase-shifted from your voltage with the inductive load. You may see that the distortion occurs at the point where the (lagging) current is zero; if so you could be looking at the wrong place for your solution. If this is the case, your voltage's PWM artifacts can't be fixed with dead time.

\$\endgroup\$
0
\$\begingroup\$

There really isn't any useful information to say whether you can or you cannot (eg present deadtime, gatedrive capability.)

Looking at the datasheet for the FET's in question:

Ton delay ~ 13ns Rise ~ 64ns toff delay ~ 49ns fall ~ 48ns

These are measured under a very specific switching setup (fig 10a) and this will not be the same for your setup. You will have stray inductance and your stray could be quite bad.

My advice would be to measure the switching losses by performing a DOUBLE-PULSE test in-situe.

by using a reasonable precise probe across the FET (and its Rds_on will serve as the current measurement) you can switching current into an inductive load and thus capture the true rise,fall and delay for your specific setup.

Based upon these results you can choose to change the deadtime

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.