Need help on this inverter

Discussion in 'The Projects Forum' started by fabrizziop, Sep 16, 2011.

  1. fabrizziop

    Thread Starter New Member

    Sep 16, 2011
    6
    0
    Hi. I'm new to this forum, so, first, hello to all of you. I'm kinda new to electronics, I've learned all online, and I've made some simple projects with spare parts like a bench power supply with an old dot matrix printer transformer and things like that.

    I tried to make a 12VDC to 120VAC inverter like a month ago, with two NPN transistors , but I couldn't find the center-tapped transformer, as in my country it's very hard to find electronic stuff.

    Now, I tried to make a better inverter with a full H bridge, so I could use all kinds of transformers, not just center-tapped transformers. I assembled it in a breadboard, and used an wall wart AC to AC adapter(just a transformer inside a nice plastic case) backwards, using the secondary 9V as primary, and it works fine, and I powered a radio and several small appliances.

    My problem is that, When the supply voltage for the circuit is less than 12V, the P-Channel MOSFETS are cold to touch, but when voltage is higher than 12V they become warm even without a load. I got a borrowed oscilloscope, and I see that it's like that the P-Channel MOSFETS take a little while to turn off. I think you can help me.

    The schematics are in the attachment. Thanks in advance.
     
  2. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    You are using very large resistors (5.1k) on the MOSFET gates, and then another 5k in series to turn them off. This causes an RC time delay due to the series resistance and the capacitance of the gate.

    However, the LM339 comparators cannot sink much current, and they cannot source it at all. You need driver circuits between the LM339 and the MOSFET gates.

    Also, you need to change the threshold levels on the comparators. You need to have "dead time" between when one set of MOSFETs are turning off, and the other set are turning on. Without 'dead time', you will wind up with the dreaded "shoot-through" where both the high and low side MOSFETs are on at the same time, causing a short circuit across the supply. This will result in burned-up MOSFETs.
     
  3. fabrizziop

    Thread Starter New Member

    Sep 16, 2011
    6
    0
    Well, look, each of the comparator outputs drives a high side and a low side MOSFET. I think the shoot-through happens when the comparator starts sinking current. What alternative do I have to a MOSFET driver?.

    Would a NPN transistor connected to an inverted output help to sink the gate current. I would do this to each of the both gate channels. This would allow me to put just an 1K pull-up resistor.

    What is the minimum value of the MOSFET gate resistors?. As I currently have 500 ohm resistors in the gates(unlike the schematic).

    Sorry for my bad english.
     
  4. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Your English is better than many who live in my country. You have nothing to apologize for there. ;)

    I will have to sort things out in a SPICE simulator, as your schematic is difficult to read as drawn.
     
  5. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK, I've re-drawn your schematic to be more standard in appearance. I drew the transformer somewhat unusally, but it's easy to see that the MOSFETs are an H-bridge.

    A problem is that you are using the pin 3 output. There is no way to adjust the duty cycle if you are simply looking at pin 3. You should be looking at the voltage on C1 instead, which resembles a triangle wave.

    Anyway, have a look. Right now, it's basically your same schematic, just re-drawn. The 2903 is basically the same as a 393 (except higher temp range, lower operating voltage), which is half of a 339 comparator.
     
  6. fabrizziop

    Thread Starter New Member

    Sep 16, 2011
    6
    0
    Well, thank you very much, but I solved the problem with your idea of a MOSFET driver. I made a totem-pole driver with a NEC D882 and a 2n2907, I made two, one for each side of the H bridge. Now they're cold and the spike on the supply voltage of the thing lowered(a lot less shoot-through). I did this: http://forum.allaboutcircuits.com/showthread.php?p=177826

    It works a lot better now. Thank you for your help. I used an IRF9540. It was cheaper and the only available on the site where I buy components.

    Is there any other thing I can improve on this?. I just wanna know if this thing is going to blow up in for some reason when I need it.
     
  7. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    It is probably too late for your application, but you can use two transformers to make an equivalent transformer w/ center tap. Each transformer would need to be ½ the rated voltage you originally needed, and you would need to be very careful about phasing.
     
  8. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK, good - glad you made the gate drivers.

    Have a look at the attached; it's changed only slightly from the previous version. The change is rather subtle; instead of using the pin 3 output from the 555 timer, I'm using the triangle wave from the top of C1, and I've changed the way the thresholds are set.

    The critical part here is that the DC component of the transformer primary needs to average 0. You can make R10 and/or R11 variable along with VR1 to find the initial values necessary. You can start by setting points Hi and Lo to initially match the high and low points of the triangle wave; if they don't, then the output won't be symmetrical, and you'll wind up with a rather high DC component in the transformer primary.

    This leads to a phenomenon known informally as "flux walking"; the DC component along with the AC input winds up causing the transformer primary to become saturated, the voltage across the primary rapidly drops upon saturation, and the MOSFETs will suddenly burn up due to the resulting high current. This can be difficult to diagnose if you are not familiar with the phenomenon.

    VR1 can then be used to adjust the effective output voltage. Instead of a square wave output, you will wind up with a stepped square wave, frequently called a "pseudo sine wave" output. It's really no where near a sine wave, but it's better than a plain square wave output, and also helps a great deal to reduce the likelihood of "shoot-through". Adding a small amount of fixed resistance in series with VR1 should prevent shoot-through altogether.
     
  9. fabrizziop

    Thread Starter New Member

    Sep 16, 2011
    6
    0
    Thanks for the idea, but I'm going to use a 120VAC to 12VAC transformer(10:1). When it's sqaure wave the RMS voltage is the same as peak one, but on stepped square wave, I'll need about 17VDC to have 12VAC RMS, and I plan to power it with a 12V battery.
     
  10. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK then, you'll have no means to adjust your output voltage. If you went to something like a 9v to 120v, then you could use the stepped square wave output. But, it's up to you.
     
  11. fabrizziop

    Thread Starter New Member

    Sep 16, 2011
    6
    0
    Yep, as I can't find anywhere in my country a power 9V to 120V transformer, only 12V to 120V as people use them to power 12VAC halogen bulbs.
     
  12. fabrizziop

    Thread Starter New Member

    Sep 16, 2011
    6
    0
    Update: Found a cause of bad performance: Counterfeit IRF9540 MOSFETs. They look like these: http://endless-sphere.com/forums/viewtopic.php?f=2&t=10655

    where it says counterfeit 4110, the one I bought at a local supplier(in physical store) are very alike(logo and numbers in wrong place, etc). I will go there ASAP and confront them.

    Be very careful, fake pieces aren't just on eBay, they're also near YOU!
     
Loading...