Current runaway in power inverter circuit.

[Don Gilmore]

Hi guys:

I'm an ME doing a little EE work. I'm testing a circuit that is basically a power inverter.

I start with 24Vdc and a transformer primary (ferrite core) with a center tap. The center tap is connected to +24V. Each end of the primary is connected to a n-type power MOSFET (with a very low Rds) each of which connects to ground. I fire the MOSFET's alternately using a function generator and an inverter (Schmitt trigger). I have also added RC snubbers to each end to filter out ringing and spikes.

Because of the nature of the application, the transformer is very small with relatively few turns of heavier gauge wire, so I drive it at a fairly high frequency. I'm trying to maximize my output for a given current input from the power supply. My secondary drives a higher-current, low-resistance load.

I start at high frequency and lower the frequency until I get the current I need. The circuit works great, and I get more and more output current as I lower the frequency. But there is a certain point where the input current begins to run away, even though the current isn't that high. I have been experimenting with different gauges and turns of wire, but here is a typical result:

f = 60 kHz
Power supply output = .14 A
Iout = .87 A

....but by the time I get down to f = 25 kHz:

Power supply output = .17 A
Iout = 1.37 A

Then, after a few seconds, Ips is .21 A, then .28 A, then .35 A, etc. until I have to shut it down before it burns something up. The output current doesn't change. This occurs at some frequency no matter what transformer winding I use and it occurs suddenly and when the output really isn't all that excessive.

Can someone tell me what is causing this and is there a remedy? Thanks for any replies.

Don

 

[Don Gilmore]

Thanks, Robert.

I had been thinking some about that. Actually it fits well with my embedded application. I planned to artificially produce the high-frequency alternating signal by connecting the two transistors to separate outputs of a microcontroller chip. It should be easy to generate a pause between "on" states of the MOSFET's in my program.

In the meantime, I'll try a 556 dual timer. Thanks again for your help!

Don

 

[Don Gilmore]

I wonder how it would work if driven with a half-wave-rectified sine wave. Too much delay?

Don

 

[RobertMacy]

On Thu, 30 Jan 2014 07:33:08 -0700, Don Gilmore <eromlignod@outlook.com>
wrote:

...snip...
Then, after a few seconds, Ips is .21 A, then .28 A, then .35 A, etc.
until I have to shut it down before it burns something up. The output
current doesn't change. This occurs at some frequency no matter what
transformer winding I use and it occurs suddenly and when the output
really isn't all that excessive.

Can someone tell me what is causing this and is there a remedy? Thanks
for any replies.

Don

Your transformer is 'walking up its hysteresis curve. it's unbalanced
although YOU think its balanced.


Two solutions:
1. FORCE a dead time where BOTH switches are OFF, this allows the
transformer core to recover.
I used to use simple AND circuitry and ONE SHOT timer. By combining the
SAME signal delay to both drives you also gain balance. And adjusting one
timer adjusts both off times.

Technique can be done this way drive A is delayed by one shot, drive B is
delayed by one shot. Adjust the one shot so your off time is approximately
1% of duty cycle, should be ok.

2. Too complicated at this time, later.

 

[RobertMacy]

On Thu, 30 Jan 2014 09:00:23 -0700, Don Gilmore <eromlignod@outlook.com>
wrote:

Thanks, Robert.

I had been thinking some about that. Actually it fits well with my
embedded application. I planned to artificially produce the
high-frequency alternating signal by connecting the two transistors to
separate outputs of a microcontroller chip. It should be easy to
generate a pause between "on" states of the MOSFET's in my program.

In the meantime, I'll try a 556 dual timer. Thanks again for your help!

Don

there is a way to calculate the EXACT amount of time to make certain that
current saturation NEVER happens.

But the fact that you got away with it for a short bit of time, instead of
a few milliseconds, says you won't need much.

Assuming you're running in the 50kHz to 100kHz range, you may only need
1uS. At that small a dead zone, you could use a two input XOR chip with an
RC input to one leg, that way you get a short spike each transition and
should be the same. save complexity to your circuit and a bit of power.

Don't know if it's still true, but use CMOS timers instead of TTL timers.
The old mask for 555 chip [extremely popular chip] used to have a flaw in
that at transition BOTH drives were ON. Meaning pull up and pull down both
trying at the same time. That translated to a 0.5nS spike of sometimes 2
amps! of current. Well, you can imagine what 2 amps into the GHz range can
do to your supply structure AND let alone becomes a radio that launches
energy that can be heard in the next state.

so use newer versions, the new masks are done by engineers who value
'quiet' circuitry.

 

[RobertMacy]

On Thu, 30 Jan 2014 10:01:42 -0700, Don Gilmore <eromlignod@outlook.com>
wrote:

I wonder how it would work if driven with a half-wave-rectified sine
wave. Too much delay?

Don

You're actually starting to discuss ways to 'regulate' the output now.

For example, that forced, tiny dead zone all the time. Now drive from no
switching to full switching and you can see that you will get out a
voltage from zero up to max. thus, regulation.

Actually there's no such thing as too much OFF. Well, until NOTHING goes
through, then THAT's too much off, ...by definition.

One thing to watch out for. Sometimes ringing can occur when you shut off
both switches. That ringing can take the signals into areas you did NOT
expect. You may need to put clamping diodes across your transformer to
keep everything well behaved.

You'll see that when one side goes down to ground while eing switched the
other side only goes up to around twice your supply voltage. But when you
switch off ?? both sides will finally equal thepower supply, but getting
there they may travel more minus than ground, or more positive than twice
supply, either which can damage the switches. Most MOSFETs body diodes
will clampon the negative going spike, but sometimes not a good idea to
use thatbody diode. It's my understanding it can be a bit slow, therefore
recovering can take a bit of time, which means it's still conducting like
a short to ground when you turn on the other side, ouch. That will make a
spike of current come out of your bypass cap on the power supply.

I assmue you have LTspice, or some simulation tool that can show you this
stuff, right? There's nothing like the real world as the best analog
computer, but sometimes it's difficult to change parts and values quickly
and misconnections can be catastrophic.

 

[Phil Allison]

"Don Gilmore"


I start with 24Vdc and a transformer primary (ferrite core)

** What style of core??

I start at high frequency and lower the frequency until I get the current I
need.

** That the output falls with rising frequency suggests very poor winding
technique.


The circuit works great, and I get more and more output current as I lower
the frequency.

** Oh dear ........


But there is a certain point where the input current begins to run away,


** Oh dear 2.....


Can someone tell me what is causing this and is there a remedy?

** The core is going into saturation = massive loss of primary inductance =
high current and then BANG !!

As others have indicated, a "dead band" is essential and a very precise
matching of the two drive wave "on" times

FYI:

When you build a real one, the start up process is gonna be a treat.



..... Phil