UC3854 PFC problems

[Spirit]

I am working on UC3854 PFC. Input is 80-250V, 50Hz, output is 385V dc.
Everything is made like on 250W pfc preregulator on Unitrode UC3854
datasheet, except inductor and power supply for UC3854 (for testing I use
external stabilised 18V dc).

For inductor I use ETD35G ferrite core with 250 turns of wire. Core
specifications are on this site:

http://www.iskra-feriti.si/index.php?t=mat-list

Problem is that pfc preregulator works ok without a load and I have 385V on
output, but when i put any load, even very small load like 20k resistor, I
get very large input current, like 2A and more, and duty cycle about 90%,
and very small output voltage, just few volts.

I checked, and rechecked all components and connections many times but
everything is like on Unitrode schematic...

Now I think that problem is in inductor.

Does anybady have experience with UC3854 inductors?

I am very interesting in if that inductor is very critical or pfc can work
with several inductors with more or less worse performance?

 

[Graham Holloway]

Boost diode back to front?

DNA

Shouldn't there be a lot of smoke?

Graham Holloway

 

[Spirit]

Boost diode back to front?

Boost diode is ok polarised.

I think inductor is the problem.

What kind of ferite corres are suggested for this type of application?
Torus, EE,...?

 

[CWatters]

Boost diode back to front?

Faulty/wrong capacitor?

Not Low ESR type?

 

[Phil Allison]

"Spirit"

For inductor I use ETD35G ferrite core with 250 turns of wire. Core
specifications are on this site:

** Did you create an air gap ?



........ Phil

 

[Spirit]

"Spirit"



** Did you create an air gap ?

Yes I tryed with varyous and without air gap, but always the same :-(

For protection I use 10 ohm resistor in serie with input AC voltage.

When I connect UC3854 to 18V, and disconnect input voltage of power part,
duty cycle is about 90%

Then when I connect 60V to input, without a load, duty cycle is about 70%,
and I get 380V at output, that is ok.

When I connect load ( 5.6k ohm resistor) on output, duty cycle is about 90%,
output voltage drops to cca 100V and mosfet comes very hot.

In the moment when I disconnect load, duty cycle is about 5% for the moment,
and then goes to 80% without load.

I tyed with one coil, with another coil, and with 2 coils in serie, and
always the same thing.

I really don't know where can be the problem...:-(

 

[Stefaan Vanheesbeke]

Then when I connect 60V to input, without a load, duty cycle is about 70%,
and I get 380V at output, that is ok.

Without any load, and a DC input voltage, after the start up transient the
output voltage should be correct, and the duty cycle zero. Because of some
losses (resistor divider for measuring the output voltage, ...), the duty
cycle will be very low.

Did you made any calculations on the value of the coil?? saturation
current?? Duty cycles that should be normal?? Continiuous/discontinous boost
operation.


PFC's are generally not the most easy things to design, if you want to go on
with the project, some guidelines (for these, an oscilloscope is needed.
Without oscilloscope, I give you very little chance to make it work
correctly)


- start from scratch with verifying basic UC3854 operation on your board :
1. oscillator running correct frequency

2. try a circuit that you can generate variable duty cycles at
the output with a variable resistor --> you can use the current error
amplifier with a simple feedback resistor in a basic opamp amplifier. This
sets the level where the saw tooth from the oscillator is compared, and
generates a certain duty cycle at the output.

- when you can generate veriable duty cycles, you can start playing around
with the power stage. Use only DC input at the moment.

1. apply small input voltage (from a current limited power supply),
DO NOT FORGET TO APPLY A LOAD RESISTOR TO THE OUTPUT!! Calculate the output
voltage with a certain duty cycle (use also this duty cycle to start
testing), the output voltage will be equal or less than the calculated
voltage (it will be less if the boost converter works in discontinous mode).

2. Monitor the solenoid current!! Put a small ressitor in series
with the soleoid (for example 1 Ohm for first low power tests, later on 0.1
Ohm or less). The resistor must have LOW PARASITIC IMPEDANCE otherwise, the
voltage measured on it will not be equal to the actual current in the coil
(it will show steps).

3. play around a bit with the duty cycle, load resistor and input
voltage.

4. evaluate the inducatnce and saturation of the inductor used. The
inductance can be calculated measuring the slope of the current and the
voltage across the inductor. Saturation is the point where the current
starts to rise faster than linear. The soleoid should always be used below
saturation!

5. At the point where you reached the full input voltage and load
conditions ( = output voltage set correctly with the potentiometer, and full
load applied), evaluate output ripple, thermal behaviour of the MOSFET,
solenoid, ... Also some early measurements about efficiency can be done.


When the tests above are finished with success, a first milestone is
reached. Remember you are still working in a DC environment, not with AC
input, so the final result will be different (according to efficiency,
heating, output ripple ...)

- now it is time for the control loop. This is not a piece of cake. The
control loop settings depends highly on the inductor value and output
capacitance. This chip uses current mode control, wich is a big advantage
for boost converters. Some interesting documentation is written by THE man
of switching power supply control loops : Lloyd Dixon (the unitrode days).
http://focus.ti.com/lit/an/slua079/slua079.pdf

There should be a document available for this control loop type used in PFC
devices, but I can not find it right away.


Have fun.

 

[Mark]

I agree you need a scope and a current probe to even begin thinking
about troubleshooting these.

Also lok at the layout re grounding,,, a small amount of ground
coupling from the power circuits back into the control circuits can
casue unpredicable behavior. Use single point grounding for the power
and another single point ground for the low level stuff.

Mark

 

[Vanheesbeke Stefaan]

I agree you need a scope and a current probe to even begin thinking
about troubleshooting these.

Also lok at the layout re grounding,,, a small amount of ground
coupling from the power circuits back into the control circuits can
casue unpredicable behavior. Use single point grounding for the power
and another single point ground for the low level stuff.

Mark

Indeed, grounding can be very imortand. Also when doing measurements, don't
be supprised to see a lot of switching noise.
I'm always wondering how they generate the nice and clean waveforms in
datasheets or application notes, ...