SG3525 problem

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have built a circuit for a car SMPS that is regulated.

The problem is it is either full duty cycle (- dead time) or no duty cycle.:bawling: None of this has been hooked up to a full psu, it's just the controler board.

I have a perfect clock at 100KHz, outputs are square and non-overlapping but when I put the vref of the chip to the feedback pot and start turning and hits 2.5V the duty cycle goes to zero. :smash:

Why?.:confused: I don't understand why it's doing that, I've double and tripple checked my circuit and it seems all ok.

Any Ideas?

Thanks

Dave
 

Attachments

  • sg3525.gif
    sg3525.gif
    10.8 KB · Views: 1,108
Hi

U use 0.1uF after pot, since this is FB I would remove it. He can hold the voltage high long enought for you to think it is not working

Also remove 0.1uF that is a bit lower down, you dont need it
Pin9 should be connected through 0.1uF to gnd not pin2
And probably you have elements on left connected to gnd(timing resistor, dead time resistor,...)?, right
 
Fry,

Without a load to monitor and feedback to your PWM section, going from 0-99% duty cycle is expected, if your test feedback voltage is ~2.5V. This is because the feedback is coming from outside the loop, and not monitoring what the chip is putting out. For a simplified way to test this, connect a small inductor, schottky diode and capacitor in buck fashion to one of the outputs, then take the voltage reading off of that, and adjust accordingly. You should now see duty-cycles varying from 0-99% more gradually, and not so suddenly.

Also, as luka said, change your pin 9 (compensation) to ground, thru a 1000pF - 0.1mF cap.

I know what I'm trying to say, but I get a feeling that I migh not be getting it across clearly. If not, I will try to clarify things a bit. :xeye:

Steve
 
Fry,

Without a load to monitor and feedback to your PWM section, snapping from 0-99% duty cycle is expected, if your test feedback voltage is ~2.5V. This is because the feedback is coming from outside the loop, and not monitoring what the chip is putting out. For a simplified way to test this, connect a small inductor, schottky diode and capacitor in buck fashion to one of the outputs, then take the voltage reading off of that, and adjust accordingly. You should now see duty-cycles varying from 0-99% more gradually, and not so suddenly.

Also, as luka said, change your pin 9 (compensation) to ground, thru a 1000pF - 0.1mF cap.

I know what I'm trying to say, but I get a feeling that I migh not be getting it across clearly. If not, I will try to clarify things a bit. :xeye:

Steve
 
I'm not sure as far as the 3525 goes, but I know that with others like the 3524, and the TL494, you can connect your varying PWM control voltage directly to the compensation pin.

That pin is equivalent to the output of the error amplifier and drives the PWM circuit directly.

This will allow a continual variance of pulse width with a varying input voltage between ground and vref.

If you don't like doing it that way, connect a resistor divider between the compensation pin and the inverting input pin. This will set the internal error amp to have a specific DC gain, just like an op-amp.

If, for example, you put a direct short between the compensation pin and the inverting input, the gain will be set to 1. You could then control pulse width directly with the non-inverting input.

I have done both of these methods with complete success.
 
luka said:
Hi

U use 0.1uF after pot, since this is FB I would remove it. He can hold the voltage high long enought for you to think it is not working

Also remove 0.1uF that is a bit lower down, you dont need it
Pin9 should be connected through 0.1uF to gnd not pin2
And probably you have elements on left connected to gnd(timing resistor, dead time resistor,...)?, right

O.K. I will remove caps on divider and FB, and change comp cap to ground.

Duo said:

That If, for example, you put a direct short between the compensation pin and the inverting input, the gain will be set to 1. You could then control pulse width directly with the non-inverting input.

I have done both of these methods with complete success.

Woundn't you want gain though in a normal work PS or are you stating the simple test procedure?
 
Duo said:
The gain is achieved by having the output sensed by one side of the error amp and the other connected to the vref through a resistor divider. (Gain is very high and regulation is tight.)

For testing, you don't need any gain besides 1.

Ok so I'll remove those caps and tie the comp to the inverting input pin and post my results later tonight.

IS the gain almost infinite when no feedback is applied to the error amp?
 
So I did the modifications that were suggested and I can now vary the duty cycle though it is very very sensitive (this is with out tying comp to the inverting input pin ) but it does go from 0 to 100% (-DT) so it works.

One thing that gets me is if I have it right at the threshold and I blow on the chip to cool it the duty cyle can change from full to nothing. Is this right? is it because theres no feedback for the error-amp?

Also is it more stable with feedback for the error-amp or not?

Thanks Dave

P.S updated schematic with fixes applied


Edit: fixed missing ground in schematic (thanks luka ;) )
 

Attachments

  • sg3525.gif
    sg3525.gif
    6.7 KB · Views: 501
luka said:
Hi

Does it heat up?!?! Let me see... :xeye: ohh yes I didn't see that you dont have any transistor base resistors. Put 47-100 ohm ones directly between SG output and B of transistors. Heat will then disappear. Everything else looks good to me.


Sorry, I must of explained it wrong, nothing heats up when it's running, but if I heat it up or cool it down it changes duty cycle when set close to the threshold.
 
Just a single resistor per transistor pair should be fine.

If you're running your 3525 closed-loop (regulating), then you aren't going to be technically setting the gain of the error amplifier per-se.

(The gain is being set, but indirectly.)


You should be able to connect vref directly as a reference to the error amplifier and then use feedback from your DC output voltage to allow regulation. (You must use a resistor divider from output to ground to set the feedback voltage equal to vref for the desired output voltage.)

A potentiometer from vref to ground with its wiper to the error amplifier would allow for adjustable operation. (You can pad this pot with resistors to set limits on its range; a good idea).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.