SMPS(flyback question)

I would normally use a Schottky diode every time rather than waste time on a so-called ultra-fast type. I have inherited SMPS units with issues, and surprisingly often a change to Schottky diode(s) changes the whole unit back to proper operation, and with a dramatic reduction in diode temperature.

Thanks , I'll have a look around for a suitable replacement. Cheers Rob

The hissing could be a side effect. When I was working with a uc3842-based flyback supply in 1988, I heard some instability that most people didn't that I quickly traced back to feedback circuit that drove the optoisolator. Check to make sure that your feedback is stable at those higher loads.

1) It's not necessarily a bad thing for a rectifier to run hot, since efficiency increases with the consequent decrease in forward voltage. 2) At higher output power, peak current obviously rises, and this has many effects that lead to instability, especially: increased energy stored in leakage inductance --> larger voltage spikes --> instability due to noise 3) Duty factor also increases: research on keywords 'sub-harmonic oscillation' and 'ramp compensation'

Changing the type of rectifier diode will sometimes help out with #2 above.

Paul Mathews

snipped-for-privacy@yahoo.com wrote: (snip)

Just a guess... At higher current, the supply is more likely to hit continuous current mode (inductor does not fully discharge before end of cycle). This means that the power switch is turning on, while the output diode is still conducting. The reverse recovery diode current gets added to the normal switch current (normally, mostly capacitive current from winding capacitance) during turn on, and this normally gets higher when the diode gets hotter. At some point, the switch may be hitting a current limit and skipping a cycle (causing the audible noise).

Hi to all. Thanks for the pointers. Would I be able to see if the supply is running in continuous mode by monitoring the voltage across the current sense resistor. I assume I will , but I am not sure what to look for exactly. Fiddeling around in the past I have noticed 2 types of voltage across the sense resistor. One that rises from 0 in a linear ramp until turn off of the transistor , and one that rises sharply to some value , and then ramps linearly until transistor turn off. Is this examples of the circuit running in discontinous and continous modes. As I mentioned I am still feeling my way in this stuff. Thanks for the help. Cheers Rob

To see if you're continuous or not, I'd suggest looking at the secondary-side (output) diode. When the diode is negatively biased, the transformer is taking in energy on the primary side. When the diode is forward biased, the transformer is dumping its stored energy into the output caps. When the diode drops from its Vf (and may oscillate slightly negative to Vf) but doesn't kick back over to extremely reverse biased, the transformer is "empty" and you're in discontinuous mode.

Good guess, I guess. With the additional effect that a flyback going continuous conduction gives you an annoying RHP zero that won't do any good to the feedback loop stability if not accounted for.

To Rob, check your primary side switch current and see whether the current ramp starts at zero (discontinuous mode) or if the current jumps to a non null value on MOS switch on (continuous mode). Alternatively you can look at the voltages waveforms. In CC mode you have only 2 voltages levels, corresponding to the ON/OFF states of the switch. In DC mode you'll see 3 levels, the 2 of CC mode plus one idle voltage when all the core stored energy has been transfered to the load side. During this phase you'll see some damped voltage ringing (the transformer magnetizing inductance is resonating with all the parasitics capacitance it sees).

If John guessed right, you should have a zero pulse start below your observed limit, and a rising non zero pulse start above that same limit.

CC is often not desired because of the higher switching losses, and because the RHPZ gives you a slower transient response. In that case the cure could be to redesign the transformer.

OTOH DC puts more stress on semiconductors (peak currents are higher).

--
Thanks,
Fred.

Yes. Sense resistor waveform will be trapezoid (plus a leading edge spike) for continuous mode versus triangle (plus spike) for discontinuous. Subharmonic oscillation becomes more likely as you increase duty factor to near 50% and beyond. However, you could easily have multiple interacting causes of instability if your layout and construction have excessive parasitics. Paul Mathews

The easiest place to see it is by looking at the secondary voltage feeding the rectifier. If the current is discontinuous, there will be a small period of ringing before each charge up pulse. When the current becomes continuous, the voltage will jerk from forward biasing the rectifier right to the inverse charge up voltage.

Yes. Continuous current mode means the current never ramps all the way to zero between charge ups. Firing power into a conducting rectifier is more likely to produce a brief spike on the current sense resistor that trips the current limit, even though the current after the spike is well below the limit. Sometimes adding a small RC time constant (shorter than the normal ramp time) between the current sense resistor and the current sense circuit, will smear this spike well enough to let the supply run in continuous current mode. However, both the rectifier and power switch will be running hotter because of the spike.

Hi all. Thanks to all who have replied to my questions. The voltage I am measuring across the sense resistor is trapazoidal at all but very low power levels.Does this mean I'm running in continous mode most of the time? I'm not sure I understand the posts about taking the voltage from the input of the diode. At the input of the diode I see a (almost perfect) square wave , 25% duty cycle going from -50 to 14V. There is a little ringing at the top of the rising edge and the bottom of the falling edge , but that is all. The primary inductance is about 600uh , sec inductance around 18uH and the switching frequency is about 110Khz.The current sense resistor is

0.375 ohms. The feedback to the current sence input to the 3842 is via a 1k with 470pF to gnd. Thanks to all who are helping me here. Cheers Rob

yep

thats cos its ccm

whats the load & Vout?

cheers terry

Hi all again. Doing a quick simulation shows it definately seems to be running in CC mode. There is still significant current flowing in the secondary winding then the next cycle begins. What ways are ther to solve this. The output voltage is 13.7V Load is 4.6A ( 12V halogen bulb)

situation. Still CC mode but the current has ramped much closer to 0 before switching begins again. I currently have 1000uF -> 10uH - >

1000uF output config. Reducing the caps to 220u improves matters.I'll do some more experimenting here. I could reduce the swithcing frequency I suppose , give more time for the current to decay to 0 before the next cycle , but I'd rather not. If I have to redo the transformer , what do I change?

More turns on the primary (or less on the secondary), I would say.

Tim

-- Deep Fryer: a very philosophical monk. Website:

Hi there. Reading a bit it seems that CC mode is not a bad mode to be running in. It may be better to change my compensation to account for this.The datasheet for the uc3842 shows compensation for for a flyback in cc mode.It does not show values unfortunately. Not realy sure how to calculate them. It is basically just another resistor and cap to create another pole ,but at what frequency. This is where my ignorance on the subject becomes sorely apparent!! Bit more experimenting required. The circuit is basically running quite well. I can deliver up to 150W( have not pushed it any further) with a fan heeping things cool.The switching tranny and diode are only on small heatsinks (20 *30 mm with 2 forward facing sections of about 10*30mm) As I said , bit more experimenting required :0>

Cheers Rob

I agree. It lowers the RMS winding current relative to the DC output current and in some ways, lowers the noise the supply emits. It also has some negatives, like slower control response to output error and rougher turn on transient.

I am a bit rusty on the compensation, but the gist of the problem is this:

When in discontinuous mode, the supply can respond to a change in error in the very next power cycle, because each one is completely independent of the last. All the energy stored in the inductor gets dumped to the output, each cycle. If more power is needed for the next cycle, the on time is increased, the energy goes up, and that larger energy comes out of the rectifier by the end of the cycle.

In CCM, there is energy left in the core when the next power pulse interrupts the output diode current. So the consecutive cycles are not independent of each other. In order to raise the output current (in response to error) the input duty cycle rises. But the effect of this on the first cycle after the increase is both an increase in the diode current (during the discharge part of the cycle) and also a decrease in the conduction time available for that current. For the first few cycles, this results in a net decrease in average DC current out. The longer the drive time is boosted, the more severe the output current drop. Eventually (many cycles later) the stored energy gets so high that the shorter duration output pulses finally get strong enough that the average current increases enough to start bringing the error down. At that point, the charge up time starts to back off. The initial reaction to this is that the very high output current lasts longer (all that is left of) each cycle. So the output current rises as the input duty cycle falls, till the stored energy in the transformer gets dumped down.

This reverse initial reaction is the difficulty called a response zero that complicates the control. It requires a loop compensation that fixes any problem more slowly so that the reverse acting initial response is swept under the rug. Sometimes, having more output capacitance helps get by with this slower control.

This same effect takes place when you steer a bicycle. In order to turn left, you have to first, briefly steer right to move the line of tire contact with the pavement to your right, causing the bike to be leaning left. Then you can turn. This limits how fast you can begin a turn.

Also, when changing the horsepower out of an internal combustion engine, if you slap the gas pedal down when the engine is idling, you risk stalling the engine with the power consuming compression stroke that must precede the increased power producing power stroke. You hide this problem by accelerating more slowly, or add a bigger fly wheel to the crank (which also slows acceleration but hides this problem).

Hi there. Thanks for the explanation of what is happening in cc mode. Really helps a lot. Still lots of studying to be done on my part though :0( I assume I have to add a R & C to slow response down a bit. My maths is not so hot anymore ( never was I suppose) , so a bit of trial and error experimentation is in order. Maybe if I can find some app notes with this kind of cct on the net that should point me in the right direction. Cheers for now. Rob

(snip)

I haven't studied this, but a quick scan looks pretty promising.

It took me a couple tries before it came up without timing out.

snipped-for-privacy@yahoo.com wrote: (snip)

A similar one: