UC2856 Current mode PWM controller IC

[theoldwizard1]

I have been researching building my own inverter (some you might have seen some of my other posts).  The most common design uses a DC-DC SWPS circuit, typically with a ferrite transformer to boost the low voltage DC up to high voltgae DC.  After that, the high voltage DC is used to power a power a Class D amplifier which is being feed a 50/60 Hz sine wave.

There are several ICs on the market that do a good job of PWM control for DC-DC conversion.  The UC2856 has dual feedback error amplifiers and something called "pulse by pulse current limiting".  This part of the IC is in the bottom left corner.

Any ever use this IC and can you explain how the "pulse by pulse current limiting" works ?

[T3sl4co1l]

Looks like a somewhat more featured variation of the UCC3808... which is basically a UC3844 with the alternate toggle output taken out to a pin.

Like them, you need to use a high ripple (discontinuous or nearly so) design, otherwise chaotic behavior results.  Slope compensation helps, but you can only do so much.

The advantage for a push-pull device over single-ended is, you can use a different transformer, potentially saving on space (it's a transformer driven full wave, not a coupled inductor driven from a single switch).  The disadvantage is, you need two (or more) of several switching components (transistors, diodes), and you still need the energy storage component (the inductor, which has to handle high ripple), so it tends to be bulkier regardless.

Tim

[theoldwizard1]

Looks like a somewhat more featured variation of the UCC3808... which is basically a UC3844 with the alternate toggle output taken out to a pin.

Like them, you need to use a high ripple (discontinuous or nearly so) design, otherwise chaotic behavior results.  Slope compensation helps, but you can only do so much.

Correct !  Dead time is important so that the magnetic field in the transformer can collapse releasing its that energy.

The advantage for a push-pull device over single-ended is, you can use a different transformer, potentially saving on space (it's a transformer driven full wave, not a coupled inductor driven from a single switch).  The disadvantage is, you need two (or more) of several switching components (transistors, diodes), and you still need the energy storage component (the inductor, which has to handle high ripple), so it tends to be bulkier regardless.

Correct again.

The good news is that high power, low _RdsON MOSFET are getting cheap.  High speed diodes (Schottky) are used in a full wave bridge rectifier.  And yes, these are fed to a classic LC filter, but the L does not need to be overly large.

[oldway]

"pulse by pulse current limiting" is a great reliability improvement on average current limiting (TL494 for example).
As far as there is a little  inductance in serie with the Mosfet for limiting di/dt, the pulse by pulse current limiting is able to prevent catastrofic failure in case of something going wrong.

Better use push pull configuration only for low voltage converters.
Be carefull with core saturation risks of the ferrite transformer !
Primary windings must be bifilar.
 

[theoldwizard1]

"pulse by pulse current limiting" is a great reliability improvement on average current limiting (TL494 for example) ...

Great info so far, but it does not answer my question.  How does this part of the IC "work" ?

[theoldwizard1]

"pulse by pulse current limiting" is a great reliability improvement on average current limiting (TL494 for example) ...

Great info so far, but it does not answer my question.  How does this part of the IC "work" ?  (The EA op amp is for voltage beedback.)

[oldway]

Great info so far, but it does not answer my question.  How does this part of the IC "work" ?  (The EA op amp is for voltage beedback.)

Not so much to explain.
When Is exceed the set value (by R1/R2), the pulse stops and the Mosfet is blocked before current reach a too high value.
The current limit transistor is limiting output voltage of EA at a value depending of R1/R2.
Limited Output voltage of EA is then compared with amplified Isense.
When amplified Isense exceed this voltage, output of comp switch and stop the pulse.

[T3sl4co1l]

The voltage on the COMP pin is proportional to the amount of peak current the switch will be turned off at.

If the pin is weak (probably the case), you could even overdrive it from an external op-amp if you wanted.

This is essentially what they're doing internally, except it looks like it's got a passive pull-up (the current source), so everything can be wired-AND (the PNP emitter follower is only active in the pulling-down direction, and likely, the same is true of the internal error amplifier).

The error amplifier is the only linear op-amp in the IC; the current sense is amplified, but not operationally (you don't get to play with the +in, -in, out pins).

Alternately, you can think of it as, the COMP voltage is divided by 3 and compared to the differential current sense voltage.  Same thing, just semantics (though quite different to implement).  The UC3842 series shows this: the COMP voltage range is 0-3V while the current sense input is 0-1V (ground referenced, not differential).

If you wanted to operate this device in constant-peak-current mode, you can simply strap the error amplifier for unity gain (resistor feedback only, no compensation capacitor) and connect it to a suitable reference voltage (thus setting COMP to a constant voltage).  This would be useful for simple applications, like LED lighting (where the load is not expected to vary, and a constant current is desirable).

Voltage regulation is achieved with the error amplifier.  You could add even more stages externally, say if you wanted to control the voltage such that it... I don't know, sets a constant or adjustable power level in some load.  You could also connect such an error amp in parallel (driving COMP via diode, or by the current limit pin), so you don't have chained loops.  Both ways have their advantages and disadvantages.

Tim

[theoldwizard1]

The current limit transistor is limiting output voltage of EA at a value depending of R1/R2.
Limited Output voltage of EA is then compared with amplified Isense.

STOP !  You did a "hand wave" over part I don't understand !

What are the current source, diode and 0.5 "battery" doing in the circuit ?

[T3sl4co1l]

Don't know what the diode means (they don't indicate bias current through it, so what determines its voltage drop?), but stuff like that is usually done to indicate or suggest there is a voltage drop there.  Which therefore tells you the relation between the COMP pin voltage and what the current limit actually is.

Tim

[oldway]

Explanation can be not very clear because we don't have the true internal schematic of the UC2856.

EA output is probably an "open collector" configuration, so the 0.5mA bias current is needed.

0.5V negative offset on COMP input is also needed to ensure to block pulses when there is no EA output. (pin 7 = 0V)

It's also possible that the 0.5V is the voltage drop of the diode. A perfect diode (without voltage drop) more 5.5V forward voltage drop.