EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Circlotron on June 26, 2016, 05:52:09 am
-
Looking at updating an old board that uses a SG3525 PWM controller to drive a pair of side-by-side boost converters in an automotive environment. Been making these for many years and it is quite a reliable old thing so I don't want to change much. Output is 16V at 15 amps. It works down to 11.1 volts input whereupon it hits max duty cycle. The datasheets says max duty cycle each side is typically 49% and minimum of 45% but this one runs out of puff at 40%. It runs at just over 100kHz and I'm thinking maybe it has some internal propagation delays that limit the on time at high frequencies? I could fairly easily drop the frequency a bit and add a bit of inductance to the two boost toroids if that would help. I don't really want to change the chip type unless there is something really better that would basically drop straight in. Does the propagation delay / frequency / max duty cycle sound likely?
-
There is a fixed deadtime determined by the timing capacitor and dead time resistor (if used).
-
...Output is 16V at 15 amps. It works down to 11.1 volts input whereupon it hits max duty cycle. The datasheets says max duty cycle each side is typically 49% and minimum of 45% but this one runs out of puff at 40%. It runs at just over 100kHz...
Theoretically, 40% duty should be just enough to allow for 0.7V of combined drop across the switch and inductor and 0.5V across the output diode (ie - 10.3V input and 16.5V output). It is entirely possible to lower these voltage drops, particularly if the switch is a MOSFET.
The amount of inductance doesn't affect the duty cycle, just the peak to peak ripple, with the exception that a higher inductance tends to mean more winding resistance, all else being equal, so a slightly higher voltage drop. Conversely, a lower inductance means a higher peak current in the switch, so a higher IR drop across the switch if it is a MOSFET (probably not much difference if it is a BJT, however). That said, your inductor should be in the range of 3.3uH to 10uH - with 4.7uH ideal - and as it will need to withstand around 25A peak without saturating it ought to have less than 10 mOhms of series resistance.
The output diode should be a 30V Schottky for lowest voltage drop and to eliminate reverse recovery losses from running in continuous conduction mode. You might find a higher current rated diode in the same package which will help to lower the forward voltage drop, but I wouldn't expect huge miracles here; a decrease of 0.1V would be about the best you can expect.
Aside from adjusting deadtime via a resistor between Ct and Discharge, you can also try lowering the timing capacitor value and increasing the timing resistor value. Finally, you could also try a different manufacturer of the '3525. Old ICs like this get run on old fabs which have the worst tolerances, and the ICs are rarely (read: never) fully tested. You might just have a bad batch because a 40% max. duty cycle would render this unusable for most offline forward and flyback converter designs.
-
It's a bad design anyway, because in general, CCM inductor currents won't track. (And if it's DCM, it's still bad because the ripple is humongous.) If there's current feedback from the total, dynamics won't be bad, but I would suppose it's also a typical voltage control circuit, which is just ridiculous.
Tim