Say your designing a multiple output smps - forward type, for argument sake - with +5 and +3.3V. I've seen two ways (more actually) :
1. Two separate windings with a coupled output inductor. PWM feedback is from either 5 or 3.3V.
2. One 5 Volt winding tapped at 3.3 Volts, giving 2 outputs. The pwm operates on the +5V winding.
I like method #2 because the tapped load will load down the +5 output, and so the pwm will respond, keeping cross regulation in line.
Any comments on these 2 methods?
1. Two separate windings with a coupled output inductor. PWM feedback is from either 5 or 3.3V.
2. One 5 Volt winding tapped at 3.3 Volts, giving 2 outputs. The pwm operates on the +5V winding.
I like method #2 because the tapped load will load down the +5 output, and so the pwm will respond, keeping cross regulation in line.
Any comments on these 2 methods?
Method #2 is good, as long as the portion of the secondary carrying current for both the 5V and the 3.3V sections can handle both currents.
Steve
Steve
For low powers, a flyback is fine. In such a converter, it's advantegous to use two independent windings, respectively with 3 and 4 turns for 3.3V and 5V outputs, and place them in the same layer in a bifilar fashion to get the best cross-coupling possible.
For medium powers, you may use a buck-derived converter, forward or push-pull, with 5V (6V) secondaries tapped at 3.3V (4.5V) and a coupled output inductor.
For high powers, the best solution is also a buck-derived converter, but with a single 5V secondary feeding two independent sets of output diodes and output inductors, one directly and the other through a saturable reactor in order to get independent duty cycle adjustment. This is the smartest approach and it's found in most ATX computer power supplies.
For medium powers, you may use a buck-derived converter, forward or push-pull, with 5V (6V) secondaries tapped at 3.3V (4.5V) and a coupled output inductor.
For high powers, the best solution is also a buck-derived converter, but with a single 5V secondary feeding two independent sets of output diodes and output inductors, one directly and the other through a saturable reactor in order to get independent duty cycle adjustment. This is the smartest approach and it's found in most ATX computer power supplies.
How much current doe each rail have to be spec'd for? You could do a single 5V rail with a low dropout 3.3V linear regulator running off the 5V rail. Or a even a single 7V or so rail with LDO 5V and 3.3V linear regulators.
Linear regulators come very handy when the same transformer has to provide not only 5V and 3.3V at low currents but also other outputs whose voltage has to be regulated over a wide range of loads, thus making cross-regulation very difficult.
Implementations of that are found on almost every modern CRT monitor and TV, where the same transformer has to provide one or more well regulated +50V to +150V outputs for the deflection circuits, +-15V for the audio power amplifier, +12V and +8V for linear ICs, and +5V and +3.3V for logic stuff. The lower outputs usually come all from a single 12V to 16V rail through low-cost L78xx regulators or similar.
However, this approach is only practical for low currents. Anyway, a simple flyback will provide good cross-regulation if it only has 5V and 3.3V outputs and current demands are low.
Implementations of that are found on almost every modern CRT monitor and TV, where the same transformer has to provide one or more well regulated +50V to +150V outputs for the deflection circuits, +-15V for the audio power amplifier, +12V and +8V for linear ICs, and +5V and +3.3V for logic stuff. The lower outputs usually come all from a single 12V to 16V rail through low-cost L78xx regulators or similar.
However, this approach is only practical for low currents. Anyway, a simple flyback will provide good cross-regulation if it only has 5V and 3.3V outputs and current demands are low.
well, in this case the power isn't very high. At present, power requirements are about 10 watts @ 5V watts, and maybe 10 watts @ 3.3V. Even allowing some headroom, the total supply is going to be 30 watts or less.
Can someone clue me in on the mechanism that keeps cross regulation work. In my case #2, it seems pretty clear that the lower voltage tap loads down the entire leg where the pwm is sampling. But how does a winding that is not directly sampled, cause the pwm to respond?
BTW, my example #2 is drawn in Fairchild data sheet on page 14
http://fairchildsemi.com/ds/FS/FSDM0365RN.pdf
In fact, it has BOTH a tapped winding and additional windings! At least this thing makes some sense to me. Unfortunately, they reference an appnote for using the part in forward mode and that notes suggests using coupled inductors.
Can someone clue me in on the mechanism that keeps cross regulation work. In my case #2, it seems pretty clear that the lower voltage tap loads down the entire leg where the pwm is sampling. But how does a winding that is not directly sampled, cause the pwm to respond?
BTW, my example #2 is drawn in Fairchild data sheet on page 14
http://fairchildsemi.com/ds/FS/FSDM0365RN.pdf
In fact, it has BOTH a tapped winding and additional windings! At least this thing makes some sense to me. Unfortunately, they reference an appnote for using the part in forward mode and that notes suggests using coupled inductors.
In a flyback, it is cross-coupling between secondary windings what ensures good cross-regulation. There is no need to derive the feedback signal from both outputs, altough taking the average from them is usually useful (not shown in your PDF).
Also, the layout of the sample transformer shown in that PDF is not necessarily optimum, but good enough for a low-cost DVD player.
Also, the layout of the sample transformer shown in that PDF is not necessarily optimum, but good enough for a low-cost DVD player.
Really - some designs average multiple outputs? I'll have to think about that one a little bit.
What is it that you don't like about that transformer stack? Would you want bifilar windings instead as a way to improve coupling?
What is it that you don't like about that transformer stack? Would you want bifilar windings instead as a way to improve coupling?
- Status
- Not open for further replies.