choice between regulator and DC/DC convert

[[email protected]]

Hi,all
I am designing a power supply circuit(12V-3.3V, 12V-1.2V)for a board
which include two fpga,a/d,dsp and the total power dissipation is about
10W+. Now I am a little confused about the function of regulator and
DC-DC converter. In my first opinion, the DC/DC convert had more
efficiency. But the FPGA EVM board from altera use LM2678 and LT1085 to
provide the power. Does I really need a DC-DC converter?

 

[Rene Tschaggelar]

Hi,all
I am designing a power supply circuit(12V-3.3V, 12V-1.2V)for a board
which include two fpga,a/d,dsp and the total power dissipation is about
10W+. Now I am a little confused about the function of regulator and
DC-DC converter. In my first opinion, the DC/DC convert had more
efficiency. But the FPGA EVM board from altera use LM2678 and LT1085 to
provide the power. Does I really need a DC-DC converter?

While a regulator dissipates the unwanted voltage
of 12-3.3 = 8.7V or 12-1.2 = 10.8V times the nominal
current, the DC/DC converts at 90% or so efficiency.

You're not going to dissipate 10W easily and noiseless.
10W require quite a heatsink.

Rene

 

[Frank Bemelman]

Hi,all
I am designing a power supply circuit(12V-3.3V, 12V-1.2V)for a board
which include two fpga,a/d,dsp and the total power dissipation is about
10W+. Now I am a little confused about the function of regulator and
DC-DC converter. In my first opinion, the DC/DC convert had more
efficiency. But the FPGA EVM board from altera use LM2678 and LT1085 to
provide the power. Does I really need a DC-DC converter?

Well, the LM2678 *is* a switcher - not an ordinairy regulator.
So it is much more efficient than a linear regulator. If you
need one, you have to decide yourself.

Infact they are all DC/DC converters, but when speaking
about DC/DC converters, it's often switched mode supplies,
often with galvanic separation too. I suppose you could
even call a resistor divider a DC-DC converter

 

[Roger Hamlett]

While a regulator dissipates the unwanted voltage
of 12-3.3 = 8.7V or 12-1.2 = 10.8V times the nominal
current, the DC/DC converts at 90% or so efficiency.

You're not going to dissipate 10W easily and noiseless.
10W require quite a heatsink.

In the context of the original post 'wording', the LM2678, _is_ a DC-DC
converter. Hence the prototype board has got a DC-DC converter. However
such units only using a simple inductor, or a PWM without an inductor, are
commonly referred to as 'switching regulators', with the phrase 'DC-DC
converter' tending to be used more commonly for devices that involve a
transformer, or some other form of voltage 'shift'. Hence the prototype
board is using a switching regulator to keep the power losses to a
reasonable level (I'd guess generating the 3.3v with this), and then using
the LDO linear regulator, to give the 1.2v. The advantage of the switching
regulator, is the high efficiency. It's disadvantage is noise.
As said above, the device needs a switching converter, or a large
heatsink. Remember also that using a device like this, lowers the delivery
requirements. The poster talks about power dissipation, if he actually
means that the final circuit 'consumes' this much, and making a guess that
most of the current is drawn from the 3.3v rail, with (guessing), perhaps
1A at 1.2v, and 2.6A at 3.3v, then using linear regulators to deliver this
would give a power dissipation in the regulators of (12-3.3)*2.6 = 22.6W,
and (12-1.2)*1 = 10.8W. The system would be using 9.8W itself, bringing a
total required from the incoming supply of 43.2W. 3.6A at 12v...
However using the 2678, and the LT1058, would give instead 3.3v at 3.6A
(because the linear regulator is running from this rail), with an
efficiency of (say) 90% (the switcher can do slightly better than this),
so (3.3*3.6)/0.9 = 13.2W from the incoming supply, and only about 1.32W
needing to be dissipated in the switcher. The LT1085, would dissipate
(3.3-1.2)*1 = 2.1W. The current drawn on the 12v rail becomes only 13.2/12
= 1.1A.

Best Wishes

 

[Mac]

Hi,all
I am designing a power supply circuit(12V-3.3V, 12V-1.2V)for a board
which include two fpga,a/d,dsp and the total power dissipation is about
10W+. Now I am a little confused about the function of regulator and
DC-DC converter. In my first opinion, the DC/DC convert had more
efficiency. But the FPGA EVM board from altera use LM2678 and LT1085 to
provide the power. Does I really need a DC-DC converter?

You seem to have gotten some good answers to most of your questions.

I will just point out that switching supplies should have input and output
filtering on them unless you don't care about putting a lot of noise on
the input power supply for some reason.

I overlooked this once.

--Mac