36V to 5V step down converter with pulse-skipping in light load

Hello,

this is my step-down converter:

..it is for ...
Vin = 36V
Vout = 5V
Pout = 1.67W Max

SCHEMATIC:
https://i55.tinypic.com/28lwh9s.jpg

It must go into skip-cycle in light load, hence the comparator.

It has a lot of components, but i see no cheaper way of doing this...since i need a good efficiency in light load, and a controller with low operating current.

I found some Buck IC's but they had high internal FET current limits
and so they would have high "staircase" current at start-up and POTENTIALLY blow the 500mA fuse which lies ahead of the above circuti on the 36V rail.
-Also, the Buck IC's were extremely expensive, and most dod not have pulse-skipping in light load.


by the way here is the LTSpice simulation file:-

2shared - download 5V SEPIC__with tripler bias_FORUM.asc


do you know of a cheaper way?

use LM7526

LM22674-5.0 : 1.76$ according to National's webbench application.

stefaan

Unfortunately the LM22674 has no pulse skipping in light load, and draws up to 6mA of quiescent current....6mA is too much......and the 6mA is the non-switching operating current....when switching the operating current will be higher.

I cannot find any details on LM7526.

-are there really no chips which can do my task i am now wondering?

Ha, I see, I didn't look careful enough

Maybe my next proposal is better : LM5007 or LM5010

It doesn't has pulse skipping, but it is designed for low power bias supplies. They use hysteretic regulation, so there is no minimum duty cycle, and the switch is only operated when the output voltage is too low. However this method results in a bit higher output ripple. So the effect is the same, reduced switching events.

If you decide on this component, I can tell you more. It's my favorite chip for low power, high input conversions. I didn't propose at first instance, because your input is only 36V, these chips go to 75 or more. This is of course no argument of not using it

It draws still 650µA typ when not switching :-(


Stefqqn

you can find details here;
**broken link removed**

hi
thanks for looking in to it....
though LM2576 draws 10mA of quiescent current.......and LM5007 has too much output voltage ripple , which is only quelled by adding a lossy filter...LM5010 does not have skip cycle.

For the LM5007, there is a nice trick to let the system work with a lower output ripple. Can you tell more on the requirements of ripple voltage, quiescent current and efficiencies at several output power levels?

What is the budget for component cost doe this? You normally can not have all good things together (low cost, less components, low ripple, low quiescent current, good efficiency, space requirements, ...)

Stefaan

Regarding cost, just as cheap as possible....there are no upper cost limits.....just the cheapest that achieves the following:-

Vin = 36V
Vout = 5V
Pout = 1.67W Max
It must go into skip-cycle in light load, so as to maximise light load efficiency.

The long story of this is that it is the 5V regulator at the output of an offline SMPS which has an isolated 36V secondary rail.
-due to reverse compatibility reasons, we cannot add a 5V winding to the SMPS transformer, hence we need this 5V smps.

.....the worst thing is that this SMPS must draw no more than 500mW power from the mains, when there is a 170mW load on the output of this 5V step-down regulator (which feeds off the 36V rail of the SMPS)

-so we need this 5V smps to be efficient in light load.........especially since its load is not 170mW continuous....but this is the avrage output power....in fact, the output load is pulsed on and off at low frequency (~1Hz with 5:1 off/on duty cycle)

eem2am, I don't understand you are looking for pulse skipping, but don't allow the ripple of for example the LM5007. To skip pulses, the algorithm needs to know somehow the output is higher than normal in the first place. This means a higher ripple at light load.

I'm even more confused with your explanation. In the first spec you say maximum 1.67Watt (about 0.33Amps). A bit further down you talk about 170mW output load (about 0.03Amps) or do you mean 1.7Watt here??
However if you say the SMSP can not draw more than 500mW from the input, you either never get the 1.67W output (like stated first) or you have room enough to get the 170mW out of it (like stated further down).

I think the solution is near for your problem, but you need to be very clear on the spec. Of course I can just wrong interpret your post.

Stefaan