Power Supply (Redesign)

One of my very first home built projects was a variable power supply. Sure, I could have bought one but where’s the fun in that? Plus, I wanted to learn more about them and be able to repair it in case the need arose down the road. I found a simple design for a variable output supply (5-30V @2A) on the internet based on the LM350 linear voltage regulator. Using that circuit plus a little additional info from the spec sheets gave me what I needed. I laid out and built my own PCB, collected the parts and built it back in the winter of 2013/2014. It worked…but, I forgot about one critically important factor. Linear voltage regulators are very inefficient and with high loads coupled with large voltage differentials (Vin-Vout) generate a lot of heat. I didn’t account for that with my circuit. Instead, I just stuck on a modest, clip-on heat sink thinking that would suffice. It didn’t. The first time I used it at 5V with a moderate load, it shut off. Upon inspection I discovered that the LM350 was extremely hot so it was obviously going into thermal shutdown mode.

Fast forward a couple of years...I finally decided to completely revamp the design and start over. I made the decision to build three supplies into my existing enclosure: two new fixed supplies of 5V and 12V @2A each and a "properly designed" variable supply covering 5-25V @1A. For the two fixed supplies I choose to go with buck converters (switchers) because of their greater efficiency (ie less heat generated) under load. I went with the Texas Instruments LM2596 chip. It only requires the addition of a couple of external capacitors, a diode and a power inductor and you're done. The spec sheets actually contain a sample PCB layout along with a step by step guide for choosing the proper external components (including heat sink) based upon your desired maximum load. I scratch built the boards, ordered the parts (I got the inductors as free samples from the manufacturer) then assembled and tested them. No problems, and they held up well even under full load.

Next came the more challenging task, the variable supply.  This time I made sure to incorporate power dissipation and component thermal resistance values into my design to insure there would be no more heat issues.  For my variable supply, Vin = 30V and the lowest Vout = 5V.  At 1A this works out to be 25W:

Power (in watts) = V (volts) x I (load in amps) = (30-5) x 1 = 25W dissipated by the LM350!!!

Obviously, way too much to handle, so I made a slight design modification.  I added a bypass power transistor to improve the load distribution (see below):

This is a well-known and proven technique to increase the load handling of a linear regulator. The value for resistor R1 is selected to develop .7V drop when a specific current passes through it, thus turning on transistor Q1.  The power transistor then handles the majority of the load while the regulator takes care of the rest.  I decided to set the maximum regulator current at .25A and let the transistor handle the remainder (.75A).  Recalculating the power consumption led to much better numbers:

Power (regulator) = (30-5) x .25 = 6.25W…okay

Power (transistor) = (30-5) x .75 = 18.75W…okay

With the power dissipation completed I could now focus on the thermal resistance (TR) calculations.  These numbers are available on the individual spec sheets of each component.  TR is expressed in units of °C/W.  In other words, for every watt a component must dissipate, it will rise “x” number of degrees above the ambient temperature.  Therefore, the lower the total TR the cooler the device will be at a particular load.  The goal is to keep the junction temperature of each device below its maximum rated value (125°C for the regulator and 150°C for the transistor).  I found a couple of decent heat sinks that offered good performance for their size, then did my calculations:

transistor		regulator
	TR (°C/W)		TR (°C/W)
J-C	1.39	J-C	3.00
C-H	1.09	C-H	0.20
H-A	3.10	H-A	5.10
total:	5.58	total:	8.30

note: J-C = junction to case; C-H = case to heat sink; H-A = heat sink to air

Finally:

Junction temp of transistor under max load = 5.58 x 18.75 + 25°C (ambient) = 130°C

Junction temp of regulator under max load = 8.30 x 6.25 + 25°C (ambient) = 77°C

Both are under the maximum values with room to spare so everything should be fine.

The completed unit is shown below: