LM338 PSU 0-30v (hunting for good circuit)

Thread Starter

oidium45

Joined Apr 24, 2010
138
Hello, I am getting ready to start a 0-30/32v regulated PSU using a single LM338T. I have been browsing/searching for a good circuit with this chip. It seems that most of the schematics that I have found use 24-25v input. I was hoping to use somewhere around a 34-36v input. Does anyone have a good schematic with 30-32v output? As of now I am leaning toward the one on the data-sheet.

Also, most of the circuits that I have found use a 4700-10,000uf cap for a filter. I have a vary large Mallory 8000mfd 40vdc (max surge 45vdc) aluminum electrolytic capacitor that I would like to use. It says "computer grade". Any suggestions would be greatly appreciated!
 

Thread Starter

oidium45

Joined Apr 24, 2010
138
I was hoping someone could recommend a circuit other than the one listed on the data-sheet. But thank you for your feedback.
 

SgtWookie

Joined Jul 17, 2007
22,230
Why do you feel that you need such a large voltage range?

If you're going to be using the supply mostly in the 25v to 30v range, then that's OK.

However, if you are also planning on using it for lower voltage projects (like 3v-10v) then you will have a power dissipation problem.

Let's say just for example that you have a reasonably stable 35v on your filter capacitor after rectification, and it'll stay within a few volts of that whether your regulator is putting out 10mA or 5A (it would have to be powered by a transformer rated for >10A to do that...)

Let's also say that you are powering a project that requires 5A @ 30V.
So, you have 35v in, 30v out, 5A current.

The load will be dissipating 30v*5A=150 Watts of power, and the regulator will be dissipating (35v-30v)*5A = 5v*5A = 25 Watts of power.

Let's also say that you want to run the supply in a room that is nominally 25°C/77°F.
The LM338T (TO-220 package) has a thermal resistance to a heat sink of 4°C/Watt.
The LM338T (TO-220 package) has a thermal resistance to ambient (no heat sink) of 50°C/Watt.
The LM338K (TO-3 package) has a thermal resistance to a heat sink of 1°C/Watt.
The LM338K (TO-3 package) has a thermal resistance to ambient (no heat sink) of 35°C/Watt.

Both will shut down at 125°C.
So, with no heat sink and operating at ambient 25°C, the LM338T will shut off when it is dissipating 2 Watts of power (50°C*2w+25°C ambient=125°C), and the LM338K will shut off when it is dissipating under 3 Watts of power (35°C*2.86w+25°C ambient=125°C).

You will need a heat sink big enough to make up for the difference between the 25 Watts and the 2 or 3 Watts that they can dissipate themselves.

Now let's try powering a 3V load that needs 5A current. We still have 35V on the filter capacitor.

The load will dissipate 3v * 5A = 15 Watts of power, and the LM338 will dissipate (35v-3v)*5A = 32v * 5A = 160 Watts of power. Now you have a really, really big problem of heat dissipation. :eek:

Let's just look at the LM338K-package (TO-3 case). As a review, the LM338K (TO-3 package) has a thermal resistance to a heat sink of 1°C/Watt. Remember, the regulator will shut down when it's internal temp reaches 125°C.

With the TO3 package, dissipating 160 Watts means that you will need a PERFECT heat sink that can keep the outside of the package at 125°C-160°C, or -35°C just to keep it from going into thermal shutdown. Basically, you'd need to run the thing at the North Pole - with a huge heat sink. Even freezing it into a block of ice would not be good enough.

You might use water cooling. Water boils at 100°C at sea level. However, that only gives you the basic 25 Watts of dissipation from the steel TO-3 case.

So, what is your plan to get rid of all of the heat you're going to subject your regulator to?
 

Thread Starter

oidium45

Joined Apr 24, 2010
138
Thanks for the good info! So what you are saying is that if i want to use the lm338 for a variable supply it would be best to keep it somewhat closer to the input voltage to avoid excess heat in the chip? Maybe I should look into other regulators as a possibility?

My intention was to build a variable regulated bench power supply. I could not get myself to purchase a cheap 1-2A model when I could possibly build a beefier one much cheaper and possibly learn a few things in the process... Can you recommend any low cost circuits that may fit my needs. I was looking at somewhere around 0-30v+ and approximately 5 amps. I am sure that I could easily get by with 2 amps but there is always the occasion where you need a bit more...

The transformer that I am using is setup for 12/24/36/48v and can easily produce 5-10 amps without too much strain.
 

SgtWookie

Joined Jul 17, 2007
22,230
You can cut your regulator heat dissipation requirements considerably by using the various secondary taps on your transformer. Then things (the power dissipation problem) become much more manageable.

With linear regulators, your big problem will be (input voltage - output voltage) * current = Watts. The output from a linear regulator may be smooth as silk, but unless you can get rid of the heat, they will shut down on you.

If you want to build a 0v-30v linear regulator, it will not be cheap to do, and you will not get by with a single regulator IC. There is just not a practical way to get rid of all of the heat. You will need the regulator IC to drive multiple output transistors that are mounted on VERY LARGE fan-cooled heat sinks.

I'm not kidding.
 

eblc1388

Joined Nov 28, 2008
1,542
You can cut your regulator heat dissipation requirements considerably by using the various secondary taps on your transformer.
Good advice.

You can easily achieve that by using center-tapped secondary windings and arrange the rectifiers configuration as either bridge or full-wave using a simple toggle switch.

Using the toggle switch will immediately cut the rectified voltage level by half, reducing the regulator heat dissipation at low output voltage.
 

Thread Starter

oidium45

Joined Apr 24, 2010
138
Thank you! I am going to give that a try! I am using a MOT tapped at 12/24/36/48v. I was thinking of making three different circuits with three chips (0-12/12-24/24-32 i know "horrible idea") but I like the idea of just using a switch to control the input voltage. Much better then making three different circuits. Thanks!
I am still a rookie. What can I say...
 

k7elp60

Joined Nov 4, 2008
562
I have built a number of adjustable power supplies in the range you are talking about (2-30V). I have tried 3 terminal regulators like the LM338 and a LT1083CP. Even with a large enough heatsink to dissipate a large quantity of watts, I have found that the internal circuitry of the 3 terminal regulators sense the in/out differential voltage and the current and often shut down even with the large heat sink. I have had the greatest sucess with a LM723 and external pass transistors. It has an internal reference source, but requires an external resistor for current limit.
 

retched

Joined Dec 5, 2009
5,207
Using your center taps into a rotary selector switch, you could choose which tap enters the circuit.

But even 12v to 2v @5a is a good amount of heat to dissipate. Do you have a PC power supply?

You can use that for a 3.3, 5, and 12v supply, very easily, then use your transformer to build a high-voltage supply.

You could put it in an enclosure (even the old PC) and drill some holes on the front to mount some connectors and even a panel meter or two, for voltage and current.

You will have the PC supply now to get you going on your home built supply.

If you now decide on a 14 - 30v variable supply with your transformer, it will be a much easier project.

A high/low switch can switch from the 24v and 36v taps. So you could essentially build a 14-24v and a 24-30v supply circuits.

You wont need a large variability in any one circuit, so you will only be selecting between voltages for different projects.
 

Audioguru

Joined Dec 20, 2007
11,248
The LM158 and LM358 voltage regulators are smart.
When the voltage between the input and output is more than only 10V then they reduce the max output current to as low as 1.2A to protect themselves.

But you still need dry ice, liquid nitrogen or a powerful fan to cool the huge heatsink.
 

studiot

Joined Nov 9, 2007
4,998
Austin's excellent observation (post#2) seems to have got lost in the discussion so I will repeat it.

You need to generate a negative base rail for the 338 to operate down to zero volts.
 

retched

Joined Dec 5, 2009
5,207
Austin's excellent observation (post#2) seems to have got lost in the discussion so I will repeat it.

You need to generate a negative base rail for the 338 to operate down to zero volts.
Good point. That will add a little to the complexity of the project.

The datasheet of WHATEVER you are using should be read ad nauseam.

The circuits in the datasheets are valuable design tools. (Even if you dont want to use them exactly)
 

SgtWookie

Joined Jul 17, 2007
22,230
Not going all the way to 0v shouldn't be a big deal, really.

What circuit are you going to operate on less than 1.25V, anyway? A Joule Thief?
 

studiot

Joined Nov 9, 2007
4,998
Well I find my power supply good for testing components at less than 2 volts, in current limit mode. Trying to test them in constant voltage mode at >1.25 volts is a waste of time I might just as well transfer them straight into the waste bin.
 
Top