Help using an LM338T regulator

Thread Starter

JConnor

Joined Dec 15, 2014
15
Hey all,
New member here, looking for some help with a very simple circuit. See below. The same circuit applies to the LM338T regulator, except it can handle higher currents (up to 5A).


I have used this circuit before quite successfully many times, using a .50 Ohm resistor, three 10W LEDs (Vf 3.0-3.3V each), and a 12V bench power supply (adjustable from 11.0 to 15.0V). Typical behavior has the current entering regulation at 12.5V. After this point, the output current remains rock-solid at 2.5A as I increase supply voltage up to the maximum 15.0V. The excess voltage turns into excess heat. Thus, the circuit requires about 2.5V over the combined LED voltage to remain in regulation. These chips were labeled Fairchild Semiconductor LM338T JM14RG P+.

Here's the issue. I recently purchased a new batch of regulators from a new supplier. I connected one of them exactly the same way, and at 15.0V I am only seeing 1.8A. I am not shorting any of the leads, and the setup is identical to what I used before. The new chips are ON Semiconductor labeled LM338T RBQ209G.

So, I bought yet another batch, thinking they were bad, from the supplier I originally purchased from. These are Fairchild Semiconductor labeled LM338T JM32RG P+. I connected it in the same way as before, and they still only reach 1.8A.

So I am using the same chips, from two different suppliers, in the same configuration, and getting the same results, both of which are different than they should be.

I do not have an electrical engineering degree and struggle with the theory of how this circuit works. I know there is a 1.25 reference voltage created between the output and adj terminals, which is how you calculate the resistor required.

Can anyone shed some light on why this isn't working, and what I can possibly test for to diagnose why?


edit: Here is the data sheet for this type of linear regulator. http://www.ti.com/lit/ds/symlink/lm138.pdf
 
Last edited:

Papabravo

Joined Feb 24, 2006
21,159
There are only two obvious possibilities.
  1. The resistor is wrong. That is, it not 0.5 Ohms any more
  2. The Vref is not 1.25 volts any more on the ON Semiconductor Part and the Fairchild part
The operative equation is:

\(I_{OUT}=\frac {V_{REF}}{R}=\frac{1.25 \text Volts}{0.5 \text Ohms}=2.5 \text Amperes\)

So can you measure the resistance of the resistor?
And can you measure Vref?

Either the resistor is ≈ 0.7 Ohms, or Vref is ≈ 0.9 Volts
If I was a betting man I'd place even money on the resistor and take very long odds on the reference voltage.
 

Thread Starter

JConnor

Joined Dec 15, 2014
15
I took out a new LM338T and new 0.47 Ohm resistor, to be sure. My Fluke leads have 0.2 Ohm resistance. I measured 0.7 Ohms resistance, so the resistor is ~0.5 Ohm. There is only 0.X resolution on resistance measurements.

Load voltage 13.48V.
Resistor 0.47 Ohm
0.79Vref
Current 1.6A

Same exact behavior as before. It's not achieving the correct Vref it seems.

So, I tried something else. I used a 27 Ohm resistor.

Load voltage 13.48V.
Resistor 27 Ohm
0.80Vref
Current .046A

Not sure what was happening here exactly, but when I would touch my multimeter probe to test Vref, the light output would change, as if the multimeter was affecting Vref. Another curious thing was, when I lowered the source voltage down to about 11.5, the Vref went up, and eventually reached 1.25V at or below 11.5V.

Really odd behavior. Thanks for your help.
 

ronv

Joined Nov 12, 2008
3,770
Looks like the Vf on your LEDs is higher than expected. They should have clamped the load voltage at 9.7 volts unless your saying the output of the regulator is 13.48 - even so, same problem. Try it with 2 LEDs.
 

Thread Starter

JConnor

Joined Dec 15, 2014
15
Looks like the Vf on your LEDs is higher than expected. They should have clamped the load voltage at 9.7 volts unless your saying the output of the regulator is 13.48 - even so, same problem. Try it with 2 LEDs.
I measured the Vf of each LED individually at around 3.15V each. They are rated to be 3.3V at 3A, so this is in-line with the expectation.

I can try with 2 LEDs and see what happens though!


EDIT - What I meant with "load voltage" was the power supply voltage under load. With no load it is about 13.58V and drops to 13.48V under load.
 

ronv

Joined Nov 12, 2008
3,770
Ahh, I see. So not enough input voltage. The regulator needs about 2.3 volts across it to regulate:
upload_2014-12-15_20-14-10.png
Another 1.25 volts for the resistor gives you only enough if everything is perfect. Should work ok with 2 LEDs.
 

Thread Starter

JConnor

Joined Dec 15, 2014
15
So you're supposed to add the dropout voltage and the reference voltage? In my experience, I only needed about 2.5V over the LED voltage, which matches the chart you posted above for Iout = 3A. Even so, when I raised the source voltage up to 15V, it never went into regulation, and that's 5V over the LED voltage.

Here's another guide I referenced. See 1:30.

Thanks for the help thus far :) I am considering buying a component from a US supplier to verify my sanity.
 

ronv

Joined Nov 12, 2008
3,770
That's ok for the voltage regulator, but the current regulator drops the 1.25 across the resistor so it adds in. It should have worked with 15 volts. Another possibility is that it is oscillating. If the power supply or the LED is a ways away they sometimes do that.
Here are 3 simulations so you can see the drop, an oscillation and a fix for it.
 

Attachments

Thread Starter

JConnor

Joined Dec 15, 2014
15
That's ok for the voltage regulator, but the current regulator drops the 1.25 across the resistor so it adds in. It should have worked with 15 volts. Another possibility is that it is oscillating. If the power supply or the LED is a ways away they sometimes do that.
Here are 3 simulations so you can see the drop, an oscillation and a fix for it.
I shorted one of the LEDs and tried again. As I varied the supply voltage from 11 to 15 volts, the current went from 1.6A to 1.9A and then once I hit 14V, it started going DOWN to 1.2A. This is very abnormal.

Thanks for helping me walk through the problem and consider all possible issues. I'm going to buy more and see what they do.
 

Thread Starter

JConnor

Joined Dec 15, 2014
15
Yep, the LM338, resistor, and LEDs are all on a heatsink. I know the heatsink of the regulator is the same as the OUT pin, so it's electrically isolated from everything else.
 

JohnInTX

Joined Jun 26, 2012
4,787
Are you sure its not going into thermal shutdown as you increase the power dissipated with increasing input voltage? The fact that it begins to work OK then reduces its output over time and as the input voltage is increased smells like the thermal shutdown is active.

A few rough jabs at the Lil Professor shows with 2 LEDs you have 15V - 6.6Vled- 1.25Vref = 7.15Vacross the regulator. At 2A, the regulator dissipates 14.3 Watts. To keep Tjunction <125C (100 above ambient) needs a Thermal Resistance junction to ambient of no more than 100degC/14.3W = ~7 degC/W. The junction to case is already 4degC/Watt for the TO-220 (NDE package). That means your heatsink must have a thermal rise rating of no more than 3degC/Watt. Yours doesn't look big enough.

All said, I don't know why TI works and Fairchild/ON semi don't if they are legit and truly the same part.

Good luck.
 

Thread Starter

JConnor

Joined Dec 15, 2014
15
Are you sure its not going into thermal shutdown as you increase the power dissipated with increasing input voltage? The fact that it begins to work OK then reduces its output over time and as the input voltage is increased smells like the thermal shutdown is active.

A few rough jabs at the Lil Professor shows with 2 LEDs you have 15V - 6.6Vled- 1.25Vref = 7.15Vacross the regulator. At 2A, the regulator dissipates 14.3 Watts. To keep Tjunction <125C (100 above ambient) needs a Thermal Resistance junction to ambient of no more than 100degC/14.3W = ~7 degC/W. The junction to case is already 4degC/Watt for the TO-220 (NDE package). That means your heatsink must have a thermal rise rating of no more than 3degC/Watt. Yours doesn't look big enough.

All said, I don't know why TI works and Fairchild/ON semi don't if they are legit and truly the same part.

Good luck.
It never works, from the moment I apply power. I only needed to test it for 10-15 seconds to tell that the current isn't reaching the specified current. The components barely got warm.

The video I posted above isn't mine. My heatsink is a massive brick of aluminum with fins :)

Yeah, perplexing. I'm thinking they are fake, or mislabeled LM317's, or something else.
 

ronv

Joined Nov 12, 2008
3,770
I love simulators - if you can just give them all the information.
This kind of looks like your description.
Try it with a cap close to the regulator and ground.
 

Attachments

Thread Starter

JConnor

Joined Dec 15, 2014
15
Just a little update.
Bought new components from a ... 3rd supplier. They work as they should! The old chips were bogus.
 
Top