LM338 voltage regulator poor regulation when breadboarding

Thread Starter

Hamlet

Joined Jun 10, 2015
519
When I set this at 9.91v, then attach a load, it drops to 9.72v. That's almost 2%.



My supply is a 7.5A switcher, taken from old laptop. I also used
a 10A transformer/rectifier/cap as a supply. Same problem.

I am using thick, short wires, & I am taking my voltage readings
directly from the output terminal of the LM338K (the case) before the breadboard,
so as to eliminate voltage drop thru the limited current handling
of those tiny little holes/terminals of my breadboard. I am using a
LM338K, old stock, Linear Technology device, of which
I have several. It's from Hanford, WA, circa 1980's, surplus. Not China.
I am using a large copper cpu heatsink.

Anyway, I am using a 12v, 50w halogen as my load. I am at almost 2%
of fluctuation under load from initial set point. Is this normal?

I thought these LM317/350/338 regulators were tighter than this?
 

Attachments

blocco a spirale

Joined Jun 18, 2008
1,546
I would expect to see some voltage drop between zero and 4A loading, the drop you're seeing only amounts to around 0.05Ω series resistance. For most applications this would be acceptable.
 

#12

Joined Nov 30, 2010
18,224
The datasheet confirms that 2% is not the advertised result. You are just going to have to do some detective work in the millivolt range to see how your external lead resistances affect the circuit. If that doesn't explain the sag, you will have to try a different LM338, add a helper transistor, or use something like an LM723 so you can do some delicate amount of positive feedback when under load.
 

nsaspook

Joined Aug 27, 2009
13,079
The sense resistor must be directly connected to the output pin and regulation measured there. Any resistance on the order of 0.05Ω between the sense resistor and output pin will give you 10x worse regulation.
 

ScottWang

Joined Aug 23, 2012
7,397
The LM338 5A rating current, load regulation typical p=0.1%, normally it will be using less than I=5A*(1/3)=1.66A, when you used it for I=50W/12V=4.167A, if the heatsink is too small or without fan, I just thought that maybe the temperature could be over 60℃ (165.6℉).

Could take a photo for the sink and measure the surface temperature of LM338?
 

crutschow

Joined Mar 14, 2008
34,281
The sense resistor must be directly connected to the output pin and regulation measured there. Any resistance on the order of 0.05Ω between the sense resistor and output pin will give you 10x worse regulation.
True. (That's the output pin of the regulator).
But the common side of the sense resistor (R2) must be connected directly to the circuit breadboard output pin for best regulation.

Hamlet, where are you connecting the meter ground lead when you make the measurement?
It should be directly to the common side of the sense resistor, R2, otherwise any resistance in the ground lead can affect the measured voltage.
 

Dr.killjoy

Joined Apr 28, 2013
1,196
Have you tried adjusting the resistance to compensate for the load ?? ALso remember the voltage should be higher with no load and will drop with a load and stabilize the regulator since most voltage regulators need a load in order to work properly..
 

GopherT

Joined Nov 23, 2012
8,009
Have you tried adjusting the resistance to compensate for the load ?? ALso remember the voltage should be higher with no load and will drop with a load and stabilize the regulator since most voltage regulators need a load in order to work properly..
Good point DrKJ

@Hamlet
You cannot just go from no load to 50 watts and expect no voltage change. The minimum load, according to the datasheet, is 5mA. Add a 470 to 1.8k resistor between LM338 output and ground, then repeat.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
GopherT, I thought that R1 sets minimum load? Anyway, I added a 1k between output & ground,
took a measurement of 10.01v, added the load, and measured 9.78v. I have soldered the powerbrick
+ to the IN on the reg. I have socket terminals on the legs of the LM338. Maybe I need solder there too.

These are brief tests. Heatsink really shouldn't matter (75x75mm x7.5mm thick copper, with alum. fins, thermal grease.)

Someone asked about my power brick, it drops .5v on load. It's rated at 7.5A.

nsaspook & Crutschow said:
"The sense resistor must be directly connected to the output pin and regulation measured there. Any resistance on the order of 0.05Ω between the sense resistor and output pin will give you 10x worse regulation."
Well, these parts are breadboaded. Maybe I should solder them? This might be all the trouble, if so, that should disappear when everything
is soldered to the PCB final...?

I am not unhappy with the results, but I have been having something of an odyssey with regards
voltage regulation and power supplies. I have learned mucho, and at this stage, I don't want to overlook
anything. I have some questions about transformer sag, should that be a new thread?
 

#12

Joined Nov 30, 2010
18,224
Heatsink really shouldn't matter
This style of regulator has a "Safe Area of Operation" self protection circuit. Heat is one of the triggers. There is some possibility that heating is causing a partial response inside the chip. I have never gotten down to the nitty-gritty with three terminal regulators because I started designing before they existed. If I need 0.1% regulation, I know how to do it, but I don't know how to do it when everything is inside a chip I can't get at. AFIK, Only external (additional) circuits and Kelvin connections can achieve this quality with a three terminal regulator.

I assume the specs are measured under very controlled conditions, often so controlled that they are impractical in your average light bulb circuit.
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
The ground terminal of the meter was the collective terminal for all grounds. Nominally, it was clipped right to the bare spot on the ground lead from the power brick where it met my breadboard. All the grounds were nestled there together, bulb, powerbrick, R2.

#12 is right, I don't need perfect regulation with a lightbulb. However, I do want to try and understand what is happening. Might help me someday. Before I got things whittled down to a 2% deviation, I was seeing 10% & 20% drops! I got those fixed, so... I figure why not sort out how I can get below 1% (and call it good!)

My guess is that 1) R1 is heating slightly in the contacts of the breadboard, increasing resistance, but only with a large load on the output...
2) The socket on the input/adj. pins isn't good enough contact. Again, I might have to solder it.
 

crutschow

Joined Mar 14, 2008
34,281
At the risk of repeating myself, the meter ground lead needs to be connected directly to the R2 connection, not the circuit ground, since that point can differ in voltage due to ground currents.

Report back when you measure the voltage directly from the regulator output pin to directly at the R2 connection.
Only then will you have measured the actual regulator voltage under load.
Otherwise we may just be chasing a bogus problem. :rolleyes:
 

Thread Starter

Hamlet

Joined Jun 10, 2015
519
Excellent! Thank you crutschow & nsaspook and everyone else.
It was resistance in the breadboard contacts causing the trouble.
Problem disappears when sense resister is bonded to adj/output pins.
 

crutschow

Joined Mar 14, 2008
34,281
Glad you found the problem.
It's easy to miss the effect of circuit trace or wire resistance on circuit operation.
I once had a puzzling problem with a 16-bit DAC where the output voltage was deviating from spec by a few LSBs (which isn't much for a 16-bit DAC). This was a serious issue since the DAC was the primary signal source for a satellite test system calibration performed prior to each test.
Took me a while to track it down but it turns out the voltage shift was from the ground current varying a few mA, due to the varying digital control word, going through the small resistance of about two inches of a circuit trace.
Soldering a bare copper wire over the trace was the solution.
 
Top