Philmore PS123 Power Supply (5): Counterfeit regulators from China!

I completed the final wiring of the new controller board and tested it out on the bench.  It ran well with no load so I mounted the the board and the LM338 regulator into the cabinet. The photo below shows the control board connected to the LM338 on the left.  It was being driven at 18V by my bench supply, and produced 13.8V output with no load. 

click to enlarge

I wanted to be able to operate at 13.8V @ 2 amps continuously.  For testing I bought a 15 ohm 10W resistor and two 4 ohm 20W resistors which I put in series to make an 8 ohm 20W resistor.  The resulting currents at 13.8V should be 0.92A  and 1.73A respectively.  

After I installed the LM338 and the board in the cabinet and connected the output of the rectifier, it again ran well with no load.  The transformer is a nominal 18VAC secondary, as marked on the transformer itself.  This should give a peak voltage under no load of 25.4VDC on the filter capacitors at no load.  The actual measured voltage was more like 28V however, so the house line voltage is probably a little higher than nominal.  I loaded the supply with the 15 ohm resistor and got 13.5V @ 0.9A.  The input voltage dropped to 24.6V.    When I tried the 8 ohm resistor however, the LM338 died.  I checked my schematic and discovered that I'd wired a protection diode wrong, I assumed this was the reason the LM338 died. 

I corrected the protection diode mistake and replaced the LM338, and began further testing.  Again the unit worked well with the 15 ohm load, and again the 8 ohm load killed the LM338.   I made some measurement between terminals of the LM338, both on a good unit, and on a blown unit.  It appears that the voltage input terminal had opened up in both blown units.  I suspected that the bond wires from the LM338 input pin to the die had fused open. 

I did various test, and destroyed five LM338s in total.   Below is a scope photo of a failure.  In this event the supply is connected to an 8 ohm resistor.  The supply is cycled on, and the voltages at the LM338 are monitored.  The yellow trace is the input voltage.  You can see the voltage step from zero and increase to a steady state as it charges the 9900 uF filter capacitor. You can also see ripple peaks from individual rectified AC peaks. The blue trace is the output voltage of the LM338.  The output voltage rises with the input charging. Once the input voltage gets above the set voltage plus the dropout voltage, then the output voltage settles at 13.8 volts or so.  The cyan trace shows the Adjust pin of the LM338. When functioning properly, this voltage should be 1.25V lower than the 13.8 volts, which appears to be the case. 

But note that about 150 ms after the supply turns on the output drops to 0V.  This must be the instant when the bond wire inside the LM338 burns itself open.  The Adjust pin voltage is tied to the output through a voltage divider, so it falls to 0V too.   Once the output current disappears, the input voltage is no longer loaded, and the input voltage soars from about loaded 23.5V to unloaded 27V or so.  

click to enlarge

Doing some investigations on the web, I discovered that counterfeit LM338s are very common, and that is apparently what I've purchased.    Real LM338s should be able pass 5A continuously, and up to 12A transiently. This is apparently lower current unit (maybe an 3A LM317) that has insufficient bonding between the die and the pins. 

At any rate, I will not be able to use the LM338s that I bought and will have to probably redesign the board.   Very disappointing!