Tuesday, February 6, 2018

Counterfeit LM2596 Regulator Boards

Recently Dick Benson, W1QG, mentioned to me that he had ordered 10 LM2596 adjustable voltage regulator boards from eBay for the incredibly low price of $10.99, including shipping.


Well, this seemed like a great deal to me, too!  So I also placed an order for 10 boards.

Dick received his boards while mine were still in transit.  While testing them, he discovered that the switching frequency was 50 KHz, not the 150 KHz specified for an LM2596 switching regulator.

Hmmm...could the regulators on these boards actually be LM2576 regulators that have been relabeled as LM2596 regulators?  Or (and probably more likely) are these regulators poorly implemented copies of the LM2576 regulator that have been relabeled as LM2596 parts and that might fail later in a disastrous way?  Without looking at the die itself and comparing it with the die of a true LM2576, there's no way for me to know.


Measurements, Original Board:

My boards arrived and I ran some tests of my own...

Below is one of the 10 boards in my test setup.  I configured the setup so that the output voltage would be 5V with an output current of 1 amp (i.e. 5 ohm load resistor).  The input voltage is 12V:

(Click on image to enlarge)

(The 10 watt, 5 ohm resistor will be dissipating 5 watts and thus gets very hot, so I placed it in a clay flower-pot saucer.)

Measurements:

    o Measured Input Current: 0.511A
    o Measured Output Voltage:  5.00V
    o Calculated Input Power:  12.0 * 0.511 = 6.13W
    o Calculated Output Power: (5.00 ^ 2)/5 = 5.0 W
    o Calculated Efficiency: 5.0/6.13 = 82%

The measured Output Ripple:

    o  Amplitude:  234 mVpp
    o  Frequency:   52.8 KHz  -- This part cannot be an LM2596!

See oscilloscope capture, below:

(Click on image to enlarge)


Measurements, Counterfeit regulator replaced with LM2596:

I replaced the fake LM2596 on one of the boards with a real LM2596:


Measurements:

    o Measured Input Current: 0.508A
    o Measured Output Voltage:  4.99V
    o Calculated Input Power:  12.0 * 0.508 = 6.1W
    o Calculated Output Power: (4.99 ^ 2)/5 = 4.98 W
    o Calculated Efficiency: 4.98/6.1 = 82%

The measured Output Ripple:

    o  Amplitude:  102 mVpp
    o  Frequency:  147 KHz



Dick, W1QG, had reported poor ripple performance on his boards due to high-ESR of the output caps.  I thought I'd do some experiments of my own with some capacitors I had on hand.

First...

Replace original 220uF, 35V electrolytic output cap with a 100uF, 10V tantalum cap:


Measurements:

    o Measured Input Current: 0.500A
    o Measured Output Voltage:  4.96V
    o Calculated Input Power:  12.0 * 0.500 = 6.0W
    o Calculated Output Power: (4.96 ^ 2)/5 = 4.92 W
    o Calculated Efficiency: 4.92/6.0 = 82%

The measured Output Ripple:

    o  Amplitude:  54.4 mVpp
    o  Frequency:  146 KHz



Next...

Replace the output cap with a 330uF, 10V, 100 milliohm ESR tantalum cap:


Measurements:

    o Measured Input Current: 0.501A
    o Measured Output Voltage:  4.96V
    o Calculated Input Power:  12.0 * 0.501 = 6.01 W
    o Calculated Output Power: (4.96 ^ 2)/5 = 4.92 W
    o Calculated Efficiency: 4.92/6.01 = 82%

The measured Output Ripple:

    o  Amplitude:  52.8 mVpp
    o  Frequency:  149 KHz  



Note that the ripple amplitude is essentially the same for the 100uF and 330uF tantalum capacitors!  This implies that the ripple current amplitude, for these caps, is probably due to ESR, not capacitance value.


And finally, I thought I'd try a through-hole cap...

Replace the output cap with a 220uF, 63V Elena electrolytic cap:


Measurements:

    o Measured Input Current: 0.501A
    o Measured Output Voltage:  4.96V
    o Calculated Input Power:  12.0 * 0.501 = 6.01 W
    o Calculated Output Power: (4.96 ^ 2)/5 = 4.92 W
    o Calculated Efficiency: 4.92/6.01 = 82%

The measured Output Ripple:

    o  Amplitude:  83.2 mVpp
    o  Frequency:  146 KHz  


Note that the ripple is worse!  


Replacing the potentiometer with a resistor:

Another modification I made was to replace the pot with a fixed resistor.  Pots fail, and when they fail they will usually fail open.  If this happens on this board, the regulator board's output voltage will go to its maximum value, probably frying circuitry downstream.

So I'd much rather replace the potentiometer time-bomb with a resistor!  Fortunately, this is easy to do.

If you remove the potentiometer, you will see under it pads for an SMD resistor.  Simply load the correct value of resistor onto these pads and, voila, you will have a fixed-voltage regulator board!

If you look at the LM2596 datasheet (section 9.2.2.2.2) and compare it to the regulator board, you will see that the datasheet's R1 is already stuffed (with 330 ohms in my case) and that the SMD pads under the potentiometer are for R2.

With an R1 of 330 ohms, then, for a 5V output, R2 should be 1K ohms.

Here's the board I modified to be 5V fixed output:


(This board has also been modified with the correct LM2596 regulator and a 100uF tantalum output cap).


W1QG Measurements:

Dick also ran his own tests on his boards.  Here are his results:

As I mentioned, the ESR of both the input and output caps is substantial.
  
The PCB I used was probably one of the worst, but I cannot confirm this.
       
This 220 uF @ 35V output cap has an ESR of 0.57 ohms, and its actual C is 173 uF (not shown):
       

 In all cases, the input voltage was 12V, and the output was set to 5V into a 5 ohm load.
      
This output cap was replaced with a tantalum rated at 220uF, 10 V:
       

 The ESR is 0.058 ohms which is about 1/10th that of the "stock" version. 

Also note that it's physical size is about 3 to 4X bigger with the voltage rating of only 10V.  Its capacitance was 208 uF. 

The stock cap was replaced with this tantalum, and another measurement made:

    
     
For the last measurement, the counterfeit IC  (LM2576) was replaced with the ON Semi LM2596.

Results:
      
        
Changing to the low ESR Tantalum provided a 13.9 dB reduction in the pp level.
         
Then changing the IC to a "real" LM2596 reduced the pp ripple by 21 dB from the original. 


Other Notes:

The LM2596 datasheet contains a wealth of information regarding component selection.  It is well worth the read.

In my tests the board itself was dissipating about 1 watt of power (for a 1 amp output at 5v and and input voltage of 12v).  The board gets noticeably warm!  If dissipating more power (and perhaps even at this power), I would recommend heat-sinking the board.  To accomplish this, I would scrap away the solder-mask on the back side of the board (under the LM2596's tab) and solder a heatsink to this area (copper would be a great choice for heatsink material).

Here's a schematic I've drawn of the boards that I received from eBay:


I checked the component values (but not voltage ratings) on one of the ten boards I received, and the values are all within 10 to 20% of the values shown on the schematic.

Specifically:

  o  The input 100uF cap measured to be 87uF, ESR of 0.66 ohms.

  o  The output 220uF cap measured to be 200uF, ESR of 0.28 ohms.

  o  The 100nF caps measured to be 117nF and 120nF.

  o  The 47uH inductor measured to be 45uH.

All measurements were made at 1 KHz using a GenRad 1657 Digibridge.



Standard Caveat:

I might have made a mistake in my code, designs, equations, schematics, models, etc.  If anything looks confusing or wrong to you, please feel free to comment below or send me an email.

Also, I will note:

This information is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

1 comment:

Unknown said...

I could be wrong but to have the same ripple output the correct coil for the fake chip should be 3 times larger than the one mounted in pairs with the original chip.
These unwary people took the TI / ONsemiconductor datasheet and copied the application scheme and then mounted the fake chip.