We have failure on Dielectric test between Power wire to Ethernet Port. Failure was observed near 1.7kV. After test Ethernet port was damaged (Phy chip).

See attached Case 1 :- Where we apply High voltage between Power to Ethernet terminal. (see below case 1 setup image)

enter image description here

The SMPS we used is rated for 3.75VAC for 1 Min. https://recom-power.com/pdf/Powerline_AC-DC/RAC10-C_277.pdf

Test video can be seen on below link. https://we.tl/t-ZcJQAZtdNo

We also tested CASE 2:- where we apply high voltage up to 3kV AC for 1 min between power wire and Secondary Ground. after test unit was ok, no failure observed in this case. (see below case 2 setup image) enter image description here

Question is when SMPS is rated till 3.75kVAC and test pass with CASE 2 setup than why its failure on CASE 1? Is it because of Ethernet transformer or something else? Where is it getting its path from?

Circuit is below enter image description here

  • \$\begingroup\$ What happens if you test between input of SMPS and the 5V output, instead of the ground? \$\endgroup\$
    – Joren Vaes
    Commented Jun 21, 2019 at 9:47
  • 1
    \$\begingroup\$ Never Tested yet. \$\endgroup\$
    – Bharav
    Commented Jun 21, 2019 at 10:05

1 Answer 1


simple fix.

Use HVDC at 2.5k to 3kVdc or as std requires for 1 second but with ~100k series resistors to achieve HV breakdown rating to make the test non-destructive,yet valid to UL/IEC CE HIPOT leakage standards.

This is a common Test Engineering problem due to lack of design awareness. The AC line filter Y caps conduct leakage currents such that even if the supply may pass , your loads obvious did not due to the CM cap impedance coupling your HVac to the load.

I saw a resistor array arcing. This also tells me your test is extremely destructive. It doesn’t have to be to detect dielectric breakdown.

A good Test Engineer will recognize that the solution is to limit the Hipot tester current above the failure criteria current but hopefully below a destructive threshold . The other solution is use rectified HVAC with a HVDC cap to reduce the ripple or use a tester designed for this and ramp up slowly. Then add HV rated current limiting resistors in series just above (by ~ 25% ) for your desired test failure threshold.


When I did this 20yrs ago Hipot failures were still detected but now non-destructive using the higher HVDC standard for 1.414 +10% so that the test could be done in 1 second, not draw high AC current , yet still test for HV insulation breakdown failures.

This will not only save you scrap costs but be able to speed up HIPOT testing. Your tester should display leakage current, so you use whatever slow ramp dV/dt

Often some poor AC/DC power supplies will fail when the floating secondary is earth grounded and with flyback frequencies, cause the leakage capacitance in the transformer effectively shunt some of the voltage across the transformer insulation and induce more stress on the primary side components to chassis or earth ground. In this case I requested the major OEM ( PowerOne) to improve their assembly and test process in Mexico and perform as I requested with DC earth grounded and then fix all the insulation gap issues after they transferred the production from San Diego, Cali. again the Test and Process Engineers lacked the design experience necessary to choose a worst case test process and test for margin to failure,

  • \$\begingroup\$ we have only High voltage tester , where minimum current setting is 50mA. also there is no Y caps in the system (we removed it already before test). there is only one cap inside the SMPS module( yes I open the SMPS module) has 1nF 1KV between Neutral and Secondary Ground. so at 3KV we can except =3KV/(2*pi*50*1nf) =666uA RMS leakage current. But is case 1, I saw one more cap of 0.1uf (10/100 megajack connector shown above) , is connected between ground and Ethernet transformer center tap. So at 1.8KV leakage wil be = 40mA RMS, may be due this Ethernet failed. \$\endgroup\$
    – Bharav
    Commented Jun 22, 2019 at 4:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.