Many years ago, when I was responsible for the incomming quality of a major US corporation, I was encountering many failures of an off-line bu;k linear supply. After much investigation, I gixovered the the AC Hipot test was causing failures within the supplies. All off-line supplies, linear or switching must pass the hipot test. Here the both AC inputs are shorted, and the all the output teminals are shorted together. No ptoblem - right. NOT.

In the AC Hipot test a 50-60 AC voltage is applied to the AC input. The current is measured on the output terminals. During the AC test, the capacitive coupling of the transformer (plus any primary gnd to secondary ground stabilization capacitor) createds an AC current acoss the dielectric boundary. This, then induces an AC voltage drop across any device in series from the secondary to the output terminal(s). This stresses the protection diodes on any IC, or any low voltage MOSFET, which could lead to falure,
The DC Hipot test uses a DC voltage equivalent to the AC peak voltage. By using DC, the AC current over the transformer capacitanves are eliminated ant the low voltage secondary semiconductors are not stressed. One caution, the ramp-up and down periods must be longer than 1 second, otherwise the dv/dt on the AC line will cause transient currents in the secondary.

Marty, from the magnetics perspective we have always seen the AC voltage as a more destructive test and try to keep hipot tests to a minimum.

Please remember there is more to it than the ac current that flows according to the capacitance. You also get corona current flowing, and the test process itself is physically destructive to the device. I believe that is what Mr. Garcia is talking about in his comment. Don't overtest - you just create more damage.

From a circuit point of view, this gets worse when you put the magnetic in the circuit and surround it with the sensitive electronics parts and then do the hi-pot testing. I'm not sure - should this even be done? Look for advice from the seasoned experts out there.

Medical equipment requires HiPot testing. In order to pass international and US/Canadian standards, we tested at 1500 volts AC for 1 minute from Line & Neutral tied together to chassis. There were other HiPot tests that we performed to get regulatory approval but not on a production basis.

Early on, I designed and built a device that would perform HiPot testing, ground resistance testing and leakage current testing which was approved by UL. This was years ago when UL544 was the standard for medical and dental.

I believe Hipot tests are different depending on grounded/ungrounded (2 wire and 3 wire power cords) designs. I suspect it is also different for different categories of products.

I can see that line/neutral tied together on oone side and the output tied to the other side of the hipot to be more problematic that the line/neutral to input ground test.

I have never had to do a 2 wire product so I have not had an issue.

I'm technical support engineer on power supplies manufacturing company and when We are troubleshooting some hipot failures tested with DC, sometimes isn't easy to found where is arcing with DC applied, but with AC apllied....You can see the arc..in other words is more fast found the issue using Hipot AC....

Years ago UL required some medical equipment to pass 1500VAC and some 2500VAC. I believe the higher voltage was for medical equipment connected where something like a catheter could ground the heart. AC was used and was important because low level current at 50 or 60Hz may cause fibrillation.
Also if the parts operate in a high voltage environment there must be no corona as it is usually destructive and will ultimately cause damage to the equipment.
The High Voltage test is often a workmanship test as it is difficult to verify the workmanship of a potted high voltage module.

If your power supply cannot handle the AC hipot test because of currents flowing from stray capacitance, I would worry about sending it out into the real world.

We have found recently, since the introduction of new 3rd Edition medical standard, a lot of confusion has been generated with customers as to the requirements of AC hi-pot. Some customers are demanding 4kVac Hi-pot test in finished product i.e. without removal of Y-cap, transformers etc., whereas before type-testing was accepted.
Thus, we had good reason to perform some experiments.
It is critical that the balance between input Y-caps to chassis, and output caps to chassis is considered. If the chassis voltage becomes too high, then you may have to worry about clearance/clearance to ALL parts of your circuit, and not just primary. Also, ceramic capacitor value varies with voltage applied across it (some interesting maths!).
Secondly, and I have to agree with Ray's comments, the corona discharge under AC voltage application becomes a major determining factor especially with voltages above 2500VAC. Where meeting the DC requirement was difficult, it was manageable with due consideration to distances and capacitor values.
However, the AC testing was a whole different ball game. Arcing occurred in multiple areas, and not always the same area/component. And as previously mentioned, once breakdown occurs once, you may as well throw the PSU in the bin (as regards hi-pot tests).
The general consensus was to overcome the AC hi-pot, and to maintain a high density PSU design, some type of conformal coating or potting compound (starve the oxygen from the arc) would need to be used.

A few of things to watch for. The dielectric loss for solid insulation changes with frequency. So the loss at low frequency (DC) is different from the loss at higher frequency (AC 50Hz). This means that the power loss in a solid dielectric for a given electric field is different for DC and AC. If this loss is too high it will destroy the soild insulation. The difference in loss from DC to AC test or vice versa can be enough to permanently destroy the insulation.

The temperature coefficient of the dielectric loss for DC applied electric field is often positive and the temperature coefficient of the dielectric loss for AC applied electric field is often negative. This means that a DC test can, with certain insulating materials, cause hot spots in the insulation via thermal runaway and thereby do damage. As a result some solid insulation materials cannot be tested with DC and are not suitable for DC insulation to the same electric field levels as they can manage for AC fields.

Finally the application of sustained high voltage DC to an insulating material can cause a polarization in the breakdown strength. That is after the application of the high DC voltage the insulator will have a better withstand to one polarity than the other.

Hi pot testing has the potential to do lots of damage that might not be immediately obvious. Care is required.

Hipot testing is mandatory in production for all AC power products.
It is always useful for any insulated products where susceptibility of the components and safety of the user is at stake.  

Make sure your Hipot Test is NON-DESTRUCTIVE

Most important is to limit the discharge current with a resistor in the probe or interface test cable supplying the Hipot Voltage. If the tester has a filter cap then the discharge current with ionized breakdown can damage parts in the device under test (DUT) A current limiter controlled power supply tester is not the same. Choose your worst case short circuit current to be slightly higher than your acceptance criteria for uA using a controlled ramp rate such as 500/V ~1kV/s ( remember AC line filter caps are designed for <500uA max for Hipot AC leakage test are affected by Ramp rate. )  

When Dielectric insulation failure occurs, you do not want to induce secondary damage to the device. It may also wound devices as bad as ESD.

Fix your Test Process

I once used a famous American brand AC Power Supply in a rack mounted product that began to show Hipot test failures in production and also damaged the power supply.  So I changed the Test Process to include a current limiting resistor in the tester so that if arc occurred the storage capacitor energy (in the Slaughter or Hippotronics Hipot Tester) would be current limited just above the acceptance criteria by 50%.  The tester was capable of discharge of high current perhaps more than an ESD tester, which is a different test altogether.  

Test the product, reject the failures and offer Root Cause suggestions

Then I identified all the root causes of arcing under low light conditions and asked the supplier to fix their design AND their Test Process to ensure the DC output was grounded if they could not measure leakage current on DC output with their Hipot tester on AC input.  ( Since an Ohmeter is low resistance, that is the same as grounding the isolated DC output and measuring input current for Hipot stress)

Although it was difficult for me to convince the supplier was at fault since they measured input current and not output current on the DC. We were both measuring input current but my design had the DC common grounded to AC ground, so in effect I was measuring output current induced by transformer capacitive coupling and insulation capacitance coupling where the highest voltage would be dropped across the lowest capacitive coupling to ground and these gaps and any components connected to them were at risk. After PS Failure rates reached 10% with at least 6 root causes that occurred after the supplier apparently moved production to Mexico, I was able to reject faulty insulation breakdowns without damaging the components in the PS.

The changes were subtle like the gaps and material used for securing heavy leaded parts from moving and were overlooked because of the supplier's test process was not measuring DC leakage current or with grounded secondary.

Susceptibility is normally simulating AC source disturbance such as induced lightning surge or simiilar from fault-breaker opening on large currents with large inductance lline transformers.  But also there is a risk of ESD induced failures and worse ESD induced triggering of AC insulation failures with follow heavy currents during breakdown of the AC power source.  Normally ESD testing up 5 or 15 kV is done during the design phase or DVT or pre-production verification or PVT and not for Production testing.  Ensure you do this on every product according to IEC standards to "qualify" the design and any significant design changes affected by ESD.