I'm dealing with an application in which a DC-DC converter is used to transform an input voltage of 15V DC into lower voltages. The current configuration makes use of an LM27222 with two mosfets in push-pull configuration. The output voltage is linearized into an LC filter.

Here's a picture of the general schematic used.


The LM27222 is driven by a PWM signal with a carrier frequency of 65 KHz. The mosfets used are of type FDMS86101DC. The logic level voltage is 5V. The output current to design the system is 10A. The load is resistive. The system works fine, but some concerns have been raised to avoid collisions between the two mosfets (both transistors ON), especially during start-up/turn-off sequence. This happened a couple of times during testing of new firmwares, but, most probably, the problem was related to the testing engineer which not so carefully "touched" the output leads generating a destructive shortcut.

According to one of the new engineers we have been working with, we should:

  • Substitute the LM27222 with another driver, the L6388E, which should be much more stable than the LM27222 and avoid any change of collision, even in transition sequences.
  • Add a diode in series with the low-side mosfet. This diode should add additional protection during transient
  • Add a complex filter to the Vin line. The filter should have a common-mode and a differential mode inductor, together with a large electrolytic capacitor and several small ceramic capacitors. According to the engineer this filter should limit the demand of pulsed power from the power supply, and result in a more "compliance-friendly" design to meet the EMC (electromagnetic compatibility) regulations. The Vin comes from a AC-DC desktop power adapter and he is concerned that the connection cable from the power supply to the electronic card could radiate frequencies because of the required pulsed power. Besides, since several of these power supplies got broken during tests, the engineer is also suspecting that we are throwing some important noise back into the power supply, leading to failures.

Here's a picture of the L6388E internal schematic.


I would like to ask two questions:

  1. Is it necessary to carry out all the steps suggested by the engineer? The change of LM27222 is particularly painful and I would avoid that if it does not give clear advantages.
  2. If we decide to keep the current configuration with the LM27222, how can we drive the LEN signal to avoid problems at start-up and turn-off of the system?
  • 4
    \$\begingroup\$ Do not add a diode in series! That does not make any sense. If you add a diode you will loose the advantage of synchronous rectifier and performance will be even worse than the plain buck regulator schematic. \$\endgroup\$ – Todor Simeonov Jun 21 '17 at 10:58
  • 1
    \$\begingroup\$ I definitely agree with Todor Simeonov, a diode would violate the whole concept of a synchronous rectifier. The recommended design from the datasheet is a good reference point and they never have such a diode in there. \$\endgroup\$ – nickagian Jun 21 '17 at 11:33
  • \$\begingroup\$ Good point there. I actually was very doubtful about it \$\endgroup\$ – Francesco Jun 21 '17 at 11:40
  • \$\begingroup\$ How do you actually use the LEN signal at the moment? And why do you say the problems occur during start-up or shut-down? Have you observed something in this direction or it's just an assumption? \$\endgroup\$ – nickagian Jun 21 '17 at 12:14
  • \$\begingroup\$ @nickagian. We keep LEN LOW at startup. We wait circa 100 milliseconds to allow the system to power up completely. Then we keep LEN HIGH and after 100 milliseconds we start giving the PWM signal to the IN pin to control the converter. Maybe we should wait some more time than 100 milliseconds at start-up? \$\endgroup\$ – Francesco Jun 21 '17 at 12:42

The L6388E doesn't look at all that good in the data sheet for your application. For instance (and this worries me) they say recommended max operating frequency is 400 kHz and that shoot through dead-time is nominally 320 ns. That means dead-time can be 13% of the overall switching cycle. They don't appear to recommend nominal operating conditions and neither do they offer a "typical" circuit in their data sheet.

Minimum Vcc for the part appears to be 9.1 volts and this well below the 4 volts to 6.85 volts in your default schematic.

Adding the diode in series with the lower MOSFET is a bizarre thing to recommend and defeats the efficiency gained by using a synchronization MOSFET.

  • \$\begingroup\$ Thank you. Could you recommend a proper design for an input filter for the DC-DC converter? Would a standard LC suffice or should we look into more complex filters? For us it would be needed to filter out most of the EMI and avoid to radiate through the power cable. Not sure how "dirty" can be the input line of such a DC converter with 65 KHz driving frequency... \$\endgroup\$ – Francesco Jun 21 '17 at 12:01
  • \$\begingroup\$ @Francesco it's guesswork and sometimes good layout can pretty much solve bad layout problems. The downside of an LC low pass filter is that load dumps can and will ripple through the chip and cause a transient doubling of the input voltage thus destroying the chip. It's tricky without something to go on so I have to be reluctant to answer in all but general terms. Think of the input filter as an RC low pass filter and then put the inductor in parallel with the resistor to obtain low DC volt drops hence design an RC filter but add the inductor as a means to give dc current a direct path. \$\endgroup\$ – Andy aka Jun 21 '17 at 12:09

The first line in the LM27222 datasheet says

Adaptive Shoot-through Protection, 10ns dead time

...and at the bottom of page 6

ADAPTIVE SHOOT-THROUGH PROTECTION The LM27222 prevents shoot-through power loss by ensuring that both the high- and low-side MOSFETs are not conducting at the same time.

(more detailed explanations follow)

I would attribute your blown FETs to a scope probe slipping... it happens... (happened to me once).

Blown FETs can have many other causes, for example bad layout:

  • Voltage spikes on switching (high supply inductance, bad decoupling...)
  • FET oscillation (too much grid inductance, long traces...)
  • Chip going nuts (lack of decoupling, etc)

The diode in series to the low-side MOSFET is not recommended. The whole idea of having a synchronous converter is to improve the efficiency when compared to a converter that has a diode instead of the MOSFET. The diode would contradict with this concept.

Regarding the input filter, if you do need to be compliant with the EMC regulations I think you will be confronted with this issue some time in the future. A common-mode and differential mode filter is recommended in this case. What I usually do is to use a common mode choke between the input and the converter. Then, after the chocke, I use a 10nF capacitor between the two input lines and also a 10nF capacitor between each input line and GND. Also before the choke, I put the same 10nF capacitor between the lines and additionally an RC network (10Ohm, 10nF) between each line and GND. And this is my filter for the EMC and it helps to block noise in both directions. Of course a bunch of decoupling bulk capacitors are also needed.

Here is a schematic of the input stage (the circuit part around the two transistors is used to limit the inrush current during start-up): enter image description here

Now, whether the absence of such a filter has something to do with your defective power supplies is hard to say. But consider this: If you are creating a short-circuit by allowing the two MOSFETs to conduct at the same time, this short-circuit is also applied to Vin! And it could happen that such a case is destructive for the external power supply.

  • \$\begingroup\$ Thanks. Could you please post a schematic of the input filter you were explaining? \$\endgroup\$ – Francesco Jun 21 '17 at 12:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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