LNK306 Opto Feedback failure ?
I am using an LNK306 with opto feedback to generate 24VDC. I used the basic High Side Buck design described in the LNK data sheet so there is no current limiting resistor in series with the input of the opto. The design has been through many hours of envrionmental testing (heat, humidity, transient etc...) without failure. We have deployed several systems for field testing and in one unit the 24VDC power supply failed. It appears as though the opto may have failed causing the power supply to go out of regulation which triggered a clamping diode and movistor (located in another device). Is the opto more prone to failure without the series resistor ?
After doing some research and contemplation perhaps the output of the power supply is not as stable as I thought. I did do some transient load testing but never really measured the loop stability. Is it probable that without the current limiting resistor the gain is so high as to cause oscillation ? I am using an FOD817D opto. The output of the supply has a 470uF Electrolytic paralled with a 0.1uF ceramic.
Any help would be greatly appreciated.
Comments
I did this design a couple of years ago. At the time I was having transient load issues. Out of desperation I placed a 1nF capacitor across the BP and FP pins and that fixed the problem. However I did not understand why. PI-Wesley responded that the extra capcitor prevented pulse grouping. Somewhere along the way that 1nF beacame 100nF. Now I'm wondering if that is causing a problem. I have been doing additional transient load testing and have not seen any problems so far. Any insights ?
I used PXI Designer for the original design then modified it for CCM and opto feedback.
PI-Wesley recommeded an opto with at least a CTR of 300% which the FOD817D has.
Can you send me the schematic and layout please?
Here are the files you requested
I haven't completely ruled out an external transient voltage or a momentary short having occured on the output of the power supply. The supply powers an external Tilt Sensor that is immediately adjacent to the power supply's enclosure. The enclosure of the Tilt Sensor is not grounded to the system and the pins that the power supply connect to are exposed. However there is a cover that encloses the Tilt Sensor but it is not grounded either. The reason for all this folating metal is a requirement by the customer (the FAA) that all metalic surfaces be anodized or in some places powder coated. I guess I should explain that the power supply is part of a CCA that resides in a Light Housing Assembly that normally sits at the end of a runway. The failed power supply/CCA was mounted in an LHA that was installed between some buildings for test purposes.
Is there any chance I might be able to get a PI Engineer to comment on my original post of June 7 (LNK306 Feedback failure)
?
Hi dlmuns.
Can you tell me if the only component that failed was the optocoupler? Did the LNKswitch fail as well, or any other component?
I do think that it would be wise to put a resistor in series with the optocoupler. The gain is very high although in this type of controller I don't think it would be causing instability (not impossible though since the output voltage is pretty high). The issue I see is that during a fast load transient (going from 100% load to 0%) the voltage on the 24V bus will want to go up, which may put stress on the diode of the opto because the current is not really limited. This is maybe not an issue in normal operation and it is a very short spike, but it may become important to limit the current during external transients, and it may become a reliability issue in the long run.
I know you mention you don't have issues with the 100nF capacitor between BP and FB, and at first sight I don't see any major problems with it either, as long as you check that there are no large overshoots/undershoots during load transients, and no excessive output ripple during normal operation.
Hi PI-Wesley and thanks for your quick response. Unfortunately my failure analysis approach was not as systematic as it should have been. I replaced the LNK306 first. Still no output. After some probing I discovered the 28V Transzorb (D39 on the schematic) was shorted. this is a 400 watt (10/1000uS) device. I removed it and U2 (a POL DC/DC converter). Applied power (should have applied it slowly) and the Zener reference diode (D4) smoked which indicated to me that the opto was not working. I did not have a scope attached so I don't know what was happening as the power came up. Replaced the opto, zener, R2 (zener bias diode) and for good measure D5 (catch diode) although it was probably ok. Powered up and everything worked as expected. I'm guessing the opto failed, the output spiked, the transzorb clamped but not before some other circuitry downstream was damaged. Because it takes 50mS for the LNK306 to detect an open circuit, the transzorb overheated and shorted. We are currently running a long term reliability test (7000 hours) on a system. We are at 5000 hours and there have been no failures. The external device that is connected to this 24VDC power supply is a Tilt Switch which is essentially a linear regulator, constant current source and tilt sensor.
If you have any other comments please let me know. My project manager has decided that unless there is a similiar failure in our long term reliability system, we will leave the design as is. Thanks for your help. In any future designs I will be using a current limiting resistor in the opto input.
Did you use a PI Expert recommended design? The PC817D has a high gain as compared to the A, B or C variant, which might be forcing more current than necessary into the FB pin.