Created power supply using LNK306 - Expecting 5.0V output... getting 10.4V output instead

Hey guys,

I used the LinkSwitch-TN LNK306 by Power Integrations to create a 5.0V power rail on my PCB. (120VAC input). I used the datasheet to come up with the component values, and looked at other example circuits on the net that also used this IC.

The problem is that I am getting a 10.4V output instead of the desired 5.0V. I double checked the feedback resistors, and they are 2K and 3.84K which should generate the desired 5.0V output. I did an experiment and swapped out the 3.84K resistor with another 2K resistor and nothing changed. I also wanted to note that I have a 3.3V linear voltage regulator and a microcontroller connected up to the 5V power, so there should be a small current load on the circuit.

I am at a complete loss on how I should go about troubleshooting this circuit. I keep going over the circuit schematic, layout, and chosen components, and cannot find anything wrong. Any help would greatly be appreciated. I attached my circuit schematic, and the PCB layout of the circuit for reference.

Here are the part #'s I am using for the Inductors, Diodes, and Capacitors for reference as well:

D1, D3: DFLU1400-7
D2: US1JHE3/61T
D4: FM4007W-W
L1, L2: 46105C
C4: GRM219R61E106KA12D
C6, C7: 475CKE450MLN
C8: 35ZLH100MEFC6.3X11
R3: WP2S-7R5JA25

I don't know 100% but:

The diodes you are using are "general purpose". This is good for 50/60hz power line use. (the first two diodes are good)
But the last two diodes are too slow for switching power supplies. (60,000hz) Probably you should use UF4007. (ultra fast) You need parts that say how many nS it takes to work. So I believe the voltage across C4 is wrong because these diodes are 100s of times too slow. This will effect your feedback.

I know the UF4007 is through hole but you can find a "fast" 1A diode. See it switches in 75nS.

Welcome to ETO!

Jattus said:

so there should be a small current load on the circuit.

Click to expand...

Perhaps it's not enough?
Are you aware that, since there is no galvanic isolation from the mains supply, any point of your circuit could become 'hot', and possibly LETHAL, if there were a component failure?

alec_t said:

Perhaps it's not enough?

Click to expand...

With the slow diodes this thing is going to act strange. (not regulate)

alec_t said:

no .... isolation .............................. possibly LETHAL

Click to expand...

If the power supply is plugged in backwards it will put the power line on your "GND". The micro and all electronics connected to it, must be in a box isolated from humans! You can not have censers going outside the box. Buttons must be double insulated. Be careful.
-------------------------------------------------------------
Sorry I may be wrong. I can not see the words on the schematic. I do not know what diodes you are using. I now see you are using three different types of diodes.
The two diodes at the power line can be general purpose.
The two diodes in the switcher must be fast.

Perhaps it's not enough?

Click to expand...

Probably right several examples show a 5mA load resistor.
------------------------------------------
The voltage across C8 (output voltage) should be the same as the voltage across C4. (or close) If not the feed back will not work! Be careful measuring C4 voltage. It should be 5V but not near ground. It will be riding on top of a 60,000hz switching waveform. (high voltage)

ronsimpson said:

I don't know 100% but:

The diodes you are using are "general purpose". This is good for 50/60hz power line use. (the first two diodes are good)
But the last two diodes are too slow for switching power supplies. (60,000hz) Probably you should use UF4007. (ultra fast) You need parts that say how many nS it takes to work. So I believe the voltage across C4 is wrong because these diodes are 100s of times too slow. This will effect your feedback.

I know the UF4007 is through hole but you can find a "fast" 1A diode. See it switches in 75nS.
View attachment 98510

Click to expand...

Thanks a lot for the reply!
I specifically chose fast diodes for the job. The input diodes (D2 & D4) are general purpose diodes, and I am under the impression that these are fine here. The important diodes are D1 & D3. The datasheet mentions that these need to be ultra-fast diodes, and should have a recovery time of less than 35nS. Because of this, I chose "DFLU1400-7" which is 25nS recovery time.

Shouldn't I be good here as far as the diodes are concerned?

alec_t said:

Welcome to ETO!

Perhaps it's not enough?
Are you aware that, since there is no galvanic isolation from the mains supply, any point of your circuit could become 'hot', and possibly LETHAL, if there were a component failure?

Click to expand...

I am thank you. This board is inside an enclosure and is 100% isolated. There are no buttons or anything for a person to come into contact with.