Lowside Opto Design doesn't regulate

Posted by: sdame on 08/27/2009

I followed the low-side Buck Optofeedback design recommended on p6 of the LinkSwitch-TN application note (LNK306), but am having difficulty on my initial testing of the circuit.

The attached schematic shows an AC connection, but for testing purposes, I'm just connecting to a variable 30VDC supply and increasing the voltage to see it work. I can get to about 15 V input and then the LNK306 comes alive and the drain starts switching at 66KHz. But the output will not regulate. It tends to just track the input downwards until I increase the input voltage to about -22V and then the output just shuts down.

Any idea where I might have gone wrong. I can see that the part is trying to work, it just doesn't regulate to the 5V I expected.

I think I have connected my schematic exactly how it is shown. The Optocoupler I'm using is the SHARP PC817XG.

Comments

Although I don't have the specification your power supply, I think you may consider the following issues:
1) Our datasheet specified DRAIN Supply Voltage 50 V. The input voltage you are testing with is too low. The minimum DC input voltage should be ≥70 V when you design/test with LNK-TN.
2) Correct Diode Selection – UF400x series diodes are recommended only for designs that operate in MDCM at an ambient of 70 °C or below. For designs operating in continuous conduction mode (CCM) and/or higher ambients,
then a diode with a reverse recovery time of 35 ns or better, such as the BYV26C, is recommended.
3) Maximum drain current – verify that the peak drain current is below the data sheet peak drain specifi cation under worst-case conditions of highest line voltage, maximum overload (just prior to auto-restart) and highest ambient
temperature.
4) Thermal check – at maximum output power, minimum input voltage and maximum ambient temperature, verify that the LinkSwitch-TN SOURCE pin temperature is
100 °C or below.

OK, I guess I was trying to sneak up on the design test before I just apply full 115VAC to the circuit. I already came across one issue which is the input diode rectification needs a bias resistor to cause the caps to bias toward the Neutral side. So I plan to put a 1K resistor in parallel with D4 to ensure the CAPs will bias to Neutral.

Also, I think I should have moved L3 to the negative side of the caps since that is where the filtering will be most effective (on the negative half wave rectified line side.

But those concerns were secondary. I was just ramping the voltage up slowly to see what would happen when I apply (up to) a negative 30V (which is the limit of my DC bench supply) to the input (VIN-). What I notice is that the Drain does start switching for the circuit as it sits around -15V. However the +5 to VGND is not stable at all and I can get (not solid) voltages in excess of 10V-15V across the R1 resistor as the input is ramping up. I plan to connect a microprocessor to this rail, so that means the process would see transient voltages in excess of the maximum rail they are supposed to see.

Is this really the way this is designed to work? i.e. you don't get a stable output unless you pop the voltage up way above (below actually) -50V? I would seem if it works this way, that the output should not start switching until there is a safe operating voltage established. Then the output should be held at 5V according to the Zener and Opto arrangement I have.

I literally just copied this circuit right from the datasheet, so I'm wondering what I'm missing.

I will try the diodes suggested for the production design, but for the moment, I'm just at room temperature. The Triac in my circuit will cause circuit heating, so that is a good idea.

The output is intended to supply up to 150mA at -5V (measured from the +5V net to VGND).

I'm thinking I must have something wrong with the design in the way I connected it. Or perhaps there is a problem with this Opto selection? It is a little hard to know with such a feedback servo type design where I've gone wrong.

I will try connecting it to the line just to see what happens and report back.

Following up on my original design, I was able to get things running somewhat, but now I'm trying to root out the remaining noise issues in the design. I would like to have a rock solid -5V supply and am close. I discovered there is a load dependent 3KHz noise and a separate recovery or tracking noise that is following the half-wave ripple. I have plenty of input voltage over over -150VDC with about an 8V sawtooth wave ripple. The negative going slope of this ripple couples into the output as a tracking problem.

Do you have any idea how we get rid of this tracking issue?

Also, I have been able to apply a brute force output capacitance of 200uF to help remove the 3KHz spikes, but if you have additional ideas on how to remove that remaining 3Khz noise it would be appreciated. I'm less concerned about the 3KHz noise spikes and more concerned with getting rid of the large 1V anomaly that effects our performance.