LYT4217 design stability
I've designed a 35W (35V, 1A output) supply based on the DER350 design. I am using a LYT4127, removed dimming components, changed the reference resistor to 24.9K and the BP cap to 4.7uF. I am picking off the peak voltage sense from before the choke, as in the DER286 reference design.
The driver comes up at 66Khz, 30% duty cycle. The FB pin current is 188uA, just what the PI spreadsheet tool predicts with a 90K resistor for Rfb1. I think the driver is operating in skip mode. Increasing or decreasing Rfb1 causes the output current to go down.
If I put a 0.01uF decoupling cap between pins 3 and 5, the driver comes up at 132KHz but the output current is 2x what I designed for. If I increase Rfb1 to about 157Kohms the driver operates at 1A output. Efficiency is good and waveforms are good.
I am concerned that there is something unreliable in this design solution and that it will not be reliable over many units or over temperature swings.
Is it an expected problem that the LYT4217 might operate in skip mode if there is no decoupling on the FB pin?
Is it reasonable that Rfb1 needs to be changed so much from the predicted value to get the desired operation? Or, is this an indication of something wrong?
Comments
Removing the OVP circuit does not change the performance. I've partially mitigated the 66KHz startup problem with the 0.01uF bypass cap from pin 3 to 5. However, if I turn the circuit off and on within about 4 seconds, it starts up and stays at 66KHz. If I wait longer before turning back on, it runs normally at 132KHz. If I decrease the output filter cap, the waiting time gets longer. Output is 35.3V 1A, bias voltage is 20.6V, feedback resistance is 157K to yield 116uA FB current. Vfb is 2.3V, Vbp is 6.4V, Pin2 is 3V, current into pin 2 is about 85uA. The design spreadsheet indicates FB current should be much higher than this.
I've checked that Ifb is greater than Ifb(ar) before 50msec, so I dont think I'm stuck in the soft start sequence. Are there any other suggestions to look for?
Hi,
Can you send some waveforms to help in analyzing this concern? Schematic will help as well.
In the datasheet, startup frequency is 66kHz during soft-start period so I expect that after 100ms the switching frequency must be 132kHz..
I think I've beat the startup problem by increasing the capacitance at pin 3 to ground. I find that if the output cap voltage is above 15V at the time of turn-on, the driver apparently thinks the load is open (or over current?) and locks into operation at 66KHz. It does not reset to 132KHz even after an extended time. By increasing the cap on the Ifb signal, the driver will startup correctly even when the output cap is fully charged, like we will see in dips and interrupt testing. I'm up to 1.47uF. RC time constant of this is now 0.23s. I have to check if this harms any other performance metrics like power factor or total harmonic distortion.
Will you be able to provide me a copy of your working schematic? I just want to check if there's something else that we have overlooked. Please send me some waveforms as well that illustrate the problem.
Thanks.
Increasing or decreasing RFB1 will definitely cause the output current to go down or up, respectively. The device will go into skip-cycle mode if the feedback current exceeds 210uA typically.
Actual IFB value depends on the bias voltage and is calculated at (Vbias-2.4)/RFB. Please make sure that the actual IFB is less than 200uA to avoid skip cycle mode.
It is not necessary to put a decoupling capacitor on FB pin unless the trace is noisy. Still, this would not cause the operating frequency to go down to 66kHz nor will it cause the device to operate in skip cycle mode.
The spreadsheet value of RFB1 provides a good estimate of what value to use in order to get the desired output current. Actual bias voltage definitely affects RFB1. Please use FB pin resistor fine tuning. If the actual bias voltage is close to 20V as per spreadsheet, then if the predicted RFB1 is 90k and the output current is 2x as much, then there is something else causing the issue.
Please remove the OVP circuit that is connected to the FB pin and recheck. If the output current changed after you disconnect the OVP circuit, then this is the one causing the problem. You may send the schematic for further analysis if you still can't solve the problem.