LTC4267 isolated 5V will not start

Hello,

Can you confirm that C5 or its equivalent is either removed or modified so that it connects between VPORTP and PGND? My best guess is that this design was intended for use with PSEs that support legacy detection, i.e. use the presence of a bulk capacitor as indication a PD is connected. This technique is not IEEE 802.3-compliant, but was more common twenty years ago in the early days of PoE. IEEE 802.3-compliant PSEs expect between 50nF to 120nF of port capacitance, or just C1 on DN361. Otherwise, can you share a waveform of the detection attempt and what type of PSE/Injector you are using?

On a side note, if you need more than 13W there are now several options available without requiring the battery circuit in DN361.

Best Regards,

Eric

Eric, many thanks. That was most helpful. Removing C5 brought the board to life. It also alerted me to another problem, so not quite out of the woods yet! Now, the LTC overheats and R7 (100R) burns out. I'm not used to this - I've been designing products with PoE for years and normally it's very straightforward. I'm beginning to think there's some kind of pcb fault...nonetheless, thanks indeed for your reply.

R7 overheating indicates there is too much power dissipation in the primary snubber. I recommend checking out LT3748 datasheet, page 15 for a guide on flyback snubber design. You can also try adding C5 back into the circuit, but downstream of the hotswap MOSFET (VPORTP to PGND) as the DC-DC converter likely requires some bulk input capacitance.

You can also refer to the DC2046A-F for a more recent 5V / 13W PD design. DC2046A-F supports up to 5V @ 2.3A with its efficient topology, and the output current is scalable with power stage modifications. See DC2046A-B (5Vout @ 25.5W) and DC2046A-E (5Vout @ 40W).

Best Regards,

Eric

Hi Eric, thanks for your suggestion. The part I don't understand is why this design is so unstable. On powering up, R7 overheats and burns out in a matter of seconds (along with the MOSFET). There isn't any time to perform tests on the current snubber network. I could switch to the LTC4276 but we're at the pre-production stage in this product and this would involve considerable extra time and expense.

Success! It turns out that the opto-coupler had been fitted the wrong way round. This alone did not solve the problem; I've also had to raise the snubber capacitor to at least 270p. Now I can measure the switching interval and adjust the snubber accordingly. It was your two suggestions that got me on the right track and I'm extremely grateful to you for your assistance. Many many thanks.

A final twist - the snubber resistor R7 on DN461 is an order of magnitude out; it should be 1K, and best stability without excessive power dissipation is achieved with C3 in the range 330-470pF for the PA1134 transformer specified. With R3 set to 100R, there's unresolveable overshoot on Q1 drain.