I would have placed U8 above or underneath T1 (on the same bottom layer to reduce the length of the HV trace. The cutout in the middle, as you pointed out it's purely cosmetic. However, I would have placed a cutout around pin 8, to prevent whiskers going to the neighboring pins.
For testing the transformer you should also consider measuring the primary leakage inductance. Short out the secondary coil and measure the primary at the intended operating freq. Having too much could kill the switch if the RCD clamp isn't tuned well. There is sooo much more that could be tested here. Hope to see a video on that soon.
We did this in another video where we looked at the design of the custom transformer and the clamp circuit. It came out to about 5%, which was actually less than I expected given it's an off-the-shelf core and bobbin.
Thank you for the anti-creepage slot. I would have taken it all the way through the capacitor and optocoupler (again, standard/best practices). Regarding the silkscreen, as a final step you can modify the footprints for individual components to meet your DRC requirements. In these cases, just breaking a few lines would do.
@@Zachariah-Peterson The board is manufacturable, but it fails a basic DRC? You put all that effort into getting to the one yard line. Push through for the touchdown!
@@Zachariah-Peterson That's fair. I've got more time than money, so I would have spent a few minutes fixing the silkscreen to avoid an engineering fee. Fell into that trap once, and the fee for a few bits of silkscreen over a board edge was almost as much as standard shipping.
HI Zach, I often see pcb layout recommendation saying that output cap of a psu should be placed close the output pin of the PSU. But wouldn't it be better to place an outpur cap closer to the ic's sinking power from it, to bypass series parasitic inductance?
If you think about it, there will be capacitance near both. For this module, it's small enough that "near the output connector" is close to "near the output inductor" so I do not expect a performance impact. When you look at the load, there is most likely some input capacitance coming into that load module or system, and then if the load is a digital IC it should already have some decoupling/bypass capacitance at the power pins.
@@Zachariah-Peterson I see, so I guess the purpose of having decaps close to the ouput is to help with ground bounce during the time the internal fets sink currents (switching off time) right?
@@AlbertRei3424 On the load side by a digital IC it is as you mention, on the power supply side the capacitance will set the startup rate and it will determine the level of ripple seen on the output.
can you tell us about how much $ were those transformers? asking mostly like how much was the core+former and how much was the coiling process since i never ordered stuff like this before.
If you're going to pay someone else to design it, wind it, and package it then it could be $50 per part, that was for winding and packaging in the US. If you buy the parts off the shelf and wind them yourself, it's much less expensive. I have used core and bobbin and done the winding myself and they work pretty well.
I'm in the middle of a PoE flyback design and this video comes in very useful, thanks !
I would have placed U8 above or underneath T1 (on the same bottom layer to reduce the length of the HV trace. The cutout in the middle, as you pointed out it's purely cosmetic. However, I would have placed a cutout around pin 8, to prevent whiskers going to the neighboring pins.
For testing the transformer you should also consider measuring the primary leakage inductance. Short out the secondary coil and measure the primary at the intended operating freq. Having too much could kill the switch if the RCD clamp isn't tuned well. There is sooo much more that could be tested here. Hope to see a video on that soon.
We did this in another video where we looked at the design of the custom transformer and the clamp circuit. It came out to about 5%, which was actually less than I expected given it's an off-the-shelf core and bobbin.
Thank you for the anti-creepage slot. I would have taken it all the way through the capacitor and optocoupler (again, standard/best practices).
Regarding the silkscreen, as a final step you can modify the footprints for individual components to meet your DRC requirements. In these cases, just breaking a few lines would do.
Ain't nobody got time to modify a few silkscreen lines just for a creepage slot!
@@Zachariah-Peterson The board is manufacturable, but it fails a basic DRC? You put all that effort into getting to the one yard line. Push through for the touchdown!
@@petersage5157 I know and I noticed afterwards, but in my head I'm thinking "let JLCPCB deal with it"
@@Zachariah-Peterson That's fair. I've got more time than money, so I would have spent a few minutes fixing the silkscreen to avoid an engineering fee. Fell into that trap once, and the fee for a few bits of silkscreen over a board edge was almost as much as standard shipping.
HI Zach,
I often see pcb layout recommendation saying that output cap of a psu should be placed close the output pin of the PSU.
But wouldn't it be better to place an outpur cap closer to the ic's sinking power from it, to bypass series parasitic inductance?
If you think about it, there will be capacitance near both. For this module, it's small enough that "near the output connector" is close to "near the output inductor" so I do not expect a performance impact. When you look at the load, there is most likely some input capacitance coming into that load module or system, and then if the load is a digital IC it should already have some decoupling/bypass capacitance at the power pins.
@@Zachariah-Peterson I see, so I guess the purpose of having decaps close to the ouput is to help with ground bounce during the time the internal fets sink currents (switching off time) right?
@@AlbertRei3424 On the load side by a digital IC it is as you mention, on the power supply side the capacitance will set the startup rate and it will determine the level of ripple seen on the output.
can you tell us about how much $ were those transformers? asking mostly like how much was the core+former and how much was the coiling process since i never ordered stuff like this before.
If you're going to pay someone else to design it, wind it, and package it then it could be $50 per part, that was for winding and packaging in the US. If you buy the parts off the shelf and wind them yourself, it's much less expensive. I have used core and bobbin and done the winding myself and they work pretty well.