0
\$\begingroup\$

This is my first board with AC.

I'm planning on using it to power an ESP32. Overall the consumption will be much less than 0.3A.

I'm putting two HiLink HLK-PM01 units in parallel, with 0.3A output each.

I also include a 0.5A glass fuse in the middle and a couple of capacitors 16V, 2200uF that I happen to have in stock.

The fuse is a fast blow 250V 0.5A glass fuse from Amazon.

The idea is that even if one of the two HiLink units fail, the other one will keep working.

I will also have a 3A fuse on the plug that will connect this to mains. The PSU + ESP32 unit will be in a plastic junction box.

Does it look safe to use? Will I burn my house down, if I leave this unattended? Can I leave this running for months (it's an IOT device after all)?

enter image description here

enter image description here

enter image description here

\$\endgroup\$
9
  • 3
    \$\begingroup\$ It doesn't look safe without you specifying creepage and clearance distances. \$\endgroup\$
    – Andy aka
    Commented Feb 7 at 16:07
  • 1
    \$\begingroup\$ You cannot parallel constant voltage converters in this way. You have built a short circuit \$\endgroup\$
    – tobalt
    Commented Feb 7 at 18:32
  • 1
    \$\begingroup\$ Since you will be installing your own design into a building you own, will the device be approved by safety agencies so it will be a legal device to install, and are you legally allowed to work on residential building installations so you are e.g. an electrician with enough certificates? If not, better not install it and leave it unattended, and make it a non-installable with a mains cord. \$\endgroup\$
    – Justme
    Commented Feb 7 at 18:48
  • 2
    \$\begingroup\$ The circuit could be made on a single-sided board with a little re-routing. No nasty mains on top while probing around. \$\endgroup\$
    – Transistor
    Commented Feb 7 at 18:50
  • 3
    \$\begingroup\$ Do those HiLink devices have any meaningful safety ratings? Tons and tons of unsafe products are sold on websites similar to Amazon, and people have died from using them. I never use a power supply unless it's UL listed, UL recognized, or ETL listed. (If a product says that it "meets UL standards," that is not the same thing as being UL listed!) \$\endgroup\$ Commented Feb 8 at 2:38

4 Answers 4

15
\$\begingroup\$

The idea is that even if one of the two HiLink units fail, the other one will keep working.

Power supplies do fail catastrophically by blowing the input fuse. So you'll need a fuse per each supply. They can also fail by shorting the output, or may start up at different rates and get damaged by back-fed current from the faster supply. You'll need isolating diodes on the outputs of the power supplies.

At a minimum, I'd change the circuit as follows:

schematic

simulate this circuit – Schematic created using CircuitLab

16V, 2200uF

C1 and C2 can be fairly small, around 100μF will be plenty unless the power supply datasheet specifies a larger capacitor. C3 can be a higher value like the 2200μF you got.

The fuse is a fast blow 250V 0.5A glass fuse from Amazon.

Don't buy critical components from anyone but mainstream authorized distributors. Throw those fuses away immediately, and buy fuses from DigiKey, Mouser, Newark, Allied Electronics, RS Components, ELFA/Distrelec. When selecting parts on the distributor website, exclude "Marketplace" parts that come from third parties and not from the distributor.

The supplies are rated for 0.1A input max, so a 125mA slow-blow fuse per supply will do the job. A 0.5A fuse is too large even for two suppliess.

I'm putting two HiLink HLK-PM01 units [from Amazon] in parallel

There may be nothing wrong with those supplies, but if you want something you depend on, and real product support in case things go wrong, do yourself a favor and don't buy components from Amazon. I would not use them in a project meant to run unsupervised and/or in anything with a plastic enclosure (vs. a metal one).

Hopefully you're still in Amazon's return window and can send those supplies and fuses back. Then head to DigiKey etc.

\$\endgroup\$
6
\$\begingroup\$

Kuba beat me to commenting on the direct-paralleled outputs, so on to other things.

Overall, you have relatively fat traces for the low current AC input, and thin traces for the higher current DC output. Consider reversing this. Neither your input nor output currents are very high, but in general, trace widths are determined by the signal/power current, and trace spacing is determined by the signal/power voltage.

It looks like connector J1 is reversed. The second line probably is a polarization back for a common pin header. With the way it is shown, the connecting wires come off in a direction over the middle of the board. Of course, without knowing the connector part number this is guessing. This might not be a problem, depending on the wiring plan for the installed board.

Also about J1, consider using a 3-pin header with the center pin removed. This is a common way to meet creepage distance rules with small connectors.

Left side of board, blue trace to the left of F1. Move it to the left so it is centered between the F1 and J1 pins. This is a better distribution of the available creepage distance.

Right side of board, diagonal blue trace to the right of the F1 pin. This is where the creepage distance is most important, because the isolated output is closest to the AC input. Move the blue trace another tenth of an inch to the right. It can be very close to the other 5 V conductors, but must be at least 4 mm from the AC conductors. More is better.

\$\endgroup\$
1
\$\begingroup\$

If you are determined to use your design and components:

The circuit schematic needs some improvement, as in Kuba's answer, I'd add some diodes to ensure there is no possible way for the power supplies outputs to feed back into the other, as well as independently fuse them.

For the PCB though that is definitely not very safe. There should ideally be an air gap between all live and neutral traces, with > 5mm spacing between the traces themselves, and they should definitely not run on top of each other. If a trace burns the PCB below carbonises and becomes conductive, and you'll have a dead short between live and neutral, which is unprotected, once plasma bridges between the two layers. Adding a spark gap after the fuses, would also help with other unexpected surges (eg lightning).

There is another much safer approach entirely though:

If the objective is simply having a redundant power supply and it is not intended as a commercial product. You could design a much safer, DIY friendly (possibly cheaper too) dual input USB-C to screw terminal board, still using the diodes on each 5v line from the USB-C sockets. Then you can find 2 identical USB chargers, which are UL listed, easy to find, and can deal with the high voltage side far better, as well as not voiding your insurance if it causes a fire.

Edit: Removed reference to 2-1 USB-C cables, as they may be directional and even if not, it would rely on on the supplies having feedback protection on their outputs, which is the point of the diodes mentioned above.

\$\endgroup\$
1
  • \$\begingroup\$ +1, it’s almost always better to buy a complete power supply instead of messing around with it yourself. \$\endgroup\$
    – Michael
    Commented Feb 8 at 14:04
1
\$\begingroup\$

If it has to fit inside a junction box it will be very tight so I don't understand why you want to use two cheapo power supplies while it would be cheaper and smaller to use one from a known reputable manufacturer... Examples:

RECOM POWER RAC03-3.3SK

MEAN WELL IRM-02-3.3

Both have MTBF > 1 million hours.

The fuse has to be slow-blow, because the fast-blow fuse will do its job and blow at power-up due to the current spike for charging the power supply's internal capacitor.

Most of these power supplies datasheets recommend a MOV on the input to catch voltage spikes, so you can indeed add one.

2200µF is not necessary on the output, a good quality 100-470µF cap is enough, also smaller.

If you use a ESP32 module with USB and a 5V input, it's a good idea to replace the AMS1117 LDO with one that actually works, like LDL1117. Also replace the output cap with 10µF MLCC instead of tantalum. This way your ESP32 will not randomly crash when coming out of sleep. Another solution is to use a 3V3 supply directly.

I won't comment on issues highlighted by others, so here's another one: mechanical design! How are you going to mount the ESP32 board and the power supply board together? Since the board is mains powered it's important to have it well mounted and secured in the enclosure. And since there is no OFF switch, if you want to pull out the ESP32, for example to reprogram it via USB if the wifi upload somehow fails... you also have to think about that. Both boards should not be able to move around and create a contact with the live traces, and live traces should be protected from fingers too.

For example you can use a shrouded fuse holder which is encased in plastic, which means pretty much all the live metal on your board will be at the back, which is safe if it is mounted against the enclosure.

If you use metal board standoffs, mind the distance between them and your live traces: on your board they are quite close. This is important if you use mezzanine standoffs to mount your ESP32 board on top.

If your enclosure has screw holes for mounting the board, make sure you print your pcb on paper and check it fits and the holes align.

On your board, changing the fuse would need needle nose pliers as it is hidden between the two supplies.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.