I'm designing an Open Hardware door-lock-controller for our hackerspace, to these basic requirements:
- Be made exclusively from available and inexpensive parts.
- Read external Wiegand keyboard
- Read external Wiegand RFID reader
- Control various door-locks and other actuators.
- Read 125 kHz EM4100 RFID directly with an onboard reader.
- Passive POE powered, to make cabling easier.
- Run off unregulated 15-30V DC (cheap and easy high-capacity UPS via SLA)
- Provide a regulated 12V DC out @ 300 mA to 1.5A peak for external gear.
- Talk Ethernet to the server.
- Keep a local list of allowable keys, so only power is needed to let us in.
- Control external 230V gear using two relays.
This is the second spin of the PCB with a few changes to fix the problems I've found, the biggest change I've made is to throw away the LDO that regulated the 3.3V rail, in exchange for a simple MC34063 based switcher.
I'm pretty sure it's going to work, but I'd like some comments to check if I've overlooked something.
Compromises I've had to make in the design are:
- The primary (35V) capacitor is stacked on top of the Ethernet transformer to save space.
- The Ethernet transformer is huge, but I have a bunch of them and they were cheap, so I'm going to use them, even if they are huge.
Main questions
- Is there a cheaper and better way to implement the two power supplies (I need around 200mA@3.3V and 300mA@12V with peaks to 1.5A) ?
- Does the MC34063 switcher circuit/layout look sane?
- I have no idea how inefficient the MC34063 is so I've tried to heat sink it as well as possible using the copper I have available, any hints for improvements?
Best Answer
I'd be really surprised if this operates, at least with any reliability. You desperately need a power plane and a ground plane with components like these.
For instance the ENC28J60SS is a current mode driver, those traces from it to the pulse transformer to the jack are going to be carrying high currents switching at high speeds with no reasonable supply or ground path for them. I'd also be concerned with the crystal as its tuning capacitors have a very long high inductance path to ground. I would guess that this design will emit a ton of EMI, possibly enough to interfere with any nearby electronics.
The layout for the switcher is also huge, i would expect you to have serious amounts of noise and supply ringing throughout this design coming from the switcher and the ethernet components.
You also mentioned POE but there is no POE controller on this board. If your planning to just plug into a POE switch there is a negotiation procedure to request power (in various levels). The only way you could power this off the ethernet port as is would be to find a mid run injector that just dumped voltage on the unused pairs in 10/100 Base-T. If you plugged this into a gigabit switch that could cause major problems. Those pairs also aren't center tapped as they should be for POE. Dumping pins 7/8 directly to ground could cause major problems for both this device and the switch.
There is a lot i could go into but 90% of it is a result of not having power/ground planes, clear that up first.
Also the mid point of R5 and R6 should not be connected to the center tap of the pulse transformer and the jack side of the Output transformer should have a small, high voltage cap to chassis ground at its center tap