Initial commit

docs/power_path.md

@@ -0,0 +1,86 @@
+# Power Path
+
+## PD controller
+
+Based on the specifications, I've decided to use the the STUSB4500 as
+a PD controller instead of the TPS25730. Mostly due to the nice interface
+and the lack of ADC-based configuration pins. The TPS25730 has one thing that
+is nicer than the STUSB4500, the way it enables its sink path.
+
+The STUSB4500 has a pin called the *VBUS_EN_SNK* which is an open drain output
+and turns on (or rather the Transistor turns on) when a PD profile has been
+negotiated according to the configuration. This means that the way to switch
+on the sink path is by using P-channel FETS.
+
+The TPS25730 on the other hand, has an internal charge pump that can generate
+a voltage above the source (source or drain will only matter during turn on
+and turn off events, since the entire MOSFET is in series and meant to be
+driven in its ohmic region). This means that it uses N-channel FETS.
+
+The big advantage of N-channel FETS compared to P-channel FETS is that
+generally N-channel FETS are smaller and have a lower R_ds(on) compared
+to the P-channel FETS.
+
+Looking at the reference design of the STUSB4500 and the TPS25730 really
+shows the difference. The STUSB4500 design uses the STL6P3LLH6 P-channel
+MOSFETS which have a max R_ds(on) of 30milliOhm per FET. The CSD87501L
+Dual N-channel MOSFET used by the TPS25730 has a combined max R_ds(on) of 9.3
+milliohms.
+
+Comparing both FETS, the loss in the switch for each of the PD controllers
+at 5A (12V) is equal to:
+
+* 30 milliOhm * 2 = 60 milliOhm * 25(I squared) = 1.5 W
+* 9.3 milliOhm = 9.3 milliOhm * 25(I squared) = 0.233 W
+
+Although a loss of a quarter of a watt is awesomely low, I would be happy
+with a total loss of around a watt, or half a watt per MOSFET.
+
+To this end, the following specifications will be used to find a better
+MOSFET for the Power Path:
+
+1. Vds > 30V
+2. Vgs > 20V (I really don't want to use a different FET if I decide to
+make a 20V version).
+3. R_ds(on) max < 20 milliOhm
+4. Vgs(th) min = 1V
+5. Smallest possible footprint.
+
+### Sink Switch Gate Driver
+
+The reference design of the STUSB4500 (the one with the MCU) uses a few
+passives as a gatedriver before connecting to the *VBUS_EN_SNK* pin. 
+R10 is logical, since we want to make sure that when *VBUS_EN_SNK* is not
+turned on the voltage between the gate and the source is 0. The size of
+R10 is not. Looking at the reference design and assuming a 20 voltage drop,
+the current flowing through *VBUS_EN_SNK* is only 160 micro A, significantly
+lower than the 3 milli A used in the electrical characteristics. The resistor
+R11 is likely used to limit the maximum Vgs, sice the STL6P3LLH6 has a limit
+of around 20V, which would nearly be reached without it. The choice of
+22 Kilo Ohm is likely to create a larger safety margin, since above a Vgs of
+10 V, the R_ds(on) would be low enough.
+
+The reasoning behind R13 and C10 is not yet known. I will test it out
+with a simulator.
+
+### Discharge circuits
+
+The STUSB4500 has two discharge paths that are automatically triggered
+in case of a disconnect or some voltage error. Although really useful,
+both pins *VBUS_VS_Disch* and *Disch* are only triggered by the internal
+logic of the STUSB4500. There is no way, except playing around with the
+voltages, to trigger it yourself. On the *VBUS* side there is no problem
+with that, the only reason I would want to discharge is a change in PD voltage
+or a disconnect. However, at the system side, the rail needs to be discharged
+if a LED strip is unplugged, since after an unplug event a plug event may
+follow with a LED strip of a different voltage. Due to this, I will need to
+add my own discharge path parallel to the one in the reference design.
+
+A fun thing to notice is that on the VBUS side, the discharge path to
+*VBUS_VS_Disch* has a bunch of passives. The pin can only handle up to
+50 milli A. Which is why a 1 kiloOhm series resistance has been added.
+However, D4 and C4 are a bit strange. It would make sense that in case
+of discharging D4 "shorts out" the resistor R1, but wouldn't it be better
+to make R4 the same value as R1 allowing up to 40 milli A of current to
+flow (assuming a Vd of 0V). It may have something todo with the discharge
+and charge timing but I need to simulate it to be sure.