Efficient Padding for UART to Minimize Power Consumption

Question

So, I am implementing a custom protocol over UART (data field in the packet is fixed at 8 bytes + contains a len field). If the data len is less than 8 bytes, 0xFF is being used for padding. I developed this idea of padding 0xFF from my experience with SPI NOR Flash memories.

However, from the POV of UART, won't this lead to additional power consumption? For example, if I am sending a single byte of data, for the next 7 bytes(garbage) the TX pin will be at logic HIGH.

Won't 0x00 for padding be a better option?

No hardware flow control or interface is being used, and this particular UART line will be connected via 0-ohm jumper resistor to an FTDI chip (FT232RL).

Answer 1

Generic logic pins today are CMOS type.

The MCU output will drive very close to GND/VCC, with a Thevenin equivalent resistance in the 30-70Ω range -- what range exactly depends on state (typically, PMOS is a little weaker), what the specified drive strength is (V_OH/V_OL at I_OH/I_OL) and manufacturing variation (the transistor size varies modestly from chip to chip).

The FTDI input pin will be essentially a capacitance; there are clamp diodes to limit excessive voltages (such as during ESD), then MOSFET gates. From the datasheet:
FT232R

The pin types are defined for these pins on p.10.

Voltage ratings, p.15. Specs continue on pp.15-17 showing characteristics at various supply voltages.

Unfortunately they do not give an equivalent pin circuit, and input leakage current is not specified, but note 3 is most relevant here.

Therefore a mostly-high (0xff) padding will draw the least current through the internal pull-up resistors.

After static current, there is dynamic current. When the pin changes state, capacitances are discharged through transistor resistance, and power dissipated. Typically the energy is constant per switching edge, and power / current consumption is proportional to rate. Eliminating transitions entirely is preferable -- using a variable-length packet for example -- but the next best is a value with few transitions, like 0x00 or 0xff. Again we prefer 0xff.

Most likely, the current draw due to pin changes is negligible in comparison to internal operating current draw by the MCU and FTDI themselves.

Answer 2

It is likely that there is no detectable difference regarding what you transmit from the UART point of view.

Sending 0xFF or 0x00 has equal amount of data transitions anyway.

Unless you know that your physical interface circuit consumes more current in some specific state, like if you have pull-ups or are driving optocoupler inputs.

Edit: Now that it is known that MCU TXD connects to FT232RL RXD, the FT232 has internal weak (200k) pull-ups to VCCIO supply to keep the line idle high when nothing is driving it, so sending 0x00 does consume more current than 0xFF, but it is consumed from the FTDI, not from the MCU, so in the end it may come from PC or whatever is powering the FTDI (it supports being powered from PC USB or locally).

However, if you really need to consider current consumption at this level, you might then want to look at the protocol itself because every byte you send will also consume some amount of current so why send 16-byte blocks if less is needed, and also all other hardware and software should be overhauled to consume the least possible amount of energy in all situations.

Or simply realize that in this case it may be irrelevant if you send 0x00 or 0xFF or any other character as the dummy idle filler character.