Skip to content

Latest commit

 

History

History
150 lines (121 loc) · 6.24 KB

pin_configuration.md

File metadata and controls

150 lines (121 loc) · 6.24 KB
Raw

Pin Configuration

The pin configuration determines the electrical behaviour of the I2C pins.

This document describes which settings I've chosen and why. It shows which settings are involved and contains the results of some experiments.

The settings are hard coded into the driver. See PAD_CONTROL_CONFIG in imx_rt1060_i2c_driver.cpp for the latest settings.

Table of Contents

Drive Strength

The drive strength determines the fall time of the I2C lines. This is defined as tfall in the I2C Specification.

I tested various drive strength settings to see the effect on the fall time. See the Appendix for the results.

Settings

Set the drive strength register (DSE) to:

  • DSE(2) - Drive Strength Enable
    • minimises undershoot spikes
    • small increase in VOL with large pullup resistors
    • DSE(1) would maximise tfall
      • but I believe the I2C requirement is designed to minimise noise and DSE(2) does this
  • SPEED(0) - Speed
    • SPEED doesn't seem to make much difference but the docs say that lower values
      • produce less noise
      • reduce current consumption
  • SRE(0) - Slew Rate Enable
    • we don't need to increase the slew rate
    • documentation advises leaving it disabled if possible

Pullup Configuration

Enable the 22 kΩ internal pullups by default. These are suitable for a 100 kHz bus with 1 or 2 devices all of which have internal pullups.

User can choose to use 47 kΩ or 100 kΩ internal pullups instead although these are too weak to be useful on their own.

User can disable the internal pullups to use external pullups only.

See I2CDriver::set_internal_pullups() and I2CDriverWire::setInternalPullups().

To Use Internal Pull-Up

  • PUS(3) - Pull Up Select
    • selects 22k internal pullup
    • this allows a 100k bus to work without external pullups in ideal circumstances
    • the other options are too weak at any speed
    • set to 2 to use 100 kΩ and 1 to use 47 kΩ internal pullups.
  • PUE(1) - Pull / Keep Select
    • must be 1 to enable internal pullup
  • PKE(1) - Pull / Keep Enable
    • must be 1 to enable internal pullup

To Use External Pull-ups Only

This disables the internal pull-ups.

Set:

  • PUS(0)
  • PUE(0)
  • PKE(0)
    • Could also use PKE(1) but it doesn't seem to add any benefit

Open Drain Enable

I2C is an open drain system so this setting is required. Apparently removing will cause a short if one device pulls high when the other pulls low.

  • ODE(1)

Hysteresis

Significantly reduces the chance that the device interprets noise as an edge. The I2C Specification requires hysteresis (Vhys) to be 0.05 Vdd (165 mV) or longer in Fast-mode and Fast-mode Plus. We know that the Teensy tends to generate rare glitches without it.

4.2 HYS of AN5078.pdf says that the hysteresis is 0.25 V. This meets the I2C Specification requirement.

See 12.4.2.1.1 Schmitt trigger for the datasheet's description of hysteresis.

Set:

  • HYS(1) - Enable hysteresis
Item Calculation Value
I2C Requirement at least 165 mV
Hysteresis 250 mV 250 mV
LOW to HIGH trigger voltage 0.5 Vdd + 0.125 V 1.775 V
HIGH to LOW trigger voltage 0.5 Vdd - 0.125 V 1.525 V

Slew Rate Enable

Setting it to 1 would increase the fall time which is already too fast.

  • SRE(0)

Appendix - Affect of Drive Strength on Fall Time (tfall)

According to the I2C Specification, tfall must have a minimum value of 12 ns at 3.3 V.

The Teensy in loopback is much faster than that. It seems to vary between 4 and 8 ns. It's a hard to be sure as the actual value is approximately the same as my oscilloscope's resolution. These figures are similar to the ones given in section 7.2 Slew rate (SRE) of AN5078.pdf.

I searched the web and concluded that:

  • fall times below 10ns are common
  • the I2C spec figure is probably designed to reduce electrical noise such as undershoot spikes

SPEED, DSE and SRE are intended to control the "strength" of the pin. I tried a lot of combinations of SPEED and DSE. If DSE is set to 1 then the fall time is approx 8 ns; otherwise, it's 4 or 5 ns. SPEED seems to have little effect although values 0 and 1 may be marginally slower.

I changed the default values to DSE(2) and SPEED(0). This still doesn't meet the spec, but it seems to reduce noise spikes. (Compared to the previous values of DES(4) and SPEED(1)).

This table shows:

  • DSE value
  • size of falling edges undershoot spike with 22k pullup
  • LOW voltage (VOL) with 1k external pullup
DSE tfall Undershoot VOL
1 8 ns -180 mV 290 mV
2 5 ns -150 mV 130 mV
3 5 ns -160 mV 90 mV
4 5 ns -270 mV 70 mV

Adding a series resistor increases tfall but the effect is very minor. For example, 100Ω increases fall time by a couple of nanoseconds but increases VOL1 to 500 mV with a 1k pull up which exceeds the maximum allowed in the spec.