Skip to content

Ambiguities

Where the datasheet is unclear or contradicts itself, the reading the driver takes, and how the tests cover the other one.

Forty-nine questions came up while reading the datasheet (revision 1.18). Each has a decision, and the decision is tested against the model. A knob is a setting of the chip model in tests/mock_tmc5160.hpp (Knobs) or of the bus (Faults) that makes the model behave the other way, so a test can run the driver against each reading. Where a row says "guard", the decision is the driver's alone (a limit it enforces) and the chip has nothing to vary; the row names the test. Nothing here was checked on a chip: confirming each reading on hardware is open work.

Id Question Decision Knob or test
A1 SPI clock polarity and phase Mode 3 (page 25 says "usually referred to as SPI MODE 3") Faults::spi_mode: the model frames mode 3 only
A2 Maximum SCK 4 MHz internal clock, fCLK / 2 less 10 percent external Faults::sck_hz
A3 Status byte age One transfer old; never acted on in the same call Knobs::status_lag
A4 Reading a write-only register Never read; Refused; the shadow answers Knobs::wo_read_mode (zeros, last value, garbage)
A5 Write echo Used only by Verify::Echo Knobs::echo_writes
A6 Microstep table alignment Bit n is the increment into entry n Knobs::lut_literal_reading
A7 MSCURACT field order As documented, sign-extend 9 bits Knobs::mscuract_swap
A8 Reset defaults depend on OTP Documented defaults; OTP_READ decoded, never written Knobs::otp_byte
A9 Break-before-make write hazard Written only with the driver off Knobs::bbm_hazard
A10 BBMTIME maximum, 375 or 400 ns The larger number in the guard guard: bridge tests
A11 S2G level, 6 or 12 above 52 V 12 to 15 when the highest supply exceeds 52 V guard: current tests
A12 multistep_filt default Always written explicitly Knobs::multistep_reset
A13 GSTAT after power-up Cleared at start-up Knobs::gstat_reset_value
A14 en_pwm_mode or en_spread en_pwm_mode naming only
A15 Threshold ordering TPWMTHRS at least TCOOLTHRS at least THIGH, non-strict guard: core tests
A16 Hysteresis sum Guard on the effective sum guard: core tests
A17 PWM_REG 0 1 to 15 enforced with autoscale guard: core tests
A18 Freewheel conditions Only with IHOLD 0 guard: core tests
A19 GLOBALSCALER 1 to 31 Never planned, rejected guard: current tests
A20 Full-scale voltage and the page 74 table The formula with 325 mV and a 5 percent margin Knobs::vfs_mv
A21 Ramp limits Register widths plus the VMAX limit guard: motion tests
A22 Trajectory exactness The model is behavioural; only invariants are checked none: no exact reading to vary
A23 TSTEP formula 2^24 * USC / (256 * V) units tests
A24 RAMP_STAT clear mask 0x10CC Knobs::ramp_stat_wc
A25 DIAG semantics per SD_MODE Per-mode fields in Diag guard: core tests
A26 Direct mode exit Hold, clear, then XTARGET = XACTUAL Knobs::direct_mode_exit
A27 VDCMIN range 23 bits, bits 7..0 ignored Knobs::vdcmin_range
A28 External clock 4 to 18 MHz accepted guard: core tests
A29 VERSION of the A version 0x30 accepted, other values DeviceIdMismatch Knobs::version_byte
A30 IFCNT in SPI Never used on SPI Knobs::ifcnt_spi
A31 Node 254 Allowed with multi_node Knobs::node_254
A32 UART reply timeout SENDDELAY plus 64 bit times Knobs::uart_reply_delay_bits
A33 UART recovery 12 bit times Knobs::uart_recovery_bits
A34 Daisy-chain order Index 0 on MOSI Knobs::chain_order
A35 QFN pin count Recorded as QFN-WA with a note none: documentation
A36 OTSELECT meaning A table in the guard guard: core tests
A37 StealthChop autotune conditions The A version; a longer autotune leaves it unfinished Knobs::at_a_version
A38 StallGuard values Raw values over 10 bits; the model scripts SG_RESULT SetSgResult(): test A38
A39 Reset indication on DIAG0 The application's choice none: documented
A40 OTP programming Out of scope; never written model write counter forbidden_writes: test A40 and A41
A41 FACTORY_CONF Never written model write counters
A42 Strap verification Checked at start-up Knobs::sd_mode
A43 Decimal encoder constant, negatives The page 107 rule Knobs::enc_decimal_sign
A44 DcStep rules DC_TIME above tBLANK in clocks for the chosen TBL (16, 24, 36, 54), toff 3 to 15, full-step and constant off time guard: test A44 flips TBL
A45 Revision and TMC5130 TMC5160 and TMC5160A only none: scope
A47 Sense resistor range 1 to 10000 mOhm accepted if the plan fits the rating guard: test A47
A48 Soft and hard stop with StealthChop An acknowledgement is needed for the switch stop (Stop() kinds are ramps) guard: switch configuration tests
A49 X_COMPARE A one-update pulse on the DIAG pin Knobs::xcompare_pulse: test A49 flips it
A50 Supply range 10 to 55 V (the cover says 8 to 60 V) guard: supply_range test
A51 TZEROWAIT and a new mode The model ends a running pause on any RAMPMODE write; page 83 c only says the CPU may end the motion during the pause with option a or b model reading: tests T252, T257, T268
A52 TZEROWAIT outside position mode The model pauses after a position-mode ramp ends in standstill, at its target or short of it (page 41: the wait after ramping down to zero velocity before the next movement or direction inversion); velocity mode has no pause model reading: T262 sweeps TZEROWAIT 0, T276 the pause after an arrival
A53 TZEROWAIT after a hard switch stop The model's hard switch stop has no pause, though page 43 says a switch stop uses TZEROWAIT before a restart model reading: no test depends on it
A54 Lowering VMAX in position mode The model follows a lower VMAX (VMAX 0 included) with A1 and AMAX; DMAX, D1 and VSTOP only when it slows for the target or a reversal (page 84, joystick control) model reading: T276, T283
A55 VSTOP in velocity mode The model ignores VSTOP in velocity mode and runs down to 0 at AMAX (pages 83, 84: VSTOP belongs to the stop at the target) model reading: T274, T283
A56 A velocity-mode start from rest The model starts at VSTART, as in position mode; the datasheet does not say model reading: T272

Readings with no model knob: A22 (the model is behavioural, so only invariants are checked), A35, A39 (the application's choice), A41 (write counters), A45 (scope) and A51 to A56 (the model's fixed ramp readings). The model does not set RAMP_STAT.second_move (bit 12) on a run-back; the driver never reads that bit. The rows that name a guard are run against both sides of the limit in the named test. Ids are not renumbered: A46 was never used and is dropped.