Skip to content

Troubleshooting

Symptoms, causes and fixes, cheapest check first.

Datasheet sections this page relies on

From Texas Instruments SBASB22 -:

  • Reset: 7.4.1 Reset, p. 49

All of them, with the driver API for each: Datasheet map.

EnsureInitialized returns Bus

The port reported a failure: either EnsureInitialized() returned false or a transfer did.

  1. Check that your port's EnsureInitialized() returns true on its own (log it).
  2. Check the bus pins in your port against the board.
  3. With a logic analyser, check that the first frame (a write to GEN_CFG1, the software reset) appears on the bus as three bytes with CS low for all of them.

EnsureInitialized returns DeviceIdMismatch

DEVICE_ID did not read 0x0004, or BANK_SEL did not read as the driver left it. The chip answered with something else, or nothing answered.

Read value Likely cause Fix
0x0000 SDOUT not connected, or CS not reaching the chip Check SDOUT and CS continuity
0xFFFF SDOUT floating high, or the chip unpowered Check the three supplies and that SDOUT is not pulled up hard
Anything else Wrong SPI mode, or a shifted bit Use mode 0; check that CS stays low for the whole 24-bit frame and that the clock is at most 29 MHz (kMaxSdoutClockHz; 60 MHz is the table limit for writes only)

EnsureInitialized returns Timeout

CALIB_BUSY did not clear after the reset. Check the three supplies, that RESET is high (or released by your GpioSet() with GpioSignal::INACTIVE), and the reference: with ReferenceSource::External the chip needs a 4.096 V source on REFIO.

EnsureInitialized returns Unsupported

Config::hardware_reset is set but the port has no GpioSet(), or the configuration asks for the daisy chain (out of scope: Limitations).

EnsureInitialized returns InvalidArgument

Validate() rejected the configuration and nothing was sent. Call ads9324::Validate(config) on the host to see which rule: the combinations are listed on Configuration.

ReadFrame returns DataMismatch

The frame is all zeros or all ones, which is what a stuck or floating SDOUT reads. A real conversion of an input at a code limit is never all 0x00 or all 0xFF across 16 channels.

  1. Check SDOUT, CS and the SPI mode.
  2. Check that a conversion happened: pulse CONVST (or use MeasureOnce()) before ReadFrame().
  3. Run RunDataPathSelfTest(): it uses test patterns, so it works with the inputs unconnected.
  4. When you are sure the frame is a real one, Sampling::reject_stuck_frames = false turns the check off.

A channel reads another channel's value

The AIN9 to AIN16 registers are in reverse order; the driver maps them, so a swap points at the wiring or at the frame layout. Check the channel count and group (ChannelSelect): with 8 channels the frame carries AIN1 to AIN8 or AIN9 to AIN16 only, and Frame::present_mask says which. RunDataPathSelfTest() returns DataMismatch when a channel returns another channel's pattern.

Reads return UnexpectedReset

The DIAG byte or RESET_DETECT_FLAG reported a reset: the chip reset since the driver started it, usually a brown-out on one of the supplies. The driver is uninitialised again; EnsureInitialized() writes the configuration again. If it repeats, check the supplies and their decoupling. Wake() returns it too when the registers did not survive standby.

A reading sits at a code limit

Frame::clipped_mask has the channel's bit set: the input is outside the range. Choose a larger InputRange, or check the input for a missing divider. An open input pair reads near zero code only in the open-wire safe type.

The voltage is wrong by a fixed factor

microvolts is computed from the channel's configured range; if you changed a channel's range outside the driver, the arithmetic uses the old one. Use SetInput(), and GetConfig().inputs[n] to see what the driver holds. Offset and gain errors of the board go into Input::offset_quarter_lsb and Input::gain.

The window alert never fires, or fires at once

The thresholds compare the top byte of the code (two's complement), not volts: ±40 on the top byte is ±40 x 256 codes. The flags are latched: ReadWindowFlags() shows what happened since the last ClearWindowFlags(). A window cannot be used with offset binary data (Validate() refuses it).

RunAdcCalibration returns Refused or Timeout

Refused: a digital filter is on (switch it off with SetFilter() first) or the chip is in standby. Timeout: CALIB_BUSY stayed set for 150 ms; check the supplies, the reference and, for the free-running mode, that CONVST really runs.

MeasureOnce returns Timeout

With ReadySource::Pin the DRDY pin did not assert within Sampling::ready_timeout_us: check that Alarm::function is Drdy, the polarity and the wiring, and that CONVST really pulses. Check that your port's DelayUs() really waits: a delay that returns at once makes every timeout far too short.