BUG: When unrolling TSE sequence unblanking signal is incorrect

[schote]

• Cause in plotting the sequence can be excluded since the scope showed the same track

• Before combining digital and gradient signal, unblanking signal looks good

• [ ] Ensure that the sequence plot agrees to the scope

[schote]

Potential reasoning:

• The highest bit is used to encode the unblanking signal, but the highest bit in int16 also encodes the sign
• The y-gradient channel is used to store the unblanking signal and is the only channel which has a negative gradient
• Unblanking in the first echo-train is correct (there is no y-gradient), bug only occurs in the second echo train (y-gradient != 0)

[schote]

Minimal example to reproduce the root of the problem:

data = np.int16(6000) >> 1 | (0 << 15)
print(f"A: {np.int16(data << 1)}; D: {np.int16(data >> 15)}")
data = np.int16(6000) >> 1 | (1 << 15)
print(f"A: {np.int16(data << 1)}; D: {np.int16(data >> 15)}")
data = np.int16(-6000) >> 1 | (0 << 15)
print(f"A: {np.int16(data << 1)}; D: {np.int16(data >> 15)}")
data = np.int16(-6000) >> 1 | (1 << 15)
print(f"A: {np.int16(data << 1)}; D: {np.int16(data >> 15)}")

Output:

A: 6000; D: 0
A: 6000; D: 1
A: -6000; D: -1
A: -6000; D: -1

[schote]

Use the following workaround: np.uint16(value) >> 1 | digital Transform the int16 value to uint16 before shifting. Since the manipulate the bitorder anyway (by using 15th bit as digital and 14th bit as new MSB of analog value) the value is not a true int16 value anymore