# Behaviour of the Force Bits in Register 0x09F of AD9361

Question asked by DanI on Jul 28, 2014
Latest reply on Jul 30, 2014 by tlili

In the 9361 Register Map document, the descriptions for Registers 0x8E through 0x9D (Tx quad cal correction words for phase, gain, Ioffsets, Qoffsets) contain two statements in relation with the two possible values of their corresponding Force Bits in register 0x09F.

1. 1. If force bits in 0x09F are clear and Tx Quad Cal has completed, then registers 08E...09D hold the TxQuadcal correction words for phase, gain, Ioffsets, Qoffsets.
2. 2. If force bits in 0x09F are set then the values written to registers 08E...09D are used as correction words for phase, gain, Ioffsets, Qoffsets.

Note that these two statements not mutually exclusive.

Experimenting with 0x09F produces paradoxal results. It appears that the TxQuadcal correction words for phase, gain, Ioffsets, Qoffsets change after the force bits are set in register 0x9F.  Also it appears that after a non-converging TxQuad cal, by simply writing the force bits, the sideband rejection and LO rejection are much better than after a convergingTxQuad cal. How is that possible?

If the force bits in 0x09F are clear and Tx Quad Cal has completed, which values are effectively used as correction words for phase, gain, Ioffsets, Qoffsets?

If the force bits in 0x09F are set, which values are written to registers 08E...09D, by who are these values written, and at what moment are they written?

Descrition of the experiment (Output 1 used only)

Run TxQuadcal with RxNCOphase=0 (expect no convergence)

Read TxLO and TxSSB convergence bits in register A7, A8 [D1:D0] = 00, LO and SSB not converging

Start Tx, measure sideband rejection and LO rejection: poor rejection (10-20dB)

Since all bits in 0x09F are clear and Tx Quad Cal has completed, then according to the Register Map document, these registers hold the TxQuadcal correction words for phase, gain, Ioffsets, Qoffsets for Output 1

Read registers 08e...095 and store their values.

Now write register 0x09F = 0xFF

Since all bits in 0x09F are set, then according to the Register Map document, the values written to registers 08e...095 are used as correction words for phase, gain, Ioffsets, Qoffsets for Output 1.

Read registers 08e...095 and store their values. TxQuadcal correction words for phase, gain, Ioffsets, Qoffsets for Output 1 are now DIFFERENT from the readings before writing force bits. Question: where did these new values come from?

Measure SB and LO: now GOOD rejection (50-60dB) , DIFFERENT from before writing force bits

Run TxQuadcal with RxNCOphase=20(dec) (expect convergence)

Read TxLO and TxSSB convergence bits in register A7, A8 [D1:D0] = 11, LO and SSB converging

Start Tx, measure sideband rejection and LO rejection: medium rejection (40-50dB)

Since all bits in 0x09F are clear and Tx Quad Cal has completed, then according to the Register Map document  these registers hold the TxQuadcal correction words for phase, gain, Ioffsets, Qoffsets for Output 1

Now write register 0x09F = 0xFF

Since all bits in 0x09F are set, then according to the Register Map document, the values written to registers 08e...095 are used as correction words for phase, gain, Ioffsets, Qoffsets for Output 1.

Read registers 08e...095 and store their values. TxQuadcal correction words for phase, gain, Ioffsets, Qoffsets for Output 1 are now DIFFERENT from the readings before writing force bits. Where did these new values come from?

Measure SB and LO: now now POOR rejection (10-20dB), DIFFERENT from before writing force bits

Questions:

1. 1. If the force bits in 0x09F are clear and Tx Quad Cal has completed, which values are effectively used as correction words for phase, gain, Ioffsets, Qoffsets?
2. 2. If the force bits in 0x09F are set, which values are written to registers 08E...09D, by who are these values written, and at what moment are they written?
3. 3. How is it possible that after a not converging TxQuad cal, by simply writing the force bits, the sideband rejection and LO rejection become much better than after a converging TxQuad cal?