DS125DF410: Questions/clarification of settings related to eye opening monitor

ticket is assigned and will relpy soon.

See below responses to the customer questions.

The default power-on-reset state of register 0x31, EQ_SM_FOM (bits 4:3, CTLE state machine figure of merit) is 0x0. In contrast, the default power-on-reset state of register 0x2C, DFE_SM_FOM  (bits 5:4, DFE state machine figure of merit) is 0x3. The documentation seems to consistently refer to the value of 0x0 as “not valid” in both of these registers.

1. Could you confirm that the default power-on-reset state of register 0x31, EQ_SM_FOM is indeed not a valid mode and _must_ be changed for proper operation? 
2. If possible, could you comment on how the adaptation will proceed if EQ_SM_FOM is left at 0x0 (the default value)?

The default power-on default value for channel register 0x31 is valid and does not need to be changed. The default operation is for SM to use both HEO and VEO. Sorry for the confusion, as it looks like there is error on the datasheet description. The correct description is show below.

CTLE (EQ) adaption state machine figure of merit settings.

00: (Default) SM uses both HEO and VEO

01: SM uses HEO only

10: SM uses VEO only

11: SM uses both HEO and VEO Additionally, if Reg_0x6E[7]=1, the Alternate FOM is used. Reg_0x6E[7] takes precedence over EQ_SM_FOM.

When overriding the CTLE setting as in section 7.5.8, the outlined procedure does not set EQ_BST_OV in register 0x2D bit 3. The default value for this field is 0x0. Per the description, it seems this field should be set to 1 when overriding the CTLE boost.

Correct, to override the CTLE and manually set the boost setting 0x2D[3] must be set to 1.

Table. Manually set CTLE Boost Value for DS1xxDF410 retimers

For the default vertical eye opening monitor used by the state machine to detect lock, could you confirm the correct interpretation of register 0x28?

If VEO_SCALE, bit 6 of 0x2C is set to 1 (default), should we interpret 0x28, VEO to be:

1. Automatically scaled to remove the effect of VRANGE? In other words, if we read register 0x28, VEO to be 255 with VEO_SCALE==1, would this indicate +/-400mV regardless of EOM_VRANGE_SETTING, bits 6:5, register 0x29?
2. A measurement scaled to EOM_VRANGE_SETTING, bits 6:5, register 0x29? In other words, if we read register 0x28, VEO to be 255 with VEO_SCALE==1, would this indicate +/-400mV if and only if EOM_VRANGE_SETTING==0x3?                                                                                  ***Please note we are seeing behavior which is more consistent with (B) above.

The VEO reading is automatically scaled. Regardless of VRANGE setting, VEO_mV = Reg_0x28(Decimal) * 3.125

Additionally, could you confirm how the VRANGE interacts with the thresholds (VEO_INT_THRESH, bits 3:0 reg 0x32; VEO_THRESH, bits 3:0 reg 0x33; VEO_LCK_THRSH, bits 7:4, reg 0x6A)? In other words, if we set VEO_THRESH, 0x33[3:0]==0x8 (default), this corresponds to a full VEO of 64 codes. Should we interpret this to mean we will require an opening of +/-400*(64/256)=+/-100mV at the internal comparator before starting DFE adaptation (when is set to VEO_SCALE=0x1)?

VRANGE does not interact with those parameters. The VRANGE only affect the EOM full eye plot window vertical size

Finally, if VEO_SCALE==0 and we manually set EOM_SEL_VRANGE in bits 7:6 of register 0x11, will the scale for interpreting the thresholds and VEO be changed as well?  In other words, if VEO_SCALE==0 and EOM_SEL_VRANGE==0, what would VEO_THRESH==0x8 correspond to in terms of mV opening? If VEO_SCALE==0 and EOM_SEL_VRANGE==0, what would VEO==255 correspond to in terms of mV opening?

If VEO_SCALE is set to zero then 0x11[7:6] allow you to manually set the EOM vertical range

Section 7.3.3 of the datasheet mentions that different stage combinations have different equalizer transfer functions.

Would you by any chance have any data you could share regarding the different possible shapes?

In our application we are finding that we likely have a source of high-frequency aggression onto the differential lines (at frequencies greater than the line rate). We are finding that this aggression coupling in some cases can be strong enough to degrade the monitored vertical eye opening. Thus, we are wondering if specific linear equalizer settings provide better attenuation at those high frequencies.

Please provide a work email address and I can directly email the available DS1xxDF410 CTLE boost curve data

For the figure of merit for adaptation, specifically referring to section 7.5.17, the datasheet says "The vertical eye opening is scaled to a constant reference vertical eye opening and the smaller of the horizontal or vertical eye opening is taken as the figure of merit".

What is the vertical eye opening scaled to? Is it possible to read this scaling or the result of the scaling?

The VEO LSB is 3.125mV. The VEO values can be read via channel register 0x28. The VEOs observed during the EQ auto adaption process are not recorded and thus are not available to the user for reading

Furthermore, equation 1 in this subsection references "HEO" and "VEO" as inputs for a custom figure of merit controlled by factors a, b, and c as set by 0x6b, 0x6c, 0x6d.

For default retimer settings, the FoM equation below is implemented. TI expects the default FoM to work optimally for most applications

o FoM = minimum (HEO, VEO)

• If the Alternate FoM is enabled then the equation below is used:.

Alt FoM =  (HEO-B)* A*2 + (VEO-C)*(1-A)*2

HEO_ALT = (HEO-B)*A*2

VEO_ALT = (VEO-C)*(1-A)*2

Are the input "HEO" and "VEO" equivalent to registers 0x27 and 0x28, respectively?

Yes