TPS546D24AEVM: Config Setting and Register Reading Problem

Bill Kuo 

Bill Kuo said:

1. I think the schematic of the TPS546D24AEVM demo board has a problem. That is Master and Slave Config pin(MSEL2) are connected together.

No, the MSEL2 pin of U1_P1 and U1_P2 are not connected.  The same net name is used within the sub-circuits, because the same sub-circuit is used for both P1 and P2, but there are no connections between the P1 and P2 sub-circuits unless they are connected at the top level schematic.

If you want to convert the TPS546D24AEVM from 2-phase to single phase operation, you will need to remove the 0-Ω resistor from R15_P1 and install a 0-Ω resistor in the R18_P1 (U1_P1 MSEL2 to AGND) location.  That will disconnect the VSHARE voltages and short MSEL2 of U1_P1 to AGND for stand alone operation.

Bill Kuo said:

2. There is something wrong with the reading value on Iout(0x8C). Ex. The reading value on the TPS546D24A EVM GUI tool is 0xC4DD, and convert to Decimal is equal to -3.14A. But in reality, the value should not be negative. It should be 0xC4DD (Heximal)=11000 10011011101(Binary) =>1245*2^(-8)=4.86A。In addition, the current value=4.86A is weird，because there isn't any load on TI EVM output.

No, the current is reported correctly.  Per PMBus standard, READ_IOUT is reported in LINEAR11 format, which uses a signed 5-bit exponent and a signed 11-bit mantissa.  b'11000 is indeed -8, but b'10011011101 is -803, so the reported current is -803 x 2^-8 = -3.1367  The mantissa for LINEAR11 is limited to -1024 to +1023

The small negative reported current is the sum of the residual offset of the two converters reported currents, and within expected tolerance at no load.

Bill Kuo said:

3. The voltage register reading value on Slave is not correct. We always read 0.92V on GUI, even we set  0x18 on 0x2 register to shut down O/P voltage.  The GUI can still read 0.92V on Slave and the current value is also incorrect.

The GUI is correctly reporting the READ_VOUT value reported by the PHASE = 01h device in the stack, but this does not represent the output voltage, because the sensed voltage is out of range and the reported voltage represents the saturation ADC input for the output voltage sense.

U1_P2 (The PHASE = 01h) has it's VOSNS pin unconnected and it's GOSNS pin pulled to 1.5V.  Due to the internal 120kΩ resistance between VOSNS and GOSNS, unconnected VOSNS pin is pulled up to 1.5V.  READ_VOUT on PHASE = 01h is trying to report the voltage on U1_P2's VOSNS pin, but 1.5V is out of range.  You can connect an external voltage to the VOSNS input to sense that external voltage.  When used in this way, the VOSNS pin should be kept below 0.8V to avoid saturating the ADC, and if a resistor divider is used to sense a higher voltage, the divider should be sized to account for the 120kΩ pull-up to BP1V5 through GOSNS.

I apologize for your confusion, but everything appears to be working correctly and as designed.

Bill Kuo said:

Regarding Q2, I would like to know that how can I adjust the offset of the current?

Because I found that if I use 3A load, the current I read is equal to almost  -3.1367+3= -0.1367A, but I hope that I can read 3A in this situation.

Each device in the stack has an "IOUT_CAL_OFFSET" command which can be programmed with an offset current to cancel the residual offset to the current measurement.  It is programmed by setting PHASE = 00, then setting the IOUT_CAL_OFFSET value for Device 0, and setting PHASE = 01 and setting the IOUT_CAL_OFFSET value for Device 1.  Alternately, if PHASE = FF (All) you can write the total READ_IOUT Offset value, and it will be divided among the devices and written to each.

Bill Kuo said:

Q2-2: We found even in single-phase standalone mode, we still get the current value = -1.7A with no load.

What's the reason that causes this offset value?

Is the current be calculated by MOS, VDS/Rds(on)?

The current measurement is a little more involved than simply Vds / Rdson in order to cancel temperature, process and VDD5 voltage variations in Rdson, but that is the basic principle.  Because the sense element is so small, about 500μΩ, variations between the automated test environment (ATE) and the application environment can introduce small offset errors.

Bill Kuo said:

Q2-3: If we use another PWM IC or Power Monitor IC, will we find the offset value when we try to read the current value?

If not, what's the difference?

That depends on the PWM IC or Power Monitor IC you select, and what it's post-trim residual offset error is. 

Some will have similar, or higher levels of offset, others lower.  Dedicated power monitors with current sense shut resistors will typically be the most accurate, since they are not attempting to use a dual function switching current sense element and are dedicated to the task of current sense.

Bill Kuo said:

Regarding Q3, thanks for your explanation, so in two-phase mode, the voltage value of Master and Slave should be the same which means that I can only read Master voltage value and ignore the voltage value on Slave, right?

Yes.  In a multi-phase stack, only 1 device has a direct connection to the output, device 0.  Additional devices have no connection to Vout and READ_VOUT is not necessary on those phases.  However, since all of the devices include a voltage monitoring ADC, we made the decision to make these measurements available to the user so that they could be used to monitor other system voltages.