TPS65982: I_PP_EXT not readable if in SINK mode

Eric,

   I thought I was told earlier that the OCP worked in either direction. I have connected my board up to read the current with SENSE_p connected to the VBUS side of the resistor and SENSE_n connected to the PP_EXT side of the resistor. This way I read the current when the system is operating as a SINK of current from VBUS. I did do exactly what you said and as noted above I can read 1.5A from the TI-EVM board when it is operating as a source with the resistors in that configuration. But on my board with the resistors flipped and operating as a SINK I don't read the 1.5A. I also move the resistors on the TI-EVM board and on my board and went the other way, and could read the 1.5A on my board as a source, but NOT on the TI-EVM board as a sink. So I could never read the current in SINK mode, on either my board or on a changed TI-EVM board, and it was much more than the 40mA you say above.

1) You are saying that you can read that SINK current?

2) If you read the 8.3.3.15 External Current Limit section of the datasheet is says that current is limited in both directions because it uses the RSENSE magnitude. I was interpreting that as it worked in either direction. Am I incorrect? If so, this is disappointing since I asked this question months ago and didn't think I had to switch the sensing in my design to get current limit operation in both directions.

Eric,
I found out that there are multiple problems with the ADC / PP_EXT SINK / 5mOhm R system.
Here is what I tried today...
I use 3 boards, my board with the PP_EXT connector and the VBUS connection and a 5mOhm sense resistor and the SENSE_P/N setup to measure SINK current out the PP_EXT connector to a power supply that can sync current. I also used a TI-EVM board set to switch setting of #6 (5V/12V/20V source) and connected to the HP laptop supply at 20V.

The 1st thing I tried was to set up a 5V/1A contract between my boards, using an external 5V/3A supply into the board set as a Provider 5V/1A contract and the other set up as a consumer 5V/1A. After establishing a successful contract I started increasing the sink current on my supply, starting at 0.5A, then 1.0A, then 1.2A, then 1.5A, 2.0A and finally 2.5A. Neither TPS65982 part (one on each board) broke the contract, even though I was over 2x the valid value. If I read the ADC value on either board using my debug port and the I2C lines and a Aardvark I2C/USB plugged into my board I read 0.0000A, no matter the current flowing. I looked at register 9 after the read and it is 0003 (the 03 being the channel - I_PP_EXT) and the 00 being the register value. I have the 5V contracts using the PP_EXT path on both boards.

I then did the same thing, this time with each having a 20V/1A contract (2nd PDO) and connected a 20V/3A supply. Both established a valid 20V contract @ 1A, one the provider and one the consumer. I then proceeded again to ramp up the current on the consumer up to 2.0A and again neither broke the contract! They were both using the PP_EXT path through the FETs on my board.

I then changed the source to the TI-EVM board, but to test the PP_EXT path I had to use the 20V since that is the only thing the PP_EXT path on the TI-EVM board can use. On the TI board it uses the SOURCE configuration for the resistors. I used switch setting #6 on the TI-EVM board, and through the I2C debugger, changed the source capability to 20V/1A instead of the 3A. I made the 20V contract, and upped the sink current up to 1.22A before it started issuing hard resets with the message from the TI-EVM board of HardReset… Unable to Source. So, it limits the current in the SOURCE mode correctly.

Then I modified one of my boards to flip the sense connection, so that I had it like the SOURCE mode and then used that board as the provider I expected similar results to the TI-EVM board at 20V. However, when I read the ADC on my board with 1.0A flowing, it only read around 0.5A!! Sure enough, the ADC is giving incorrect values, about 1/2 of the real value. So the Host Interface utility is showing about 0.5A when the real current is 1.0A. I can see the real current through an inline current monitor and also on the display of the power supply which is doing the sinking. If I put back in the TI-EVM as the source and read the same ADC channel (I_PP_EXT) I read around 1.0A. I then put my source board back in and cranked up the sink current and it didn't open the connection until I reached over 1.8A, even though the contract is at 1.0A. Remember, the TI-EVM opened at around 1.2A. The ADC value at 1.8A was finally near the value you would read if it were not halved.
I'm guessing that the problem is that I am using a 5mOhm sense resistor in my EXT path and the ADC doesn't work with that . YES, I did check the system config register, and it is indeed set up for the 5mOhm value, so that isn't the problem.

So, summing up:
The I_PP_EXT current CANNOT be read by the ADC when the SENSE_P/N connections are set up in SINK mode (current flowing from VBUS to PP_EXT) as shown in the datasheet. The path does not get shut off in this mode, even if 2.5x the current is flowing than the specified contract (I was using 1.0A). This was true at both 5V and 20V.
The I_PP_EXT ADC value isn't accurate in SOURCE mode with a 5mOhm resistor, and therefore, much more current can flow before it shuts off.

If you would like me to try something in more detail I can, and I can send you screenshots of the Utilities results. I even probed the I2C lines and observed the traffic. The values read matched what the Utilities software showed.

Here is that traffic I read during a typical ADC read using the Utilities software:
Read register Reg03 (do a mode read)
Response 04 41 50 50 20 (APP ) [04 if the TI part responding with the length]
Write Reg09 04 03 00 00 00 (4 bytes, ADC channel 3)
Read register Reg03 again
Response 04 41 50 50 20 (APP ) [Don't know why the mode keeps getting read, must be a default behavior of the Host Interface Utility app]
Write Reg08 04 41 44 43 73 (This is the ADC read command)
Read Register 03 again.
Read Reg08 04 00 00 00 00 [Command is done]
Read Register 03 again.
Read Register 09, the results of the ADC read. In my case I read 40 03 98 00 on the bus.
This is the size (64 bytes, followed by the channel (03) I_PP_EXT, and then 98, the hex value of the ADC)
There is some other stuff that follows, a register 03 read, a register 28 read (system config - I'm guessing to read the RSENSE value) and then finally a couple more reads of register 03.

But again, the RSENSE isn't being applied correctly in the Utility.
It doesn't look like the path is shutting off correctly even though this is being read.

Answers needed:

1) Do you have someone who can check to see if the limit works when connected in SINK mode (SenseP connected to VBUS, SenseN connected to PP_EXT). I tried with two TI-EVM boards even and can't get any reading from I_PP_EXT.

2) I found that in source mode (I modified my board to flip the pickoff locations), that with a 5mOhm sense resistor, I don't get an overcurrent until almost 2X the current is flowing with a 1A contract. Can this be fixed by programming? I wanted to check the reading again, so here is what I did today...
I connected my two boards together, one modified to read in SOURCE mode and connected to a 5V supply. Both are using the PP_EXT path and both have 5mOhms as Rsense.

I made a 5V / 1000mA contract between them.

I then read the ADC channel, I_PP_EXT 5 times.

.539062 5C(92), .550781 5E(94), .539062, .544922 5D(93), .533203 5B(91). This is OFF, since I see 1.0A being drawn by my sink power supply and also displayed by an inline ammeter.

I then wondered if I set the System Register on the provider/source board to pretend it had a 10mOhm resistor if the reading would change.

I then read two more times:  .533203 5B(91), .544922 5D(93). So, even with the system register saying it was 10mOhm, it didn't change the reading. So the Rsense must not be used currently in the calculation, even though it probably should be, since the reading are around 1/2 of what they should be.

Then increased the current with the circuit having a 5mOhm resistor, but the system register saying 10mOhm
-1.5A : was reading about 0.8A, still had the connection

-1.6A : 0.85-0.87A

-1.7A : 0.91-0.92A (Hex 9B ADC)

-1.8A : 0.96-0.97A (Hex A6)

-1.9A : 1.00-1.01A (Hex AC)

-2.0A : 1.06-1.07A (Hex B6)

-2.2A : 1.17-1.19A (Hex CB)

-2.3A : 1.22-1.23A (Hex D0).

-2.4A : Overcurrent. This is what I found yesterday with the TI board and the 10mOhm resistor, it quits at about 1.2A with a contract = 1.0A. But note, I am sinking 2.4A.

Changed back to 5mOhm resistor value and it climbed up in the same way but it quit instead at -1.9A and broke the connection.The readings at each current level were the same as with the 'fake 10mOhms' in the system register, even with the 5mOhm setting. I.E. and -1.8A, I was reading 0.96-0.97A from the ADC read.

I get an overcurrent interrupt at 1.9A. Still way above the 1.0A of the contract.

Again, what I've found is that with a 5mOhm resistor the ADC readings are ~1/2 of reality. I'm using the Rdg * 1.2 / 1024 * 5 formula in the firmware users guide. I don't see any note about the IHVEXT_ACC numbers being anything but 5 when using a 5mOhm Rsense resistor, but the spec only refers to 10mOhm values. Does this really mean I can't use a 5mOhm resistor? It is specified all throughout the other documentation and I just released my board, so if I can't use a 5mOhm resistor I need to know soon!
I used the 5mOhm part fully confident that it would work, because it would burn less power at 5A loads. I didn't think the circuit wouldn't shut down correctly. But maybe it works better at higher currents than the 1A-2A that I was testing. I also thought it would work in SINK mode, but this doesn't seem to be true.

The TPS65982 datasheet has PP_EXT current limit values depending on 'Settings' from 0 to 15, but it doesn't say how to enter those, and I thought it was just a matter of the contract value. If there is another way these are programmed, I would like to know, since this could solve my problem.
But none of this will help if the SINK current can't be sensed to shut things down.

Sincerely,

WST

This description matches my observations as well. 

In sink mode, no matter how much current is going through the sense resistor, the device reads back 0A and the OCP does not work. 

I could connect the two EVMs together and repeat the experiments but I'm using the TPS65982 on my own board and that is what I have connected together, one of my boards as a sink and one as a source. Both are using external FETs (the same ones that are on the EVM board) and they work fine in all other aspects. They both are using a 2.15 generated software image. It will take some time for me to set up with two EVMs and do what you ask. Both EVMs have the original software on them so that they will respond to the 4 switches. I haven't updated them to 2.15 images but could if it is necessary. I can connect to the SINK board with my Aardvark/FTDI and read the ADC as they are.

Can you tell me what combination of EVMs you have that works to read the I_PP_EXT current in SINK mode? Maybe I could duplicate that setup.

WST 

First of all, thank you very much for that excellent write up and experiment. I went through the steps you outlined and created an image (after saving away the old EVM image) using the template you suggested. I had previously switched the R2/R6 with the R3/R5 on one of my EVMs and so I re-programmed that board. I then connected to a laptop supplying 5V on VBUS and that could put out 900mA. I did measure SENSEP < VBUS and set up an external sink to draw 0.5A from J6. I made all of the voltage measurements and had the following:

VBUS = 5.025V and SENSEP = 5.011V. With my sink drawing 0.499A, I measured with the I_PP_EXT ADC reading a value that was around 0.45A, so it was successful!!

BUT...

This doesn't match the operation I want, it looks like this ONLY works in dead battery mode. To test this I disconnected the J6 jumper and reconnected my sink to just pin 2 of J6 and so it wasn't shorted to the 20V External power of pin 1. I then plugged in the power supply to the barrel jack so that the TPS65982 wasn't in dead battery mode. Using the utility tool I modified the image so that the EVM would connect in sink mode with the PD contract and made the connection.

The changes were to change the System Config register so PortInfo was CONSUMER_CCPD_CCIPU and making sure that I didn't have any Source PDOs, only the 5V/900mA sink capabilitiy. After connection I verified that indeed it was in SINK mode and drawing current from the laptop. I measured the following:

VBUS = 5.33V and SENSEP = 5.01V. However, with the sink still drawing 0.499A, I measured with the IPP_EXT ADC reading a value of 0.0000V.

So, I want to be able to read the SINK current when the TPS65982 is powered by the VIN_3V3 pin and isn't in dead battery mode. Is doesn't look like this is possible, unless I can set some other values in the registers.

Here are the pictures I took:

1) ADC reading set up as you did

PD status after making changes to have the board powered:

Status register in powered mode:

And finally the ADC reading in powered mode, always reads 0.0000A

Eric,

This is getting very confusing and I suspect I just can't do what I want, and that the datasheet and manuals of the TPS65982 don't describe the operation of the I_PP_EXT path very well (at least to me).

1) You said: "The I_PP_EXT function will not work if you have RCP enabled on the external path."
I can't find this anywhere in the Host Interface Tech reference manual, the Firmware Users guide or the TPS65982 datasheet. The closest thing I could find to this is note 4 on the IHV_ACC spec, but the IHVEXT_ACC spec doesn't refer to this RCP only operation and seems to indicate that it should work in these modes.

2) Your process above is also not documented anywhere. I was able to repeat that with my modified EVM board, but it looks like this only works if I start in dead battery mode. The Host Interface Tech Ref Manual doesn't describe status register 1A this way. It says that bits 8-15 indicate the current state of the switches, not the RCP status. So when I see 'Disabled' I think the switch is turned off.
Here is the entry in the manual:

Since I can read the current when this is disabled, I REALLY don't know what is going on or what path is being used. I read this register using the READ command and get: 0x194000d. Bits 15:8 are all '0', so according to this manual, the switch should be opened.

3) When I look just a little further down in the register 1A readout than you showed, I see that the TPS65982 is in some odd mode, not the operation that I desire:

4) I have my board wired with the RPD_G1/G2 inputs tied to GND and the BUSPOWER_Z tied though 100k to VP3V3. Because of the first, my board therefore can't boot up in dead battery mode.

Maybe I'm misunderstanding the manual/datasheet but I don't see a way to read the I_PP_EXT current when operating with a 'normal' PD handshake and while sinking current, even though SENSEP > PP_EXT. If that is true, please let me know. I can't operate my IC/board in an odd way just to get this function to work.
It also seems true that the OCP will never work the way I have wired my sense connection as described in section 8.3.3.12 of the TPS65982 datasheet, since the ADC can't read the current correctly when connected this way. Is this section invalid?

Sincerely,
WST