TMAG5273: Strange configurations of TMAG5273 and not getting a CRC the way it's described in the datasheet

Tim von Unen

Part Number: TMAG5273

Dear TI,

As you can read in my other (related) question I am working on developing software for use with the TMAG5273.

I’ve come quite far in the development, also thanks to Merke, and the communication works stable ad the sensor seems to work as expected until now.

First of all one strange thing I’ve noticed from the sensor is the following:

For the chip to work optimal for your needs you need to configure it using the writable registers 0x00 – 0x0C using the standard I2C write command from Figure 7-7.

The configurable registers 0x00 – 0x0C can’t be read-out using the 1-Byte I2C Read Command for 16-Bit data of figure 7-11 and 7-12. They can only be read out using the Standard 3-Byte I2C Read of figure 7-9 and 7-10.

Which one of the two read commands you can use to communicate with the chip is depended of the I2C_RD field of the DEVICE_CONFIG_1 register (one of the configurable registers).

The writing of the configuration registers doesn’t contain any CRC or checksum of some kind so to be sure the writing was done correctly you need to read out the registers you’ve written and check if they contain your values.

So in other words if you want to use the 1-Byte I2C Read Command for 16-Bit data of figure 7-11 and 7-12 you have to set the I2C_RD field (0x01) BUT you can never check if this was done correctly because for that you need to set the I2C_RD field to a different value (0x00). Is this correct or am I misinterpreting something?

I am just interested in the 16bits X, Y, Z. The quickest, and in my opinion the easiest, way to do this is to make use of the 1-Byte I2C Read Command for 16-Bit data of figure 7-11 and 7-12. I had implemented it this way but because I need robustness and the product I am developing must work at all times I chose to redo the work and just make use of the Standard 3-Byte I2C Read of figure 7-9 or 7-10. With this command you can read out, and validate, the configuration registers AND you can also read out the result registers 0x12 – 0x17 to get the 16bits X, Y, Z values.

Then the second strange thing came to light:

For the same robustness I was talking about earlier I implemented a CRC8 calculation and set the CRC_EN field of the DEVICE_CONFIG_1. Now if you are using the Standard 3-Byte I2C Read of figure 7-9 or 7-10 the TMAG5273 doesn’t support CRC if the data length is more than 4 bytes according to a note in the datasheet. (which doesn’t even seem to be true but later on more about that). For me I want to read-out 6bytes so I can’t get a CRC unless I chop the command in two and first read-out 5bytes (2byte X, 2byte Y and 1byte CRC) and then consecutively read-out 2bytes. Which costs way more time than just the simple 1-Byte I2C Read Command for 16-Bit data of figure 7-12 (with CRC). Again am I right or am I missing something?

Then a bit more of a technical question:

I configured the chip to be as follows using the packet structure shown in figure 7-7 (Standard I2C write):

Device Config1 (Address 0x00): 0xB4 (Crc enabled, TempCO = NdBFe, Con_Avg = 32X, I2C_RD = 3-byte read)

Device Config2 (Address 0x01): 0x12 (Low noise mode, Continuous measurment mode)

Sensor Config1 (Address 0x02): 0x70 (XYZ channels enabled)

The message that I composed an wrote to the chip (confirmed using an oscilloscope):

| 0x6A | 0x80 | 0xB4 | 0x12 | 0x70 |

Received an ack back on all bytes so everything OK

After that I try to read out the registers using the packet structure shown in figure 7-10 to confirm that the write was done correctly. I try and read 5 bytes as that is what the datasheet says you need to do to get a CRC back. What I expect to see on the oscilloscope:

| 0x6A | 0x80 | 0x6B | 0xB4 | 0x12 | 0x70 | 0x00 (reset value of register 0x03) | 0xC7 (CRC) |

Instead I see:

| 0x6A | 0x80 | 0x6B | 0xB4 | 0x12 | 0x70 | 0x00 | 0x00 (reset value of 0x04) |

If I try to read out 6 bytes, still no success.

| 0x6A | 0x80 | 0x6B | 0xB4 | 0x12 | 0x70 | 0x00 | 0x00 | 0x5B (CRC) |

What I get:

| 0x6A | 0x80 | 0x6B | 0xB4 | 0x12 | 0x70 | 0x00 | 0x00 | 0x00 (reset value of 0x05) |

Only if I try to read out 7 bytes I get a CRC back

What I expect:

| 0x6A | 0x80 | 0x6B | 0xB4 | 0x12 | 0x70 | 0x00 | 0x00 | 0x00 | 0x86 (CRC) |

What I see:

| 0x6A | 0x80 | 0x6B | 0xB4 | 0x12 | 0x70 | 0x00 | 0x00 | 0x00 | 0x86 (CRC) |

Finally I received a CRC. What am I doing wrong?

Sincerely,

Tim von Unen

over 3 years ago

+1 Mekre Mesganaw over 3 years ago

TI__Expert 7485 points

Hi Tim,

Although it may not be the ideal way of checking, you might also be able to look at the SET_COUNT bits in the CONV_STATUS data byte for 1-byte I2C read commands. These bits should increment every time you take a new measurement. If you don’t see these bits incrementing after triggering new conversions, then this might indicate that the CONV_STATUS byte is not being sent and that the 1-byte I2C read option is not being used.

You are correct in that you will need to split the 3-byte I2C read operation in half so that there are no more than four data bytes when the CRC is enabled. I am not sure though why you are not seeing the CRC byte being sent at the right time. From my side, I was able to receive the expected packet you mentioned:

Can you verify that you did not accidentally send a start or stop condition at an inappropriate time during packet transmission?

Thank you,

Mekre

0 Tim von Unen over 3 years ago in reply to Mekre Mesganaw

Prodigy 30 points

Hi Mekre,

I just realized I misspelled your name twice, my sincere apologies for that!

Thanks for the tip about the alternate way of checking, I will look into that and check if this is enough for me.

That snapshot is very handy. I’ve had a close look at it and I saw a small difference in my oscilloscope image and your snapshot.

My stop condition looks as follows: At the rising edge of the ninth clockcycle the ack bit gets clocked in and at the falling edge of the ninth clockcycle the TMAG5273 release the SDA line which makes it go to VCC (if the primary doesn’t take over pulling it low). If at that point you want to perform a stop condition you first have to pull the SDA line low, then release the clock and then release the SDA line once again.

In between the 0x80+ACK and the Read[0x35]+ACK I've implemented this stop condition and immediately after that a start condition which makes it look like this:

Zoomed in with in redcircle the stop and in green circle the start:

It seems like this one ‘extra’ flank on SDA makes the TMAG5273 start counting from zero again.

Then it counts 7 bytes and at the 8^th it inserts the CRC. (0x35R 0xB4 0x12 0x70 0x00 0x00 0x00 CRC)

Whereas if it starts counting from the beginning 7bytes, 8^th CRC (0x35W 0x80 0x35R 0xB4 0x12 0x70 0x00 CRC)

If instead of performing a stop- and start-condition I write a new ResetCondition function which leaves away this extra flank on SDA (like in your snapshot). Then I get the expected result :)

Maybe this unexpected discovery is helpful for you guys to know as well because for me it saves the day.

As I said I would like to write the configuration register and be able to check them and after that just get out 6bytes + CRC continously which now will be possible:

0x35W 0x80 0x35R XMSB XLSB YMSB YLSB ZMS ZLSB CRC

Thanks for you help and quick response once again!

Regards,

Tim

Sensors

Sensors forum

TMAG5273: Strange configurations of TMAG5273 and not getting a CRC the way it's described in the datasheet