LM75A I2C hangs

Moving to temperature sensors forum.

Hi Lalit,

The LMT75A should not be read 500 times in 1 second. The datasheet on page 4 states the wait time between each conversion be at least 300ms:

Let me know if that solves the problem.

-Michael Wong

Thanks Michael, I will check and update you.

Michael I have doubt if this is affecting I2C. As we have also used NXP part LM75A, which was hanging even after reading rate of 1 second. in NXP datasheet, there is also mentioned that "This Temp register can be read at any time by a controller on the I2C-bus. Reading temperature data does not affect the conversion in progress during the read operation.”

Hi Lalit,

This is strange. Can you provide a schematic of your system?

-Michael Wong

On increasing reading rate like 500 times in 2 second with NXP IC, it gets stuck within 1 or 2 hours.

Hi Lalit,

Your schematic seems fine to me. I made a mistake in my first post, it is possible to read the LM75 500 times a second but the temperature will not update to a new reading unless there is a 300ms wait time in between readings. The NXP LM75 is a copy to TI's LM75 and should behave the same way. Since two parts are not working correctly it seems like the problem may be with your system and not the parts. Have you tried communicating with any other I2C parts and found similar issues?

-Michael Wong

Hi Michael,

We are using other I2C parts on the same board but on other I2C port. There is no problem. For temp sensor IC, I observed that both I2C lines are getting stuck low. On giving hardware reset (reset to processor through switch), SCL line goes high, but SDA line remains low. I have also tried by adding 33pF capacitor on SDA line to Ground as mention in one of the TI application note (AN-2113), but this also not resolving the issue. On probing SCL, fall time is about 3ns to 5ns and rise time is about 750ns.

When I am increasing reading repetition rate in one second in hundreds like 100, 500 etc, it get stuck within 12 hours. In some boards with less repetition rate also, this issue was observed. I have put 2 boards with TI IC and one board with NXP IC and this board has capacitor on SDA line, with repetition rate of 3 in one second. These 3 boards are running for one week without getting stuck.

-Lalit

 

Hi Michael,

 

The issue is being observed again even with lesser rate of access.

 

With regards

Lalit

Hi Lalit,

Michael was out today so I'm replying in his place. There is no way the LM75A can drive the SCL line low. This is a logic input - no output drivers there.

The LM75A state machine resets the SDA output if it communicates a low for too long. Are you sure you are using an LM75A? This function is called TIMEOUT as described in the EC Table:

Are you 100% positive that the connector that doesn't go anywhere really doesn't go anywhere?

How many devices are on this bus?

I didn't see the pullups in your schematic so I assume they are in some device. Could the processor somehow be removing the pullups during reset thus the bus you go low? On multi-function MCU pins it's usually wise to have them go to high impedance until software runs.

It's just curious the SCL also gets stuck low if it is the LM75A it means it is malfunctioning in a very big way.

Have you tried more than one part?

Have you tried a different board do all boards do this?

By reading less and seeing that the failure rate lessens could point to a marginal timing issue:

Could the processor by mistake be seeing a false start and misbehaving (Clock transition to low while data is high). You mentioned putting a 33pF on the SDA line did that make the failure rate worse? You may try a bigger cap and see if the failure rate gets worse.

Putting a cap on SCL would slow down the falling edge of the clock that could cause the failure rate to lower or go away.

Probe the signals at the input of the processor pins with buffered probes (not cap loading) and see how close the timing is.

Can you send us scope photos of the failure as it occurs, probe SDA, SCL and VDD (I know it's a pain but it has to be done)...

I'll be out for a while so Mike will pickup the discussion again on Monday. Hope you come to quick resolution!

Take care, 

Thanks Emmy for detailed reply.

Yes, I am using LM75A. Both SDA and SCL go low.

These lines are connected to one connector in addition to LM75A. But on mating connector (i.e on another board), these pins are not connected. I2C lines provided there for future purpose.

More than one parts is not tried. Pull-up resistor of 10k ohm are put on both lines, were on different page, so forgot to show in this schematic.

This problem comes in almost all boards, but the failure rate is different.

I had tried 33pF on SCL line on one board and after that it takes more time to fail, but the problem is not removed permanently.

We don't have buffered probes, so will not be able to probe as you told. Bytheway we have probed SCL and SDA, and was able to capture these signals. Bleow are waveforms, when it stuck low and working

 

With regards

Lalit

I was able to capture I2C lines with VDD. I am attaching here.