• TI thinks resolved

TMP468: TMP468

Part Number: TMP468

Hi,

I'm looking for a remote temperature sensor that I can use in ATE, I need to monitor 9 thermal diodes (PNP).

Since the application is running in ATE, acquired temperatures need to be accessed at less than 10ms/channel.

I have an embedded controller that would access the remote thermal sensor device via I2C bus at speeds up to 1Mhz.

In the past, where I only needed to access up to 4 Tdiodes, there was plenty of time to acquire temperature, make a decision and then update the thermal solution.

But this new application where 9 Tdiodes on multiple chip module device will need to be assessed has me worried so I'm in search of  a new remote sensor .

So my question:

Does TI have a multiple channel temperature sensor chip that would meet this requirement?

Thanks,

Sheldon Grooms

MTS AMD 

512-779-6994

  • Hi Sheldon,

    The I2C bus at 1Mhz is fast enough to read all nine channel in ~0.189ms based on my calculations (#frames = 2N+3, see Figure 19). So we can eliminate the bus as a bottle neck.

    The 10 ms/ch requirement is challenging since the conversion time per channel is 17ms (max). Unfortunately, we do not have any other remote sensors that are faster than 17ms.

    You mentioned that in the past you had success with 4 channels. May I ask which device you were using?
    You can read the temperature but it doesn't mean that the temperature is changing in the registers.

    -Kelvin

  • In reply to Kelvin Le THS:

    Hi Kelvin,

    So I was using 4 of the EMC1412-A, which converts at 15.6ms in it's faster mode and I wasn't getting unique values each read. I would usually get 3-4 exact readings per channel.  I was also fore-going the low-byte read, since thermal solution is +/- 3C.  This situation worked pretty good for the DAQ of the thermal control system which had  an upper limit of ~200hz for the system response.  

    With the EMC1412 I could read the high-byte and be within 1C, but it looks like I will have to read both high/low byte with this device or risk being up to 2C off. Which is twice as many transactions, but as you mentioned at 1Mhz it shouldn't be an issue.

    Looking at oneshot vs continuous, I'm assuming that in one-shot, write one-shot register and then poll a status bit and read when not busy. Which should be 17ms max. What happens when I read the register while still busy, do I get current contents?

    In continuous mode with 8 remote enabled and local disabled, complete conversion happens every 126.4ms, what happens if I read register before the 126.4ms expires? Are all temperature registers updated at same time or as conversion completes for each channel?

    I have ordered the TMP489 evaluation board and will give this device a try.

    Thanks,

    -sg

  • In reply to sheldon grooms:

    Sg,

    Q: What happens when I read the register while still busy, do I get current contents?

    A: The status bit tells you if the ADC is busy or not.  The conversion is done in a sequential manner (round robin) and each channel result is updated sequentially also.  If you're using the one shot mode with 8 remotes enabled, the BUSY bit will change only after all of the channels are finished (ADC inactive) so you can't use this status bit to gate the end of each individual channel.  You might have to use a timer and set the interval to 17ms and then read the result. 

    -Kelvin