LM95221: Output noise

Hello Hiroshi,

Thank you for providing the information in the PowerPoint. For this device I would recommend either layering the D+/D- layer, using shielded cables, and/or adding a RC filter of the D+ to reduce the noise. Please review the app note: Optimizing Remote Diode Temperature Sensor Design (Rev. A). Do you mind sending me a zoomed in photo of the D+ signal in an active conversion?

Best Regards,

Meredith McKean

Hello Meredith-san,

 I really appreciate your prompt response. 

I'm sorry. Page 6 of the PowerPoint I created appears to be the D+/D- waveform and zoomed waveform.
The title of Page 6, "LTM4646 switching noise," was incorrect.

However, I am unclear about the difference between a failure and a normal operation, so I will confirm with the customer again.

Best regards,
Hiroshi

Hello Meredith-san,

I confirmed this with the customer.

Page 6 of the document shows the D+ and D- waveforms and their zoomed versions when abnormal temperature data was acquired.
(Only the title has been changed.)

A waveform has been added to page 7.
This is the waveform when normal temperature data was acquired, and it appears to be identical to page 6.

LM95221 issue (1).pptx

The waveform connected to the FPGA is said to be almost noise-free and even cleaner than the normal Unit A waveform.

I would like your opinion on what is happening in Unit B, as previously reported below.

Q1. Why is the acquired temperature unstable in Unit B?
Q2. What is the difference from Unit A?
Q3. Why does the temperature stabilize when the LM95221 is removed and then reinserted in Unit B?

Also, could you please answer one additional question?
Q4. Regarding the recommended layout for D+ and D-, looking at "Figure 12. Ideal Diode Trace Layout" on the datasheet, it appears that D+ and D- should be on separate layers.



However, the customer understands this to be a diagram with a guard ring formed by the GND pattern on the layout on a single layer.
Which is correct?

On the other hand, Figure "1.4 PCB Layout Considerations" in the application note clearly appears to place D+ and D- on different layers.

I think questions 1 to 3 are difficult to answer. If possible, please let me know your opinion, even if it's just a guess.
and I'm sorry for asking so much of you.

Best regards,
Hiroshi

Hello Hiroshi,

Since, LTM4646 is a high-frequency switching regulator I would suspect more noise within the internal BJT temperature measurement compared to the FPGA. The LTM4646 would likely need more protection on the D+ and D- line with filtering. Overall placing D+ and D- signal lines between the layers will add protection. I'm more concerned with these signals coupling noise from other sources than interacting with each other. What do you mean by stabilize? Was there a point when the LM95221 temperature reading was stable with the LTM4646 measurement? Can you provide a zoom in version of the D+ line within the active conversion. I want to see the 3 different levels as shown in the right figure.

Best Regards, 

Meredith McKean

Hello Meredith-san,
"What do you mean by stabilize?" means that the temperature measurement data below is relatively constant.

I asked the customer to measure a waveform in which three voltage levels can be distinguished, but the noise is too loud to distinguish them.
I have attached the waveform at the end of the PowerPoint.

LM95221 issue (2).pptx

Here's how I think about this issue:
*Similar noise is also occurring at the D-PIN, so the measurement may be correct.

*Slight differences in conditions or the influence of noise corrupt the delta-sigma modulator data, causing the measurement data to output erroneous values.

*Replacing the part improved some conditions, making it possible to measure.

*Even if the measurement is possible, the system is unstable and noise countermeasures are necessary.

The customer is still hoping for TI's opinion, so I am asking you for your opinion, but I understand that this is asking too much.

I apologize for the inconvenience, but I would appreciate your help.


Best regards,

Hiroshi

I will provide additional information.

There are no problems with the A/D conversion results of the built-in temperature sensor CH.
When the problem occurs, turning the power off and on does not restore normal operation, so it may not be a problem with the ADC core/modulator.

Hello Hiroshi,

What is the I2C register configuration of the device. Are the default settings used? 

Best Regards,

Meredith McKean

Hello Meredith-san,

I checked with the customer.
They are using 0x06 written to the Configuration Register (03h).
No other changes have been made.

Best regards,
Hiroshi