Post Go back to editing

Long Term Reliability of the LTC6909 Clock Chip

Category: Hardware
Product Number: LTC6909
Software Version: N/A

We have utilized the LTC6909 device in some of our radio products that have been in the field in excess of 12 years, in 24/7/365 operation. We have recently had field returns where the failures seen to stem from the LTC6909. The major symptom is that the output clock frequency has shifted down from the expected 600 KHz to a significantly lower frequency, as low as 220 KHz in at least one case. In addition, the spread-spectrum aspect of the clock output, which is enabled in our design, is no longer working. When we observe the resulting clock outputs on a spectrum analyzer we see a discrete line at one frequency, vs. when we compare with a good, working unit and the spread-spectrum aspect of the clock is clearly visible. 

We utilize these clock outputs to synchronize some switching regulator power supplies (buck converter topology), so the shift in the operating frequency has impacted the stability and output ripple from those regulators, which supply DC power to an Altera FPGA on our main processing board. This can lead to all sorts of subtle problems.....but the root cause appears to be the LTC6909 device. We use the Industrial grade version of the part which is specified for operation from -40C to +85C. 

It seems clear that some analog circuitry internal to the device has shifted. The power supply input to the device remains at the nominal voltage = V+A = V+D = +5.2V. The R-set value is unchanged at approximately 41.9K, and we have all three phase control pins pulled high (i.e. PH = 8) so we would have an expected clock frequency = 597 KHz. We have over 25,000 of these radios deployed in South Korea for over 12 years, and we now have at least 25 units returned with this problem. The amount of the frequency shift to lower output frequencies varies, but the loss of the spread-spectrum mode seems consistent. 

Thread Notes

Moved from Power Management to Clock and Timing on Friday, November 10, 2023 11:48:09 PM by MattGorder

  • Hi Matt,

    Thanks for bringing this behavior to our attention. I relayed it to our reliability team. We will investigate this behavior.

    Best Regards,

    Emrecan 

  • Hi Emrecan--Would you like to receive some samples of the defective LTC6909 devices for failure analysis?

    Regards,

    Matt

  • Hi Matt,

    Can you send the schematic of the LTC6909, especially the MOD pin connection? 

     Another thing I was curious about is the error spread on the failed parts. (not just the worst-case 220kHz). Is it quantized into certain increments?

    About the failure analysis, I am certain we can learn a lot from these parts. However, as a product application engineer, I don't have any control over the FA process. "Our reliability team is responsible for handling all reliability-related inquiries."

    Do you have any local field sales or field applications engineers with ADI that you can reach out to? They can help with the FA process. 

    Thanks,

    Emrecan

  • Hi Emrecan,

    Here is a schematic snippet showing the circuit for the LTC6909. We operate the device from +5.3V, and we utilize an R-set = 41.9K. As previously noted, this circuit was working fine for > 10 years, but we're now starting to see some returns from the field where the output clock frequencies are not correct. These are used to control the switching frequencies for a variety of switching regulators to provide DC power to an Altera FPGA. 

    We don't have a lot of data on the various frequencies, but it does not appear to be quantized. It appears to be more or less random. We have visually inspected the soldering of the PH[2:0] and the MOD pins and the soldering quality looks fine. Note that we don't load the zero Ohm jumper connecting the OUT7 clock output to the MOD pin, so that spread-spectrum operating mode should be enabled. 

    We can reach out to our local Analog Devices FAE = Mr. Armin Gerstberger as well.I will get him in the loop ASAP. 

    Thanks very much for your help. 

    Regards,

    Matt

  • Hi Emrecan,

    Do you have any updates regarding this case? Our local FAE is on vacation. Do you have an alternative contact that can help arrange the return of defective parts for a failure analysis?

  • Hi Matt,

    We had discussed this topic internally. Our designers' comment was that 5.3V is on the high side of power supply limits. Ripples on the power rails might have caused a violation of maximum input voltage and have harmed PH pins in a long-term operation. However, we won't be sure until the parts are inspected by our FA team.  

    Maybe you can contact Mr. Ovidiu Balaj, (manager of the Mr. Armin Gerstberger). He would know who covers your account in the absence of Mr Gerstberger. 

    Thanks,

    Emrecan