DS32EL0421: Environmental Limits

Hi Jeffrey,

An engineer has been assigned to help you with this inquiry. They will be able to help you shortly.

Best,
David

Hi Jeffrey,

It looks like we have characterized this device over temperature, but I don't see any data on performance over humidity.

Are you able to share additional details on how you're testing these devices and the pass/fail criteria?

Humidity can impact likely of ESD, but I'm not aware of its impact on device performance. I'm also inquiring internally on this.

Thanks,

Drew

Hi Drew,

I am using them as a link over CAT6A double shielded cable. One side of the link is in a small chamber with desiccated air pumped in (but not pressurized, just slight positive pressure to keep dust out) and the other is in an open to atmospheric (office) air.

Large amounts of data are being passed over the link, but the pass fail criteria is just loss of lock on the deserializer. The serializer is in the low humidity chamber (~0-10% RH)  and the deserializer is on a board open to the air.

I know less than 30% RH is not great for electronics but I am looking for maybe some more variables to test than silicon lottery of parts and ESD robustness at what I assume is a small esd event building on the insulators ie plastic case of the part (just thinking out loud).

Any information would be great and I would be happy to discuss on a call 

Thanks,

Jeff

Hi Jeff,

Thanks for the additional setup details.  I'm checking in with some colleagues on this, although this seems to be an uncommon issue.

One thing I'd like to clarify with you is in these cases where the serializer is leading to loss of lock on the deserializer, are these serializer devices getting permanently damaged, or just experiencing a temporary performance issue?  Also, does the deserializer regain lock after loosing lock?

Thanks,

Drew

Hi Drew,

Thank you for the quick responses! I very much appreciate your timeliness. It seems to be a temporary condition as taking the board out of the low humidity it is capable of regaining lock. 

I have 2 units that survive the low humidity for 6+ hours without losing lock but then a handful (10+) of PCBs that cannot lock in low humidity.

I will be testing these units but just wondering if there is anything else I can adjust to make them more robust.

In nominal humidity 44% RH I can run for 100 hours without loss of lock and the same unit will lose lock in ~30 minutes of 3-5% RH. I do not believe it is a design issue or layout implementation due to the robustness in "normal" conditions.

Any and all information is welcome. If you need more information, I can continue to provide context.

Thank you,

Jeff

Hi Jeff,

I've talked with a couple colleagues on this topic.  Based on conversations I've had, we would not expect a decrease in device performance on either a silicon or package level based on a low humidity environment, so your observations are a bit puzzling.

Is there any change in temperature in your low humidity versus typical humidity environment?  Also, do you have capability to try testing these boards at different temperatures in your test environment?  I'm wondering if we might see some different behavior at hot/cold temperature.

Also, do you have the capability to look at the output signal from the serializer on a scope?  I'm wondering if the condition where the deserializer loses lock is due to signal somehow getting muted from the serializer, or some sort of signal integrity issue.

Thanks,

Drew

Hi Drew, 

I attached some of my findings. We have a packet counter that indicates loss of lock after a certain duration. The temperature is not changing during the test after settling at the noted operating point. There are no transient power spikes or changes in operation. The SerDes pair is sending data in a steady state.

I can adjust the temperature to be both hotter and colder so I can add those as variables quite easily.  I will test these to see if we can glean anything new from those tests

I don't have a fast enough scope capable/low capacitance probe to make an eye diagram of the CML lines to check signal integrity of the 312.5MHz unfortunately.   

The data presented below makes it seem like humidity is a factor but not binary given the longer duration of >7 hrs in enclosure before losing lock. It is confounding considering the same exact unit can run for >97 hours straight with no problem. 

Hi Jeff,

Drew is out of office today. He can get back to you with more feedback later this week.

Best,

Lucas

Can you share any temperature data you have recorded? Happy to sign and NDA or private message etc for highest reliability operation or any fringe conditions that are not recommended. 

Hi Jeff,

I'll check into this, although I'm not sure if I'll have much data given the age of this device.

Also, thanks for the details about your system setup and test results.

I have a couple more clarifying questions:

1. I've assumed that you're using the same deserializer board while testing these different serializer boards.  Is this assumption correct?
2. If you haven't already done this, is it possible to test a board in your enclosure and not pump any air into the enclosure (just use room humidity).  Although it seems unlikely, I'd like to confirm that something about the setup in enclosure itself is not causing loss of lock.
   Another experiment along these lines would be to test a board in your enclosure with the pump running, but just pump room-humidity air (non-desiccated air).

Thanks,

Drew

Hi Drew,

Thank you for walking through all this with me, hopefully we can drill down.

1. Yes all these tests are done with the same deserializer PCB.

2. I have done this test. Pumping in semi humid air seems to alleviate the problem which is why I am thinking humidity in this particular implementation on the PCB matters but I am speculating. I believe it is environmental, again speculation.

is there any timing change on the CML lines that could be influenced by temp? Is there a temperature where performance becomes marginal?

Thanks,

Jeff

Hi Jeff,

Thanks for the information.  I'm definitely hopeful we can work through this.

I'd have to see if I can locate temperature data specific to this part, but in general, I think we see more jitter on high speed lines at higher temperature.

I talked with one of our IC designers about this issue.  He wasn't aware of any impact low humidity would have on silicon, but he was wondering if this could impact trace impedance.  I haven't investigated this too much, but just one other idea I wanted to share for consideration.

Thanks,

Drew

Hi Drew,

I am running a test on the bench where I am slowly increasing temp to see if I can take a system from functional to losing lock.

I will publish data soon. 

Can you share the jitter vs temp data? 

Thanks,

Jeff

Hi Jeff,

I'm working on locating our validation data for this device and will get back to you next week.

Thanks,

Drew

Would the timing skew of a delta temperature of 15C cause a loss of lock?
Does the lock of the timing period reset when the device reacquires lock or try to use the ?

Thanks,

Jeff

Hi Jeff,

I was able to locate some data on this device.  What data rate are you using?

Over temperature, I see a trend of higher jitter at higher temperature, lower jitter at lower temperature.

The lock timing period resets when the device loses lock.

Thanks,

Drew

Hi Drew,

I am using 2.5Gbps with DC Balancing and Remote Sense enabled.

Do you have jitter in percentages or picoseconds associated with a temp change? My real question is can a temp change push the deserializer lock out of the acceptable period due to that jitter?

I just tested today using an enclosure with measured temp at 31C on the case of the serializer at less than 1% RH and that setup had a loss of lock after an hour. 

However running on the bench (previous test) with the fan, I can decrease cooling until same exact PCB, same thermocouple measurement reading 44.93C at 23.44% RH that does not lose lock for 45 hours.

Then turning off the fan in the above setup and test (after 45 hours working fine no loss of lock) temperature rises to 52.1C and it loses lock in 23 minutes. 

I can't explain this behavior. Just to confirm the same PCB is still functional, I have been running another test on the bench with fan on the PCB and the serializer measuring 34.7C at 27.5% RH the unit has been running for 5+ hours and counting. I expect this to run overnight without loss of lock.

Any ideas? 

Thanks,

Jeff

Hi Jeff,

Looking at the data I have, the maximum jitter that was observed at 85C was 49ps.  Average at 85C was 44ps, and average at 25C was 37ps.

Looking at the DS32EL0124 data sheet, it has a jitter tolerance of 0.3UI, or 120ps.  Just looking at DS32EL0421 output, it seems like the jitter shouldn't be a concern.  However, Dj from the CAT6A cable could be a contributing factor.  Without a way to measure jitter on the signal from the cat6, this is more challenging to determine.

Thanks for sharing the update, I don't have too many ideas at this point.

Are you doing any configuration to the DS32EL0421/DS32EL0124?  If so, can you share this?

I could also help review your schematic and layout if you think there would be value in this.  You can send me an E2E private message if you'd like to keep this off the public forum.

Thanks,

Drew