Buck Output Capacitor Selection for Long-Term Harsh Environment (85C / 85% RH)

[tinfever]

I'm working on a project which I'd ideally like to be able to function for 10 years of 24/7 operation in indoor but non-climate controlled environments, which could potentially include high humidity (85+% relative humidity, 50C ambient temperature). The board is also extremely space constrained. The project includes several buck converters (SIC451) which I'm trying to select the output capacitors for. I'd hoped to get away with only using MLCCs but during testing I've found that even with the correctly calculated amount of output capacitance, the buck converter output is unstable (the output is correctly 5V but with a 1.8V p-p oscillation on it at 100kHz). I've found adding more low-ESR capacitance stabilizes the output, so now I'm trying to pick the right type of capacitor to use.

Given a 50C ambient temperature, I think it's possible the capacitors will be operating long-term in a 85C high humidity conditions.

I suppose I could try to fit more MLCCs, although I'm concerned that I can't calculate the amount of capacitance required to stabilize the buck converters.  Even if I find it takes 5x 10uF MLCCs to stabilize it on the bench, and then I add a 6th MLCC for safety margin, I have no idea if 5 years from now the aging, temp variations, and the 20% value tolerance might mean it becomes unstable again. Also, this board will have some flex (<2mm deflection over 250mm span) under normal use, so any MLCCs have to have soft terminations and then I'm still hoping for the best.

At first I thought I could just barely fit in a 1206 100uF 6.3V tantalum-polymer capacitor but after further research I've realized those aren't a good fit for the high humidity requirement. Most manufacturers don't provide any endurance rating for humidity while under bias, and the AEC-Q200 rated parts that do are only rated for 1000 hours at 85C/85% RH. Kemet is also playing games with the endurance rating by rating the 2000 hour normal endurance at 2/3 rated voltage on >= 125C rated AEC parts, which is almost all of them. So who knows what the real endurance of a 6.3V part used at 5V is.

It seems like the safest option would probably be aluminum-polymer capacitors. Nippon Chemi-con has some rated for 15000 hours at 105C, and 1000 hours at 85C/85% RH under load. Although, 1000 hours is a long way from 10 years (87,600 hours). Panasonic has some SP-Caps rated for 1000 hours at 85C/85% RH but at no-applied voltage, which is useless  .

I haven't looked into aluminum electrolytics much but they don't seem to offer any endurance rating for humidity, which means they are maybe unaffected by it? If I could find some with low enough ESR, perhaps this would then be best?

The issue with the aluminum-polymer and electrolytic options is that I don't think I have space for any them, even the 7.3x4.3mm package size type. I could go to double-sided SMD and put them on the back, but I've already gone to great lengths to keep all the SMD parts on one side, so this would increase the assembly cost by some unknown factor.

Does anyone have any suggestions? Are my requirements effectively impossible?

[T3sl4co1l]

What ceramic capacitor type was it?

Polymer types degrade in a similar way to electrolytics, but for inverse reasons: in the electrolytic, it's solvent evaporating through the seal; in the polymer, it's humidity diffusing in, degrading the polymer.

Tim

[tinfever]

What ceramic capacitor type was it?

Polymer types degrade in a similar way to electrolytics, but for inverse reasons: in the electrolytic, it's solvent evaporating through the seal; in the polymer, it's humidity diffusing in, degrading the polymer.

Tim

The existing MLCCs are X7S 0805 10uF 16V.

If that is the degradation mechanism, does that mean that, after ceramics, electrolytics would have the greatest endurance under these conditions (if there is no lifetime degradation due to humidity but obviously still lifetime vs. temperature considerations)?

[kimballa]

The damping factor added by the non-zero ESR of electrolytic or tantalum capacitors can be modeled as a small resistor in series with the capacitor.

If you are OK using chip MLCC capacitors, you could add connect a 10uF MLCC to the converter output with a small chip resistor (25-250mOhm) to simulate an electrolytic, or a slightly larger resistor (0.5-2.5 ohm) to simulate a solid tantalum.

On that note, I don't think you mentioned tantalum in your analysis. Do the automotive-grade options there maybe meet your needs? They're more compact than electrolytic, though they'll raise the bom cost.

[NpR]

Hello!
What sort of housing options are available? Humidity can be kept out relatively well with proper housing and even potting the electronics. Look into how they built automotive electronics in the late 80's and 90's, most of that stuff doesn't have any problem functioning to this day. They used waterproofing, clever positioning and potting (epoxy/silicone) to protect circuitry from the elements. In most cases it is easier to keep moisture out than electrolyte in, especially in those sort of temperatures.

BR
Nick

[tinfever]

The damping factor added by the non-zero ESR of electrolytic or tantalum capacitors can be modeled as a small resistor in series with the capacitor.

If you are OK using chip MLCC capacitors, you could add connect a 10uF MLCC to the converter output with a small chip resistor (25-250mOhm) to simulate an electrolytic, or a slightly larger resistor (0.5-2.5 ohm) to simulate a solid tantalum.

On that note, I don't think you mentioned tantalum in your analysis. Do the automotive-grade options there maybe meet your needs? They're more compact than electrolytic, though they'll raise the bom cost.

Sorry, I wasn't clear in my post. Ideally I would just use more MLCCs and wouldn't need to add any additional series resistor since there is no lower limit for ESR in this circuit. There is a higher limit on the ESR though, in that adding a 250mOhm ESR 68uF electrolytic in testing didn't fix the output oscillation but adding two more 10uF MLCCs did. I've just been thinking I don't have room for more MLCCs, although since I'm going to have to find room for any solution, MLCCs might make the most sense. I was also thinking I'd sleep better knowing there is some bulk capacitance even if the calculations and testing indicate it is unneccesary.

As for the tantalums (the non-polymer kind), they admittedly scare me from what I've read. I guess the output of a buck is technically protected from voltage transients and from the the upstream high current power supply, so perhaps it would be safe with sufficient voltage derating. I think the ESR is too high though.

Hello!
What sort of housing options are available? Humidity can be kept out relatively well with proper housing and even potting the electronics. Look into how they built automotive electronics in the late 80's and 90's, most of that stuff doesn't have any problem functioning to this day. They used waterproofing, clever positioning and potting (epoxy/silicone) to protect circuitry from the elements. In most cases it is easier to keep moisture out than electrolyte in, especially in those sort of temperatures.

BR
Nick

Do you have any source recommendations where I can learn more about how they do it in automotive electronics?

Conformal coating might be possible but I'm guessing it would have to be manually applied due to close proximity (and sometimes overlap) between SMD parts and connector bodies. A proper sealed housing wouldn't really be possible because the connectors would have to protrude through any housing, and these connectors aren't designed for any sort of sealing (and the connector type can't be changed).