High reliability 5V 1A power supply (no electrolytes)?

I would recommend you use PI Expert Design Tool and start with a design similar to RDR-158. You can then start lowering the input capacitance and see how low you can go while still maintaining a VMIN value of 70VDC. You can then get away by using a film capacitor as the input capacitance.

Thanks for the input!

I tried PI Expert, but it seems the software does not (yet) support LinkSwitch-II devices. I like the RDR-158 because of the CV/CC charasteristic which would be beneficial (although not absolutely required) in our project. This is why I would prefer using LinkSwitch-II over LinkSwitch-CV. Another issue for me with PI Expert is that it always uses an Y-capacitor between the primary and secondary, which would not be possible because of safety rules in our project. Or can it be removed?

However, I made some quick testing with PIXls Designer using LNK616PG like in RDR-158. It seems like the input capacitance can be lowered from default ~15 uF to about 1.5 uF while still maintaining VMIN > 70 VDC.  Not quite as low as with the LED drivers, but low enough to use film capacitors of sane package size. But there must be a catch? Surely the reference designs do not provide up to 15 uF of input capacitance just for fun? What kind of challenges would low input capacitance present? Poor EMC filtering performance or poor stability?

There are a lot of questions - but the project will be very interesting! I am also interested in other aspects of reliability than the capacitor type. If cost would not be a major issue, is there anything else that could be done to the RDR-158 to maximize it's lifetime? 

 I would recommend using PI Expert for a new design using the TinySwitch-4 since you are looking for a low component count so that you can better control your design.

The MOSFET in a power supply has a maximum duty cycle value associated with it. Small input capacitance implies a small VMIN which in turn implies that the duty cycle at VMIN is high. Therefore, one of the parameters that control your input capacitance is your switch duty cycle. At high line, you can get away with a smaller input capacitance since your minimum input AC voltage is high (around 195VAC), which is not really possible at low line.

Yes, I understand that lower input voltage causes higher duty cycle and more average current through the MOSFET. My point was: do you think that being near the allowed VMIN limit causes stability/reliability issues? How much headroom do you recommend for the input capacitance? 

I originally did not consider TinySwitch-4 because of two reasons:

1) Y-capacitor in all example designs. As I described we cannot use that because of safety standards/rules.

2) Optocoupler. Just as with electrolytic capacitors, I am worried by the limitations of their lifetime. 

But maybe, if we can find a way without the Y-cap and a very very reliable optocoupler, then just maybe it would be possible. I am still investigating the contribution of different components to the total lifetime. If you have done any such analysis for your reference desings, it would be very interesting to see. At the moment, most studies and experiments seem to result in the optocoupler and electrolytics being the weakest links. If we could eliminate them both, it would be the ideal first step in the design. 

Just a crazy idea: could it be possible to use a magnetic coupler or a pulse transformer with TinySwitch-4?

Any comments or experience from others about this topic? It's surprising how unpopular discussion topic the reliability is (also on other forums). :) 

It would be great to have some insight into the reliability of the Power Integrations ICs, especially the LinkSwitch-II family. I have found some FIT/MTBF numbers from this forum for other families and they seem very good (around 1 FIT). But in order to really use these numbers, we would need some more information about the tests and calculations. Many semiconductor manufacturers provide some basis for the reliability data, look for example this: http://www.micrel.com/_PDF/qualrel/ASSET_1.2.pdf . Something like that would be very helpful! 

Another interesting question: is there a technical benefit from using optocoupler for feedback compared to an extra transformer winding? As I see that you use both methods in various ICs. I guess the optocoupler must have some benefits (other than just a simpler transformer)?

"My point was: do you think that being near the allowed VMIN limit causes stability/reliability issues? How much headroom do you recommend for the input capacitance?"

You will encounter no stability issues if you maintain the VMIN recommended by PI Expert in your design.

"I originally did not consider TinySwitch-4 because of two reasons:

1) Y-capacitor in all example designs. As I described we cannot use that because of safety standards/rules.

2) Optocoupler. Just as with electrolytic capacitors, I am worried by the limitations of their lifetime." 

The Y-cap helps reduce common mode noise and the design can be realized without it. We do not have an analysis on component lifetimes however in case you want to eliminate the optocoupler, then the LinkSwitch-HP is an option you might want to look into. 

Hi,

The LinkSwitch-HP looks very interesting indeed. Do you plan on adding it to the PI Expert software?

I am not sure if we have such plans presently; sorry! :-(