Is it okay to bypass atx psu with film caps?

[ss627]

Is it okay to bypass atx psus with film (polyester, polypropylene) capacitors?

Some old units have a lot of space inside the enclosure if you desolder some unused wires. I vaguely remember reading something about avoiding ceramic to bypass but I dont remember anything about plastic.

[linuxguru]

Re: Is it okay to bypass atx psu with film caps?

There's not much of an advantage - these rails aren't powering high-end analog circuits. If the audio section on the motherboard is well-designed, it will have its own DC regulators for the analog section. The digital sections of the mobo are anyway relatively immune to the 50 kHz-1 MHz hash that might get bypassed by film caps.

[Logistics]

Re: Is it okay to bypass atx psu with film caps?

There are articles by Walter Jung where they state that ceramic capacitors made an audible difference even when used in the power supply for the amplifier they were working on, but believe-me-you that this is likely extremely exotic amplifiers (which doesn't mean they are complicated, it just means they use super-high-quality components) and has nothing to do with switch-mode PSU's in PC's. I once humored the idea of bypassing a PC PSU with films, but it's a complete waste of time and money. The best you could do is upgrade your PSU and motherboard with premium electrolytics.

[999999999]

Re: Is it okay to bypass atx psu with film caps?

Yes, it is ok, but I have to ask why? A ceramic has a more dynamic frequency range and performance in the frequencies we encounter. So the answer is, if all you have lying around is film caps, by all means use them, but if you are buying parts do not consider what is best in audio circuitry if it's a computer PSU supply circuit.

Do not avoid ceramic, they are the best bypass capacitors available. Only slow/old circuits can use film caps, or ones very exotic using many layers to lower impedance enough that films can be staggered in value to support the modern higher-speed components we use today.

What you have not asked is, what is necessary for the best result with the specific subcircuit, and if a PSU, what it is powering, and what the budget is, and what parts you have on hand, and what benefit vs additional cost you want to bear.

Generally speaking, there is no reason at all to prefer film caps in a computer PSU. Remember that the most significant caps for any digital chip are those immediately adjacent to it, lowest impedance next to it on the PCB for decoupling (except audio circuits when it is a coupling instead of decoupling cap), so there is little point to having film caps upstream for any circuit when there are ceramics decoupling closer to the powered IC.

The real question is, why do you care? As Logistics noted, there isn't much need to use high frequency decoupling within the PSU, because the frequency of noise coming from the PSU is fairly trivial compared to the noise pickup along the supply harness to the parts, and especially the noise produced by those parts themselves.

In other words, if you want to improve a switching PSU, your best bet is adding capacitors with lower impedance at the switching frequency than what are already installed. That could mean only an electrolytic upgrade, could involve adding a parallel, additional electrolytic, or a tantalum, even a larger value ceramic sometimes... but practically never is a film cap called for unless the immediate next stage is an audio subcircuit, no re-regulation or decoupling after that point (which would already be a bad design, if it is that subject to noise it would already have appropriate decoupling in it's immediate vicinity on the PCB).

[ss627]

Re: Is it okay to bypass atx psu with film caps?

I have a lot of components and film caps are the bulk. Yes I realize they're not practical due to size and energy density. I do keep electrolytics but I sell 3-5 years old stock on ebay and order fresh date codes from digikey or mouser.

I was just curious and wondered if atx supplies were built the way they are due to cost cutting and space limitations. They are pretty lightweight and sparse looking considering the amount of amps they have to provide.

[999999999]

Re: Is it okay to bypass atx psu with film caps?

I spoke too soon about a couple of areas film caps may be used in a typical ATX PSU, to get higher voltage tolerance than a ceramic would allow in the high voltage side, and a timing capacitor for the regulator frequency.

Beyond those, yes PSU are often built to price-points and with limits in available space, but this is not why they avoid film capacitors necessarily. There is an ideal in a design and then there is a practical requirement. The ideal isn't necessary for proper function, it powers digital circuits which do not need the ultra clean output a tweaked linear supply would provide.

The requirements are generally met with decent quality electrolytics alone and any further decoupling of supply lines would be better at the powered component and that is also how the powered components are designed to have their own decoupling such that, if properly designed, they should run acceptably from any PSU operating with the ATX specification.

In other words, you could add film caps to an atx psu if there were room and it may result in cleaner power, but unless the PSU was misengineered in the first place there is no realized benefit to doing so. In areas where a fast small value capacitor is required for the PSU to function properly, a ceramic works fine so if a film is larger and costlier there is no reason to use one.

[ss627]

Re: Is it okay to bypass atx psu with film caps?

Thanks for the info. When I look at electronics, I place extra importance on durability and over-engineering instead of only "good enough". My curiousity mostly stemmed from reading a bit on military and medical standards and applications. I remember reading an article explaining the reasons for avoiding large value/sized ceramic capacitors in smps environments is weakness against vibration leading to micro fractures and temperature fluctuations.

[999999999]

Re: Is it okay to bypass atx psu with film caps?

Me too re: durability and over-engineering, but in that case you'd want banks of paralleled solid caps instead of low value film caps bypassing 'lytics, or to have a supply engineered with higher frequency switched fets, less solid cap capacitance, and a few ceramics. Either way basic things count a lot... non-defective caps, high quality dual ball bearing fan, surge protection built in or before the PSU. Given these things the PSU should last longer than is useful.

In some types of circuits, yes sensitivity to vibration, and temperature, are issues with ceramics. They don't see significant vibration in a PSU and with the exception of the timing cap for the regulator IC, a drift in value with temperature is not very important.

The best thing to do to ensure durability and over-engineering is use over-rated transistors, diode bridges, excessive heatsinking and cooling, and the lowest ESR capacitors you can find. The thing about bypassing these with film or ceramic is it has such diminishing returns because you are bypassing with a mere 0.00nnn to 10uF at most which is not enough to keep significant ripple current from the electrolytics. If it were a different type of circuit instead of a PSU producing dozens of amps, such a low capacitance value from a film or ceramic would be more significant... and yet in such a circuit the electrolytic wouldn't be so stressed either.

Basically we only see cap failures when it's an overall bad design, the PSU is overloaded, or the caps were just not suited for the circuit. Take a decent cap and you'll never notice a gain by tacking on a low value film or ceramic in parallel, except perhaps on a scope you may see some smoothing of high frequency noise which as mentioned previously isn't much of a factor as the powered parts are designed to accept the noise we expect a switching ATX PSU to produce.

The biggest issue I see with capacitors these days is when manufacturers only allocate 10 or even worse 8mm diameter space on the PCB so they aren't able to pick high quality caps with the capacitance the PSU needs to meet it's ratings long term. In my opinion if you see a questionable cap that is smaller than our best/popular ones with the same voltage, capacitance, ESR rating (or at least close), it's not likely to last long term unless the PSU isn't loaded near it's implied capacity.