Webench is great tool, but what about comparison for reliability and ability to pass EMI?

Dan,

For an outdoor industrial application, if reliability is the key care-about, then in reality the environmental stresses (temperature, humidity, shock/vibe, immunity to electrical surges/transients/external interference) and the PSU mechanical design (case, PCB type and design, shielding, etc) are probably the most important factors. The PCB's will see severe mechanical stress at the day/night/seasonal temperature extremes, and the thermal expansion mismatch on the 2 sides of single-layer PCB will add more stress.

The reliability of the PSU controller such as the UCC28710 should be well down the list of concerns. For the PSU design, for better reliability, you should choose components everywhere that are adequately rated, and apply generous derating on worst-case voltage/current/power/temperature stresses to increase the lifetime and reliability. One big advantage of using a PSR controller like the UCC28710 is that there is no feedback opto-coupler - so the PSU is no longer subject to the long-term ageing effects such as opto drift.

The output capacitors are possibly the shortest lifetime components in the design. They need to be rated to absorb the high ripple current at the secondary winding of the Flyback transformer. Typically electrolytic caps are sued because they have low esr, high ripple current rating and are efficient from a cost and size perspective. However, they do have relatively short lifetimes, depending the specific current/power and temperature stress. The lifetime can be extended by reducing the current/power stress, and the ambient temperature. By using several capacitors in parallel, the ripple current can be spread among them, reducing the individual cap stress. Most cap manufacturers provide lifetime info vs ripple current and ambient temperature. The PSU internal ambient temperature depends on the size and on the case mechanical design, as well as external ambient - for outdoor applications, the case will probably be sealed to prevent moisture and dust ingress, and if exposed to direct sunlight, the internal ambient can probably get very high. So despite current derating, the internal ambient temperature may already be too high to achieve the desired lifetime.

Polymer or solid electrolyte type caps may be a better choice. But be careful not to use a total capacitance value that is too low (even if the ripple current rating is plenty adequate), since the PSR control loop needs a min amount of output C to keep the loop stable.


For EMI, a pass with good margin is readily achievable with the right input filter design, PCB design and transformer design. The UCC28710 includes oscillator frequency-dithering and valley-switching, both of which help to reduce the PSU emissions to help with EMI compliance. However, EMI is very difficult to model, especially common-mode paths and effects which depend so much on the PSU construction - this makes it beyond the scope and capabilities of modelling packages like Web-Bench.

For good EMI, the internal transformer layer structure and arrangement is important to minimise the CM coupling from primary to secondary. Make sure the high dv/dt nodes (such as the primary FET drain) have minimal Cu area to reduce capacitive coupling. Make sure high di/dt loops enclose minimal area, to reduce radiated noise. Use snubbers to eliminate any sustained ringing - convert to a critically-damped impulse with the right R-C snubber - change from a unique high-amplitude single-freq in the freq domain, to a wider-band lower amplitude spread of frequencies. Make sure any large surface area high dv/dt radiators are grounded - heatsinks, transformer core, etc.


At present, WebBench does not specify or recommend the transformer design/implementation, just the required inductance and turns ratio. However, there are several TI reference designs available using UCC28710 for 5 V/2 A outputs. All of these will use an off-the-shelf or custom-designed transformer for Wurth or another magnetics vendor, so the same transformer can be sued, or the vendor can work with you to tune or customise to meet your needs.

Suggested reference design from the UCC28710 product folder:
85-265 Vac input, 5 V/2 A USB output - http://www.ti.com/tool/pmp9202
85-265 Vac input, 5 V/2 A USB output - www.ti.com/.../PMP11185 - similar but uses different transformer

Suggested TI-Designs (using the TI-Designs parametric search):
85-265 Vac input, 5 V/2 A output - http://www.ti.com/tool/PMP9561 - Uses UCC28730, newer IC, same family as UCC2871x, acheieves zero standby power (< 5 mW)
85-265 Vac input, 5 V/2 A output - http://www.ti.com/tool/PMP8363 - Very high density compact design using UCC28710


I hope this info helps you out.

Regards,
Bernard

Dan,

1. I will send the files to you by email. I need to touch base with the original designers to make sure I am using the correct/latest versions.

2. I don't know if we have any designs with only ceramic output caps, I will check. There may be issues ensuring sufficient cap value to satisfy loop stability requirements. I think that there are some designs using polymer caps (like Sanyo Os-Con etc) are thye suitable for you? For stability, you will need to meet Cout > (100 * Iocc) / (Vout * fmax) where Iocc is the output CC-mode current limit, and fmax is the design target fmax at max power (i.e. at the CV/CC operating corner). This may make the use of ceramics prohibitive, esp if the roll-off in cap value is taken into account (i.e. the change in effective C value with temperature, applied DC voltage bias and AC voltage ripple - all can be very significant for ceramics).

3. In WebBench, if you click on a part, e.g. the output cap, a small pop-up window will appear, showing all the relevant parameters for the part. At the top of that window, there's a grey box with the text "Select Alternate Component", click on that to select a different part from the WebBench library. A list of suggested alternatives (if available) will be provided. It does not allow you to arbitrarily change the part, or to define your own new custom part.

4. All Flyback power supplies will experience inrush current at startup, regardless of the chosen controller. This current charges the bulk storage cap. The only limitation on the current is the series impedance of the AC cord, EMI filter chokes, etc. So at high line, this inrush current can be quite big. Some applications require this current to be kept below a specified limit (e.g. maybe 50 A max). If so, then an inrush limiting NTC resistor is often used. The NTC (negative temperature coefficient) resistor has a high resistance value when cold, to limit the charging current into the bulk cap. Under normal loaded operation, the I^2R dissipation in the resistor will cause it to get hot pretty quickly, causing its resistance to fall, so it dissipates less power. Clearly, when hot it will not be as effective in limiting the inrush current (e.g. hot re-start). Ideally, the inrush NTC should be as small as possible so that it heats up as much as possible to reduce its resistance and dissipation as much as possible. But inevitably, this inrush NTC will dissipate significant power and will impact the overall power supply efficiency.

5. WebBench is intended for functional evaluation of the IC, and suggests the best external R's, C's, turns ratios, mag inductance etc to meet the target electrical spec. CM EMI is very difficult to model or predict, since it depends so much on internal transformer construction/implementation, PCB layout, mechanical design etc. So any EMI filter components in WebBench are usually only indicative of what could be used. The final power stage design will dictate the EMI filter required. With a good transformer design (with internal shielding and CM mitigation steps), a CM choke may not be required, depending on many other factors. However, it would be good practice to include provision for a CM choke in your design - in most cases it will be needed.

I will get back to further on points #1 & #2 above.

Thanks,
Bernard