Design your Antec PSU

[pentium]

Re: Design your Antec PSU

By the way, Antec units have poor ventilation. better ventilation will prolong psus.

...that's why I proposed liquid cooling instead of fans.

[999999999]

Re: Design your Antec PSU

Since Antec doesn't design PSU, I wonder how effective anything we wrote would be. Antec may be able to specify some changes to the base designs used in their units but it would be good to know how much latitude they really have, to make this a more effective use of time.

This is not a knock against Antec either, it's just how the market works.

First priority should be lifespan at max rated load. I don't want independant regulated rails or active PFC until that area is addressed. I don't mean "addressed" like they thought they were addressing it already, I mean bullet-proof.

Naturally given the origins of the forum, better capacitors are a given. It should also be a given that any monitoring and shutdown circuitry has to be on a riser board as much as possible (meaning technically possible, every last discrete component down to the last resistor is not necessary), because it makes the necessary space available on the mainboard. I realize they do this, but felt it worth emphasizing again and to the extent of making it even shorter they might use SMT components instead of the tombstoned transistors, resistors, etc which reduces that much more the airflow through the area once the wiring bundles are there too. Airflow though this part of the PSU relates to upstream flow past the caps.

I'm not always opposed to 80mm exhaust instead of 12cm intake fan. While some are, and that's fine, remember there are those who may have atypical needs in some cases and that is a large part of why retail PSU sell at all (beyond the other obvious capacity and quality issues).

More weighting on the 12V rail. No more than 2-3 12V rails, I'm not concerned with the antiquated ~18A limits. Even better - a weighting dial or switch, to adjust based on current per rail requirements of the system. Even better if independent regulation but that at higher cost, I realize multiple price-points have to be met.

No TO220 diode packs, no ultrafast. Minimum 30A schottky on 3V, dual paralleled on 5V, dual or three parallel on 12V depending on 12V current rating for the unit. OVERspec'd.

No BJT transistors on the chopper, mosfets please.

Staggered length power connectors, some only need be 10" long (if that) but others progressing up to 30". Mobo leads don't have to have a sleeve around them, but they do need at least one nylon wire-tie every 4-6". Secure, reduce movement of the wiring harness more at entrance to the PSU casing.

Relating to available board real-estate, on units with the large ~12mm fan, don't let the space where some had 80mm fan go to waste by being empty, extend the PCB there because the amount of room on the board is getting really important on these ever rising wattage units. Likewise, don't be hesitant to go with a longer metal casing, it is obvious PS2 form factor has been outgrown, power density getting too high.

Consider putting a standoff near the middle of the PCB, or using a thicker PCB material as they start to bow a little with the heavier components on them.

Find a way to shift the fan further towards the wiring harness side of the PSU so larger HV filter caps can be used. It not only goes towards obvious factors but some less so, like reserve time when a well endowed system is powered from an offline UPS so it incurs a switchover delay it has to survive.

More robust 5VSB circuit, these days USB peripherals are everywhere, it would be nice if we didn't have to consider how much load is on that rail from a mistaken (or needed) motherboard 5VSB jumper setting, plus a little more overengineering would make it more resistant to the passive cooling when a system is in an off state.

These were only generic issues, some of which already met in any particular model, but most effective would be to take each PSU one at a time and if the basic design isn't changeable for cost reasons, then a better idea of what latitude there is to improve certain things per unit would be most effective.

[pentium]

Re: Design your Antec PSU

......
Bluh?
I got lost on that one.

[PeteS in CA]

Re: Design your Antec PSU

Hmmm ...

* Active PFC
* Current mode PWM (e.g. UC384X family)
* Separate +3.3V, +5V and +12V regulator-inverters for high-end and server types
* All major caps must be long-life types, low impedance for O/P caps, 105C types for I/P caps; only the three 'cons and Panasonic as vendors for all lytics
* MOSFET PFC and inverter switch devices
* Transformer cores made by major Euro, Japanese or US companies; O/P inductor cores Micrometals -52 material or -70 (or -60) (200C) material
* All O/P rectifiers rated for at least 2X the rated current of the respective O/Ps
* OVP on the +3.3V, +5V, +12V and +5V Stby O/Ps
* Individual current limit on the +3.3V, +5V and +12V O/Ps
* Temperature controlled fan; ball-bearing, major brand fans only (e.g. Comair-Rotron, EBM, NMB, Nidec)
* Heatsinks ... heavy metal is goodness

[chow]

Re: Design your Antec PSU

Antec may read this?

O-K - I used to install & recommend the product years ago, but not any more. Once you have ticked off your buying demographic, do you really think they are going to come back, even if you claim to have addressed the problems? I point to the big three "American" auto makers as an example.

Seems the designs are essentially decent, but somewhere along the way the beancounters got to the production model and nickel & dimed the process, or QA became a thing of the past. You expect buyers to keep buying cr@p? Buy another manufacturers product.

[linuxguru]

Re: Design your Antec PSU

> Active PFC

AFAIK, P-PFC is more durable. Nothing can go wrong except insulation breakdown, and that rarely happens with high-quality enamel. A-PFC has silicon, and lots of things can go wrong with it when inputs are out of spec - in particular, it doesn't handle extreme events like spikes well.

> MOSFET PFC and inverter switch devices

Again, I prefer BJT for durability, especially on the primary switches. BJTs can have very high Ucbo ratings, and BJTs don't necessarily fail permanently when Ucbo is exceeded once or twice. MOSFETs must have a sharp-clamping snubber, usually some kind of avalanche diode or similar for reliability, and the snubber diode has a tendency to fail with cumulative energy absorbtion.

Perhaps it might be worth re-examining a 70s concept: A MOS-BJT cascode switch, also known as Emitter turn-off. The lower MOSFET is used as the switching element for fast turn-off and easy gate control, and the upper BJT is used to block the switch-off voltage spike.

> All O/P rectifiers rated for at least 2X the rated current of the respective O/Ps

Difficult, but not impossible. It may be possible to use bifilar winding on the high-current rails, and split the output into separate 30A dual-Schottkys for each of +3.3V and +5V. But this will require 4 such rectifiers - which will be a tight squeeze on the secondary heatsink. Perhaps back-to-back mounting can help things along.

> More robust 5VSB circuit, these days USB peripherals are everywhere

Yup, have to agree. Also more robust primary switcher and control on the auxiliary supply - it's often left on permanently, must be able to withstand transient events like overload, external shorts, etc. on +5Vsb. Would also prefer a BJT switcher here rather than a MOS integrated switch like a TOPSwitch.

[999999999]

Re: Design your Antec PSU

It shouldn't be hard to make the O/P rectifiers double the sustainable current. Any time more than 2 would be needed in parallel, the supply ought to be a stretched form factor anyway, allowing longer 'sinks.

These paralleled regulators, along with the mosfets vs. BJTs, will be among the lesser evils to arrive at the industry's target 80% efficiency. Synchronous rectifiers would help too, but then we're drifting towards a niche priced product, particularly since this large a deviation would require another design instead of minor parts substitutions and a shifting of the layout if necessary.

IOW, I don't see Antec doing a custom supply, and one properly built shouldn't have to cost over $100 to support most enthusiast class systems. If we were talking sky's-the-limit, just skip a "PC" supply entirely, server class are often larger for the inherant benefits in having more available space. Towards this end, I think Intel is the one who needs to make changes first, a new spec for PSU that enlarges the casing as their 80% efficiency guideline is only a stop-gap measure, the improvement in efficiency is lower than the increase in power consumption over the past decade we've had ATX.

[PeteS in CA]

Re: Design your Antec PSU

Active PFC gets rid of the voltage selector switch. Big advantage, IMO. I've seen the I/P current waveforms for active and passive PFC and I'm surprised passive PFC complies with the harmonics regs. One stroke of a bureaucrat's pen ...? And then there's that 50/60Hz boat anchor ... errr ... inductor.

MOSFETs are necessary for higher switch frequencies, which enable smaller magnetics, higher power density. Driving BJTs requires more power and you have to be careful about how low the beta of the BJT gets at higher current. Some drive schemes get marginal at low-line. OTOH, proportional drive, which is not sensitive to the line voltage, require care in turn-off and can lead to weird self-oscillation during brief line interruptions or sags. Inequalities in Turn-On, Turn-Off and storage times require care in inverter circuit designs.

I believe fast-switching (for BJTs) power BJTs are likely to become difficult to source - second or third tier companies. First-tier semi companies are focusing development $$ on MOSFETs. Newer, better MOSFETs are continually being brought into production; various vendors' versions of Motorola's late 1970s MJE1300X series are still prime parts. In that time frame MOSFETs have gone through quite a few generations (IR's planar technology parts are at Gen 10, with newer trench gate parts now in production), moving from economically impractical lab curiosities to mainstream parts. I "cut my teeth" in power supplies, using Mot's 2N654X and MJE1300X BJTs, custom current-mode schemes and the SG352X family voltage-mode PWM chips. For the most part, MOSFETs and current-mode PWM ICs (e.g. the UC384X family) are the present and future.

Wanting rectifiers rated for double the current of the respective O/Ps is probably a challenge, even greedy, but when you see 16A rectifiers used on a 14A or 15A O/P, .

[gonzo0815]

Re: Design your Antec PSU

Well, i think there is no doubth that the A -PFC is a matured technology.
Todays MOSfets and IGBT are used for heavy duty applications in very hazardouse environment and they can coupe with all the problems, a big industrial plant has to battle with.

A PSu for PC systems is usually not exposed to that bad mains, and with the right parts you can make any A PFC circuit at least transient proof like any other p-PFC psu.

Shure, if something is going wrong, the repairtech would have to coupe with the new circuits, but that is the price, we have to pay for any andvance in technology.

From my point of view, P-PFC is the desing of the past and i will certainly do not buy something with P-PFC anymore, unles there is a real price difference (wich i can`t see ATM when comparing quality PSu`s atm).

[999999999]

Re: Design your Antec PSU

We might claim there isn't a price difference among many quality PSU, but what if improvements to Antec PSU meant you found some of their lesser PSU improved upon enough to be an acceptible quality level?

What if Antec quality didn't matter in their next-gen PSU, it was only a matter of selecting the wattage a system needed?

The ideal to some people is to be able to buy a new case and use the PSU that came with it, and Antec is popular enough that it's common to find case w/PSU offered for $50-85 USD. If they put more thought into some areas I'd be a lot more likely to use such a bundled PSU on a typical system (anything "PC" oriented except a higher end gaming system). To Antec that is the difference between making a sale or losing one.

[gonzo0815]

Re: Design your Antec PSU

Well, i think buying a case with PSU will never do well in this market.
People who are buying a case with psu are usually on a bargain (may be except of a few experts..). And there you will usually not be the topseller, if you sell good quality at an higher price. So you build a good case and add a cheap psu (may be the cheapest you can source) and you are able to compete very well.

From my point of view, i can settle with a price of 50€ for a quality psu with good caps (only one of the better one, so no Teapo etc.) even for cheap systems. May be there is still the need for an 30€ psu with second grade caps like Teapo, Ost.....but i doubth this.

[Sparky]

Re: Design your Antec PSU

Honestly, there are only 2 things I'd like to see antec change:

1. NO MORE of those fuhjyyu crap caps! Put something good in them.
2. Fix that wiring mess that exits the PSU Almost all the Antecs I've had (which is several) have always had a tangled mess exit the PSU before going to the connectors. Would be nice for the wires to get separated out neatly then the connectors put on so that there isn't this huge knot right there. If other companies can do it, so can antec.

Other than that, I think the truepower II line was fine. Good stable rails, quiet, fairly inexpensive - couldn't complain about that. Just fix those two problems and I'd be happy. Active PFC would be nice as well but for a cheaper-priced PSU it isn't needed.

[willawake]

Re: Design your Antec PSU

Anyway antec is not very well distributed here in Greece. If they want good distribution then they should get their products in this chain of stores. http://www.plaisio.gr/

and they just did

[tazwegion]

Re: Design your Antec PSU

Are you secretly working for an Antec thinktank willawake? LMAO

Seriously, my 2 biggest beef's with Antec PSU's are the capacitors they choose & the airflow created by the fans fitted, I understand some people prefer quiet PSU's but a user control for fan RPM would be nice too!

My (Samxon recapped) TruePower380 pulls like a train with minimal voltage fluctuation on the rails (even during 100% load), so I'll keep buying Antec and recapping them... but it'd be really nice if I didn't have to, you know?