ColdWatt

[narrasuj]

Here's a link for a supposedly new technology that reduces the heat produced by power supplies:

http://weblog.infoworld.com/sustainable ... owers.html

If this is actually true and trickled down to consumer level power supplies, this would be exciting news, IMO.

[disphenoidal]

I don't see what's so revolutionary here. It sounds like a slightly more efficient power supply. Their figure says they can generate 200W DC using an additional 25W for power conversion. I'm ignoring their figures for air-conditioning because we have no way of knowing how efficient the air conditioners are, and how much energy the air-conditioners use dealing with heat from other components. Anyway, no consumer level hardware uses air conditioners. So, 200W / (200 + 25) is ~89% efficiency. Definitely good, but not revolutionary.

I don't know what kind of efficiencies server power supplies normally provide, but many consumer supplies are almost as efficient. They claim existing supplies are 200W / (200 + 96) ~ 68%, which may or may not be true.

[Oleg Artamonov]

They claim existing supplies are 200W / (200 + 96) ~ 68%, which may or may not be true.

68% is the minimum allowable efficiency as stated in the EPS12V Power Supply Design Guide.

For the ATX12V 2.2 PSUs required minimum efficiency is 72% at typical load and 70% at max load, recommended efficiency -- 80% and 77% respectively.

[star882]

Digital power conversion technology is not that new. In fact, virtually all new PC power supplies are based on digital technology. The way it works is by the mathematical principle of integration. A digital signal made up entirely two voltages (usually ground and a certain input voltage) will become a steady voltage of intermediate value when put through a low pass filter. Use some digital logic along with a voltage sensor and it's possible to regulate the output voltage. It's pretty complicated but it works very well. You can even buy DC/DC converter modules that use digital technology. There are even a few designed for retrofitting analog power supply modules. I have installed one in my breadboard power supply and it greatly improved regulation at high loads.

BTW, digital audio amplifiers use very similar technology. But instead of holding the voltage at a preset value, it can jump from value to value very quickly in a virtually unpredictable manner. However, they're actually easier to design as the load (speakers) is constant rather than wildly varying. Digital power inverters (like the ones used in hard drive motors, ECM fans, and UPSes) also use that technology. In all cases, the idea is to improve efficiency by operating transistors at the extremes (cutoff and saturation) whenever possible, minimizing efficiency losses.

[Oleg Artamonov]

A digital signal made up entirely two voltages (usually ground and a certain input voltage) will become a steady voltage of intermediate value when put through a low pass filter

This is called "Switching Mode Power Supply" (SMPS).
a) You don't need any digital logic to control SMPS, most SMPS/PWM control ICes are analog ICes, not digital. However, there are also digital SMPS/PWM controllers, which offers some advantages, both in efficiency, cost and power supply design process.
b) All and every PC power supplies use SMPS/PWM technology since original IBM PC, but almost (?) all modern PSUs use analog SMPS/PWM controllers.

Technologies that ColdWatt PSUs may use
* digital SMPS/PWM controller;
* high switching frequency;
* synchronous rectifier.

[disphenoidal]

I looked at the wikipedia page for Switched-mode power supplies, but I don't understand a few things.

As I understand it, the switching regulators regulate the output voltage by switching the output of a transformer on and off. This improves efficiency over a linear regulator like a 7805, which wastes power by running a transistor in its linear region.

However, why is the AC input rectified and the chopped before being sent to a transformer? Why not simply transform the AC to the desired voltage, then rectify it? How does adding this extra step improve efficiency?

[Oleg Artamonov]

Why not simply transform the AC to the desired voltage, then rectify it?

http://www.xbitlabs.com/articles/other/ ... ology.html
"First, the low frequency of the alternating current in the power grid (50 or 60 hertz depending on the country you live in) calls for a big and heavy step-down transformer – a 200-300W transformer would weigh several kilograms"

[star882]

A digital signal made up entirely two voltages (usually ground and a certain input voltage) will become a steady voltage of intermediate value when put through a low pass filter

This is called "Switching Mode Power Supply" (SMPS).
a) You don't need any digital logic to control SMPS, most SMPS/PWM control ICes are analog ICes, not digital. However, there are also digital SMPS/PWM controllers, which offers some advantages, both in efficiency, cost and power supply design process.
b) All and every PC power supplies use SMPS/PWM technology since original IBM PC, but almost (?) all modern PSUs use analog SMPS/PWM controllers.

Technologies that ColdWatt PSUs may use
* digital SMPS/PWM controller;
* high switching frequency;
* synchronous rectifier.

Most new power supplies are actually hybrid power supplies. The sensing/feedback circuit is analog, but the PWM is mostly digital (some logic circuits with comparators sensing analog signals). However, most new UPSes are even more sophisticated, using an ASIC to control the power electronics. Of course, it makes a lot of sense in a UPS, as a single ASIC, in high quantities, is much cheaper than the complex discrete logic and opamp circuits.
Now, there are some power supplies that are "mostly analog". That implies that the there's a little bit of digital circuitry (often just from the comparator to the power MOSFET) to improve efficiency. (A LM393 along with a power MOSFET, a fast recovery rectifier, and inductor, and a few resistors and capacitors can make a very efficient power supply using only one of the comparators in the LM393!) Of course, they become harder and harder to design for high voltages. But for low voltage and low power, they work great. In my homemade router, I will use a LM339 for two of the low power circuits (5vsb and 12v) and some power supply control functions.

[Oleg Artamonov]

Now, there are some power supplies that are "mostly analog". That implies that the there's a little bit of digital circuitry (often just from the comparator to the power MOSFET) to improve efficiency. (A LM393 along with a power MOSFET, a fast recovery rectifier, and inductor, and a few resistors and capacitors can make a very efficient power supply using only one of the comparators in the LM393!)

LM393 and other comparators are classified as analog circuits, not digital. "Digital PWM controller" means ADC + DSP + digital PWM, see here or (more detailed) here.

Modern PSU don't use LM393 as well as any separate opamps, they use specially designed PWM controllers such as TL494, KA7500B (actually, TL494 clone), CM6800, ML4800CP, FSDM311 and so on. All named ICes are analog, but there are digital PWM controller ICes which combine A/D converter with DSP, programmable memory and PWM control circuit.

[star882]

Now, there are some power supplies that are "mostly analog". That implies that the there's a little bit of digital circuitry (often just from the comparator to the power MOSFET) to improve efficiency. (A LM393 along with a power MOSFET, a fast recovery rectifier, and inductor, and a few resistors and capacitors can make a very efficient power supply using only one of the comparators in the LM393!)

LM393 and other comparators are classified as analog circuits, not digital. "Digital PWM controller" means ADC + DSP + digital PWM, see here or (more detailed) here.

Modern PSU don't use LM393 as well as any separate opamps, they use specially designed PWM controllers such as TL494, KA7500B (actually, TL494 clone), CM6800, ML4800CP, FSDM311 and so on. All named ICes are analog, but there are digital PWM controller ICes which combine A/D converter with DSP, programmable memory and PWM control circuit.

Technically, the output of a comparator is digital as it ideally is in one of two states: either full voltage or no voltage.
As for DSP-based control circuits, that's exactly what most modern UPSes use. For all the functions that must be implemented in a UPS, they found that it's much cheaper to make custom ASICs that do the job with few additional components as opposed to using lots of standard components. The old UPSes do use mostly standard components, but they were expensive and had some issues determining the charge left as they had to rely on voltage alone. Nowadays, a UPS using an ASIC can determine charge level very accurately as the DSP can calculate charge left based on battery current and voltage as well as make up for voltage droop under load. Some can even report the condition of the battery (typically done by disabling the main rectifier and running the loads from the battery and calculating internal resistance of the battery) and inform the user how long the batteries will last before replacement. Some of the newest ones even have the DSP control battery charging, improving battery life. (Many UPSes charge the batteries by running the inverter DC/DC converter in reverse.)

And as for digital audio amps and ECM motor inverters, DSP use is even more common as they have to output a varying waveform as opposed to a steady voltage. In the case of the digital audio amp, it can take a stream of PCM data and convert it to analog at the very last step - in the low pass filters in the speakers. For an ECM motor, the inverter can output PWM-modulated sine waves instead of square waves to the motor coils, improving both noise and efficiency.

But for PC power supplies, I'm not sure how many use DSP. I have a 200w Mitac out of a small file server that uses an Infineon digital ASIC. It could take anywhere from 90v to 260v AC without a switch. Efficiency is high and it actually draws far less current than a normal 200w power supply powering the same load.

[Oleg Artamonov]

But for PC power supplies, I'm not sure how many use DSP

No one.

I have a 200w Mitac out of a small file server that uses an Infineon digital ASIC

There's no digital Infineon PWM control ICes. They all are analog.

[Oleg Artamonov]

But for PC power supplies, I'm not sure how many use DSP

No one.

I have a 200w Mitac out of a small file server that uses an Infineon digital ASIC

There's no digital Infineon PWM control ICes. They all are analog.

It could take anywhere from 90v to 260v AC without a switch

Which means that PSU has Active PFC circuit.