Forgot your password?

Comment: *something* in, rubbish out... (Score 1) 55

by serbanp (#48412479) Attached to: Machine-Learning Algorithm Ranks the World's Most Notable Authors

Bram Stoker being #1 in the 1910 decade, way ahead of someone like Mark Twain? In what universe?

The list is full of mediocrity floating at the top, while profound authors being ranked way lower (Calamity Jane > Chekhov for instance).

The complete failure of this ranking experiment just shows how true AI is still 20 years in the future (as it has been for the past 50 years)...

Comment: Re:Overstated or misrepresented? (Score 1) 403

by serbanp (#48094723) Attached to: Fuel Efficiency Numbers Overstate MPG More For Cars With Small Engines

My 17yo Civic has a complete mileage log for all of the 205k miles it traveled, i.e. how many gallons to full (I never do partial refueling) and how many miles traveled. If anything, the mileage is consistently better than what was on the windshield when I bought it new.

In fact, looking at the numbers I can see the seasonal patterns (2MPG lower in winter), probably caused by my state switching to the winter blend (more Ethanol).

Comment: Re: I dunno about LEDs, but CFLs don't last (Score 3, Interesting) 602

by serbanp (#48004927) Attached to: The Great Lightbulb Conspiracy

No LED-based lighting fixture uses a voltage regulator and a series resistor to limit the current through the LED.

The LED is driven by a current regulator; yes, it usually has a small sense resistor to measure and adjust the driving current, but that's something totally different from a resistor used to limit the current.

Remember, the 'D' in LED stands for (semiconductor) "Diode", which means that in forward conduction it has an exponential relationship between current and voltage. The only way to control the brightness is to control the current, you can't rely on the equivalent "resistance" (i.e. Vd/Id).

The premature failure of the LED bulb is caused exclusively by the embedded electronics degrading at high temperature. Since the LED is still dumping a lot of heat when working, the heat must be dissipated (hence puny LEDs and bulky, machined Al heatsinks) or the electronics get fried.

Guess what the drive to reduce the manufacturing cost will do to the quality of the LED bulbs?

Comment: Re:so... (Score 1) 260

by serbanp (#47534195) Attached to: Google Offers a Million Bucks For a Better Inverter

The input is DC so there is no need for low frequency input decoupling

Unfortunately, that's not true. The output is AC and the input-referred current will reflect it, i.e. will be quite high when the output reaches its highest (340V). Since the input source has an equivalent internal resistance of 10Ohm and would droop too much, some bulk capacitance is needed to provide these 120Hz current surges.

I agree with you that the design revolves around the placement of power components (capacitors, inductor, power switches and some sort of heatsinking), then the control circuitry be crammed in the spaces between them.

Comment: Re:so... (Score 1) 260

by serbanp (#47524895) Attached to: Google Offers a Million Bucks For a Better Inverter

Having read the actual spec, I realized that my question was stupid. Input voltage is 450V, therefore the inverter is just a full-bridge DC/DC converter

The challenge is actually perfectly doable, with the only difficult points being the size of the switching inductor and the input bulk (most likely tantalum) and output filtering (must be ceramic) capacitors. IOW, the winner will be the one with the best connections to component manufacturers.

A LISP programmer knows the value of everything, but the cost of nothing. -- Alan Perlis