Date: Mon, 1 Dec 2014 11:46:39 -0500 From: Alejandro Imass <aimass@yabarana.com> To: Erich Dollansky <erichsfreebsdlist@alogt.com> Cc: kudzu@tenebras.com, Perry Hutchison <perryh@pluto.rain.com>, FreeBSD Questions <freebsd-questions@freebsd.org> Subject: Re: OT: UPS for FreeBSD Message-ID: <CAHieY7Rq6CBASdEkm_tNOuKF8kzK3_f6TCoQD=RzgHtbhCymHw@mail.gmail.com> In-Reply-To: <20141201140037.584faf7e@X220.alogt.com> References: <CAHieY7QGp2ELF-R91eu=vSrPsimVmVNJQ4kfucQ56PR7EEZmig@mail.gmail.com> <m57qdq$did$1@ger.gmane.org> <54777AB1.9010800@bluerosetech.com> <m581p1$65m$1@ger.gmane.org> <54779629.302@bluerosetech.com> <alpine.BSF.2.11.1411271433320.60866@wonkity.com> <5478BD4F.7020306@yahoo.com> <5478BEE6.30308@bluerosetech.com> <5478CC08.9090307@yahoo.com> <20141128204722.561f948e@archlinux> <5478F16A.80605@yahoo.com> <CABhTyc9m7fOoeV170dj=foAhmyYWphzc8KD8wBacu5gNRPhT%2BQ@mail.gmail.com> <54791d3a.w/pI0kak03d%2B3nKC%perryh@pluto.rain.com> <CAHu1Y71vVbdx6Yd1VbE7kb_8k9O5UG93RXEaORPU0tULCpMsCQ@mail.gmail.com> <20141129113405.3d1bd1d6@X220.alogt.com> <54798883.saa13h6lE6rPwZCf%perryh@pluto.rain.com> <20141130212746.6d5eaf1f@X220.alogt.com> <547bd5bd.dKE49fHIj28ERZyT%perryh@pluto.rain.com> <20141201140037.584faf7e@X220.alogt.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Dec 1, 2014 at 1:00 AM, Erich Dollansky <erichsfreebsdlist@alogt.com > wrote: > How do you do a DC-DC regulation without high losses? > > The best known way is going via AC to reduce the losses. The moment you > Switching / PWM regulation is technically not the same as AC. DC can be regulated with Switching/PWM more efficiently than AC with transformers; this is essentially what switching power supplies do. But is _a lot_ more complex so it's currently not practical for electrical grid applications. > are at AC, the rest does not matter anymore. You can now use standard > AC voltages with standard AC - DC converters to minimise cost. The cost > include here also the operating cost. > AC and DC each have their advantages and disadvantages and applications. AC is currently more versatile for power grids (generation, transmission, distribution) as it can be transformed much simpler, and it is definitively more reliable because an AC system has fewer parts than a DC one, at least with the technology we have today. AC has several drawbacks like the skin effect, capacitance problems, synchronicity (current and voltage becoming out of phase given different load conditions and types), losses to heat in transformers, etc. but these challenges are easily (and cheaply) mitigated by different techniques, and although it's not perfect it's still much cheaper, practical and much more reliable than DC grids. Nevertheless, DC is in fact used in _some_ transmission applications, particularly very long distance transmission applications and a few others. Would all-DC grids be more efficient? probably so, but are currently impractical with the technology we have today. Perhaps as new technologies become available (super conductors?) DC could become more prevalent. Back to the issue of using all-DC UPS _seems_ to make a lot of sense at first glance, I believe someone already mentioned that telephone system do exactly this, BUT.... The power consumption of a distributed telephone system is much less than modern computing applications so at high power you would still need to employ high voltage to decrease loss in the distribution, or you would need to employ ridiculously large conductor bars to be able to distribute low-voltage DC even in the most modest data center, so it's really not practical, cost-effective or reliable. So even though current UPS design seems stupid (transform AC to DC and then DC back to AC) is really much more practical in larger scales. Take for example a modern "efficient" bitcoin miner which consumes 1.4kWh @ 2THs. At 120V this is around 11A which requires at least 14 gauge wire. At 240V this is around 5-6A which can use regular 16 or even 18 gauge without even warming it. Suppose we would use a 5VDC distribution in the data center, that would mean 280A continuous __per miner__ and that is a pretty thick and expensive wire. So for practical reasons you need to step-up and then step-down the voltage in order to be economically viable (or even physically viable just in terms of conductor size). Stepping up and down is much more practical and cheaper (not necessarily more efficient) with AC transformers, so the DC conversion needs to be done close to the CPU. Another solution would be for CPUs to use high voltage to begin with, but this would make them inviable because of the heat density, that's why VCC in CPUs has been dropping over the years to about 3V and even 1.8V nowadays. BTW, both heat density and miniaturization are reaching their limits in silicon-based semiconductors. I believe the minimum is around 50 atoms for a transistor junction to work and I think we are pretty close to that limit, so Moore's law has an actual limit for transistor-based computers. Anyway, the point is that with the technology we have today, AC is still more practical for electrical power distribution. Maybe as new technologies come along DC or even wireless power grids will become more prevalent, but not for a long time. Best, Alejandro Imass
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAHieY7Rq6CBASdEkm_tNOuKF8kzK3_f6TCoQD=RzgHtbhCymHw>