It happens to the best of us. We power up our project and immediately run into issues. Be it spotty communication or microcontroller reset or any number of bugs that have us mystified and picking though our code… only to find that it’s a power supply issue. Anyone who has tried doing Raspberry Pi stuff and depended on the USB power from their PC has certainly been bit by this.
It’s the same with larger, more power hungry projects as well. [Nerd Ralph] has been running a mining rig for a few years now, and has learned just how important proper power supply management can be. His strategy involves using interlocks to ensure everything powers up at the same time to avoid feedback problems, running a separate ground wire between all GPU cards and the PSU and running the supplies at 220 for the NA folks.
Be sure to check out [Nerd Ralph’s] blog for more details and tips to power your own mining rig.
Interesting. That interlock is simple and elegant, very nice.
For the voltage setting, I thought that wattage equaled volts times amps? Shouldn’t the waste heat be the same regardless of whether it’s drawing ten amps at 120 or five amps at 240? Or is it more complex and nonlinear than that? I’m curious.
Five vs Ten amps is a big difference in heating in the power factor correction circuit of his power supply. If you find a data sheet for similar power supplies, they are always 1-3% more efficient on 220Vac and run cooler because of it.
Gotcha. Thanks for responding, good to have a ballpark figure on this effect.
Heat is a result of the friction of electron flow. Half the current at double the voltage should result in roughly half the heat if using the same size wire. It’s why our grid transmits at 100k volt over long distances.
Norm, remember your high school electrical classes, and don’t confuse the young ‘uns here. I’m sure you’ll recall that halving the current reduces losses (heating) in a conductor to one quarter: losses are proportional to the square of the current.
I don’t know where your grid is, but in North America, high voltage lines are not 100 kV. 19 kV and 69 kV are typical for local distribution. 115 or 138 kV for intra-city, and 230 or 500 kV for most long-haul lines, not counting the whacky HVDC or EHV stuff.
You also have cables running from the power distribution system and over to the servers, if we halve the current in these lines, then we quarter the power dissipation in these cables. This is both true for the power cords, and for the internals in the power supply as well. At least for the “AC” side of it.
On top of this, if we halve the current, then we can also use thinner and therefor cheaper wires in our power distribution system.
And less current in the AC distribution is less electromagnetic interference.
And because what you need is power and not current, and because of voltage drop in your wiring, the current is slightly more than twice.
If you have a load of 1200 Watts, so 10 A in 120VAC and 5 A in 240 VAC. With wire around 8 ohms/km (roughly wire of 2.5mm² section or 14 AWG), you get a voltage drop of 8 Volts in one case, 4 Volts in the other. So 6.67% of your main voltage vs. 1.67% of your main voltage. Already 5% of efficiency loss in your system…
But it’s even worst because in this case you don’t get 1200 Watts at your load, only 1120 Watts ((120-8) * 10) and 1180 Watts ((240-4) * 5)
To get your 1200 Watts at your charge, you need 10.77 Amps and 5.08 Amps. So you get a overall efficiency of 92.8 % at 120 VAC (1200 / (10.77*120)) and 98.3 % at 240 VAC. (5.5% of efficiency difference).
(So your distribution wire can be thinner, and so cheaper, mainly because you have less current, but also because you can have a slightly higher voltage drop with less consequence !)
Technically it depends on the amount of resistance, effectively proportional to the amount of the conductor the electricity is able to pass through. With AC that means the skin effect comes into play but that’s a whole other bundle of fun math. With DC it’s essentially amps times volts.
And this AC vs DC case when it comes to skin effect, is a reason for why some Data centers use DC distribution, as most server products uses switch mode power supplies and therefor has an internal bridge rectifier.
And another actually bigger reason for using DC instead of AC is when one wants x amount of mega watts of battery backed power, and converting that from DC to AC to DC again is rather inefficient, compared to just sending out the DC directly.
I forgot to say that 220 volt AC is effectively 310 volts peak, so therefor one could instead use 300 volts DC instead, giving roughly 25% less current to carry in the wires to get the same power in the end.
Most PSUs are not designed to receive DC even if you calculate it correctly.
The 4 rectifier diodes on the primary side are evenly used with AC but with DC only two get hot, the other two do nothing.
It can and will most likely work but you may decrease the lifespan of the PSU.
Yes, there is a difference in running AC and DC through a rectifier. As as Limroh states, only two of our rectifying diods will be used, and therefor be under higher stress.
But at the very same time, if the rectifier dissipates enough energy for this to be a concern, then it is probably a poorly designed power supply to start with. As most power supplies will be designed for the voltage range of 110 to 250 volts AC. And as it will draw the most current down at 110 volts, then this is where the manufacturer has their specifications chiseled in stone, as if we want 1100 watts, then we need at least 10 amps when we have 110 volts AC on the input.
This means that we “effectively” have 5 amps to dissipate per diode.
While if we take the same supply and drive it with 300 volts DC, then it will only need a bit less then 4 amps. And this is in fact going to put our rectifier under less load then when we run it with 110 volts AC.
But if the power supply isn’t made for the 110-120 volt market, then yes, this can be a legit concern.
But at the same time, if one wants a high efficiency power supply, then the manufacturer typically have placed in a rectifier that is technically over specified for the job, as these devices are generally not that expensive, and are an easy way to cut down a few watts overall without adding any substantial extra cost.
Skin effect does not come into play at 60Hz!
50Hz, not 60…
US are 60 Hz, Europe and Africa 50Hz
I2R losses means that higher voltage can mean less wate heat. I*R=V.
I’m confused… “Wire up the outlets for your mining rigs for 240 instead of 120”, shouldn’t you rig it up to use whatever enters your home. Apparently there is a choice, how is that possible, do they run their own generators over there?
In some places, you can get 120 and 240, or you could wire up a large transformer at your mains distribution board and get what ever voltage you wish. Might need to call the local electrician to not loose any legal battles if the house burns down.
I don’t know about anywhere else, but in North America, power enters the home via two 120V hot and one neutral. The two 120V hot legs are out of phase with each other, and when combined, you get 240V. So, from the breaker panel, you can pull 120V or 240V, depending on the type of breaker you insert at a given position. Some large appliances (water heaters, central AC units, electric stoves, large power tools) are 240V in North America, but most everything else is designed for a single leg 120V feed. Then just to make things even more complicated, large industrial businesses often have three phase 480V for the really big equipment.
Yup. That’s how you install an outlet for a big dryer or a welder or something than needs the extra power. Never occurred to me to utilize it for a computer PSU, but I suppose it’s pretty nice when you need to squeeze out every little bit of efficiency.
The two 120V legs are *in phase* with each other. Residential power in the US is single phase, sometimes called split phase because they’re both 120V relative to the center tapped neutral.
VAC says: “The two 120V legs are *in phase* with each other.”
Of course they are not. *by definition* if they were in phase there would be no voltage difference between them.
The two 120V legs are 180 degrees different from each other, so the difference between them is 240V. Which is (of course) the supply voltage.
Virtually every home in North America is wired with 240V (NOT 220V!) split phase, providing 120V to ordinary receptacles, and 240V (NOT 220V) to larger appliances. Most kitchens are wired with 240V split phase to the (split) receptacles at the countertop, providing a very convenient 240V source if you just wire up the appropriate adapter to take the hots from each phase.
3-phase installations like industrial facilities and some apartments are fed with 208V 3-phase, which provide three phases of 120V to neutral, or 208V phase-to-phase.
Actual 220V is present elsewhere, and is a convenient global average of 240V (UK,) 240/208V (NA), 230V (EU), 200V (Japan, though like NA it’s spit to 100V at normal receptacles). So if a manufacturer designs for 220V +/- 10%, it’s probably usable worldwide.
220 volt +/- 10% isn’t really a good design standard. As it doesn’t live up to the requirements within the EU to start with.
As 220 + 10% is just 242 volts, and the EU standard is 230 +/- 20 volts. So 250 volts is within the standard, and should be expected. And then we have Australia that has 240 to 250 volts.
And in my own experience, I have seen switch mode power supplies designed for 400 volts DC (280 volt AC), taking clearance for such into consideration throughout the whole design.
Yes, agreed, actual power supply design must take into account the allowable possible range of input.
My house nominal “240V” supply routinely hits 252V, for example, and is allowed to go as high as +6% (254V).
“International” power supplies usually are specified for 100V (Japan) to 240V +/- 10%, i.e., 90-264V.
In Finland pretty much every household is wired with 3x 230V phase, with main fuses between 25-63A. Nominal voltage from phase to ground is 230V and phase to phase 400V. Load is split between phases, so that phase voltages would stay same, usually this is done by grouping different rooms to different phases, commonly 10A 230V per room sockets and lights.
Washing machines etc. have their own 10-16A fuses. 3 phase 16A power is used for larger equipment, most commonly for semi-permanent installation like oven&stove combo and electric sauna stove. Those are usually connected to star, so single element in equipment will still see 230V.
Larger motors used in rural settings and light industry may aswell use 400 voltage between phases by triangle connection. 3 phase 16A sockets are almost never seen in apartments, but are usually found from houses that have garage or other workspace with larger machinery. Larger 3 phase sockets are very common in farm setting and light industrial buildings.
E.g. our hackerspace has air compressor that is fed by 3x 400V 25A line. It doesn’t even have neutral line, motor is connected to triangle and 24V for control logic is stepped down from 400V with transformer.
I have never understood why North America stayed in their stupid 120V system, that uses very high currents over inferior aluminium wires, connected with lose screwcaps and fuses that may explode out of the wall. And it would be probably federal crime and against the code to use proper copper wires connected with good push lock blocks and DIN rail mounted fuses and contactors…
Australia is similar, except 3ph is uncommon in a house. There’s 3 phase at the kerb and each house is connected to one phase.
Everything industrial is 3 phase though; it’s 240V phase/neutral or 415V phase/phase here though. You can certainly get a 3 phase connection to your house and/or shed though – it costs about $1000 to replace the meter plus a bit more for wiring – if you have a problematic power-tool habit like me.
We, uh, we don’t have sauna heaters down here. We have air conditioners :) In semi-related news, it was forecast to be 40C today…
There are of course odd exceptions: my grandmother’s old house from the 1920’s had an instant water heater. It was a single 415V element, connected between two phases. Highly unusual, and nearly killed an electrician who didn’t realise that one side of it wasn’t neutral.
120V comes with a nice plug (NEMA 5-15P) that is smaller than EU’s and you can fit more of them onto a power bar. There are always two outlets in a wall receptacle. Unless you have wet fingers, 120V won’t kill you (usually). Ask most kids that stuck things in the wall outlet and have lived to tell their tale… it is very common. Aluminum wire for branch circuits in homes was all but banned three decades ago or more. I think you have embellished things a bit. Look up the history why Europe went 50 Hz, instead of a higher frequency like 60 Hz… your lower 50 Hz results in heavier transformer cores. Find out why Japan is 100V, and the country is split 50/60 Hz! There is always politics, and not pure technical reasons.
While we are on this topic, in Canada a common building distribution voltage is 347/600V. Typically supplied to any medium sized business that requires more than 100kW and less than what would justify their own high voltage vault. For you electricians out there, a neat trick with 347/600V is that the current on a balanced 3 phase line nicely multiplies out to the power. So at 3 lines x 347V at 100A is 104kVA.
Simple, do you know how expensive it would be to change from 120v to 240v? It would be an insane amount and there would be virtually no benefit.
Aluminum is only used for specific applications like very heavy feeds and they have special connections for it. Wire nuts when properly used are very reliable. Push locks are not all they are cracked to be sometimes too. We do have them, I have installed literally tens of thousands of WAGO connectors. Generally we do not use fuses here, almost all circuit breakers and most new houses now have arc detect breakers too. They are used in industrial situations where you have 100A+ service to a single machine.
The standard in UK is 230 V not 240 V, like the the rest of the EU.
The continent was mainly using 220 V until 1986 and UK 240 V until 1986. The 230 VAC was choosen by the IEC as a compromise between the 220, 230 and 240 system. So actually main voltage in EU should be 230V +/-6% at distribution panel, and equipement should support 230V +/-10%. (voltage drop in the internal house distribution should be less than 4%)
But a transition period of 20~25 years was used, during the transition period main voltage at distribution panel in UK should be in the 230 VAC +10%/-6% range, and in the rest of EU it should be in the 230 VAC +6%/-10% range.
(Other 240 VAC country like Australia and NZ switch to 230 VAC at least officialy)
Ah, right you are. I’m out of date — I last dealt with UK power in 1981, and have not been back since. Thanks for the update.
> and some apartments are fed with 208V 3-phase, which provide three phases of 120V to neutral, or 208V phase-to-phase.
This causes some headaches for a friend who has to deal with unaware contractors installing 240VAC or 220 VAC devices in condo kitchens what have 208VAC, not 240VAC. Some devices don’t work, some don’t work well.
Well, once I replaced almost my entire PC ’cause I figured the motherboard was toast only to find out that my PSU died and would not power up my setup.
It needed an upgrade though so all ended up well!
I recently spent 2 days troubleshooting my main PC because it was shutting down under even normal load, let alone trying to mine cryptocurrency. (I mine on my daily-driver hardware, not investing in a dedicated rig)
Turns out my old GTX670 draws a lot more power than my new GTX1080 and, based on napkin calculations, brought my total power draw a bit too close to my 750W PSU’s limit. (200W CPU doing me no favors…)
Of course, after buying a 1000W PSU, I kept getting intermittent shutdowns and freezes after 30-60 minutes of gameplay or CAD work, despite remaining perfectly stable while mining for days at a time. The GTX670 has no backplate and last year a fan screw vibrated loose and fell on top of it while everything was running. The motherboard was toast so I ungraded it and the CPU (AMD FX9590), but the GTX670 itself seemed okay. Apparently not.