My Dream System (I think)

[joeschmuck]

100kW is a bit much for my system.

 

[jgreco]

100kW is a bit much for my system.

Demonstrating a lack of imagination.

 

[rogerh]

When worrying about the safe operation of a line-interactive system it helps to remember that (in the USA) a power supply is receiving a rounded pulse of current every 8.3 milliseconds, falling to near zero for few milliseconds between pulses. So the power supply is used to not having an input for some milliseconds. This is an oversimplification but it is clear that it can be designed to be OK.

 

[joeschmuck]

This is an oversimplification but

Very oversimplified because in a transformer when the magnetic field collapses that in itself is energy. Also, if you placed the lower edge of the sinewave on a DC scale, then all of the voltage would be in the positive direction with a brief zero voltage level. It's all about perspective and how a generator/power supply was designed.

 

[rogerh]

Very oversimplified because in a transformer when the magnetic field collapses that in itself is energy. Also, if you placed the lower edge of the sinewave on a DC scale, then all of the voltage would be in the positive direction with a brief zero voltage level. It's all about perspective and how a generator/power supply was designed.

I agree with your second statement, but in an SMPS there is no 60Hz transformer, and the inductance is only storing energy for microsecond periods. At the line power frequency I guess it is capacitors storing the rectified power on the primary and the secondary sides.

 

[Black Ninja]

I found this beautiful explanation from APC White Paper about the pros and cons of line interactive and online double conversion for ups's from 750VA-5000VA range. After giving it more consideration I am giving up the idea of getting double conversion for myself.
Big thanks to jgreco for pointing the right direction, again


Q&A:

Q: Black Ninja , Are you saying that double conversion is not good and therefore it should not be chosen over line interactive ?

A: No ! In some cases double conversion is the only option to make it work, but for the rest of the cases where it is not, it will mostly hurt you than help you.


I am the Black Ninja and I approve this message.

 

[joeschmuck]

I agree with your second statement, but in an SMPS there is no 60Hz transformer, and the inductance is only storing energy for microsecond periods. At the line power frequency I guess it is capacitors storing the rectified power on the primary and the secondary sides.

Ah, just no transformer but still I'd think a full wave rectifier would be used and yes, capacitors would still be storing & filtering the power otherwise the power would look like crap on an O'scope, not something I'd want on a computer at all. On a 100kW output or larger, that some good circuitry and big capacitors. But you would need the capacitors either way. Maybe I missed this in the conversation but why does anyone need 100kw of DC power? I don't work in server rooms so I'm not sure why that would be needed.

 

[Bidule0hm]

Maybe I missed this in the conversation but why does anyone need 100kw of DC power? I don't work in server rooms so I'm not sure why that would be needed.

Because doing DC --> AC in the UPS to then do AC --> DC in the PSU is stupid (especially for the efficiency) :)

 

[jgreco]

Ah, just no transformer but still I'd think a full wave rectifier would be used and yes, capacitors would still be storing & filtering the power otherwise the power would look like crap on an O'scope, not something I'd want on a computer at all. On a 100kW output or larger, that some good circuitry and big capacitors. But you would need the capacitors either way. Maybe I missed this in the conversation but why does anyone need 100kw of DC power? I don't work in server rooms so I'm not sure why that would be needed.

Well, now we're drifting off into different territory. There's been DC power in server rooms and the data center basically forever, because telco gear typically uses -48V DC, and it was pretty easy to ramp up the size of that. Between the batteries and the rectifier, you get a nice stable source of clean power.

Lately some people have been doing that at a smaller scale: http://datacenterfrontier.com/google-unveils-48-volt-data-center-rack-joins-open-compute/

But some people are looking to other efficiencies, such as higher voltage for distribution, http://electronicdesign.com/power/400-v-dc-distribution-data-center-gets-real-0

It all comes down to the idea that at scale, if you can identify a way to reduce the amount of elctricity you're wasting through needless conversions and manipulations, the better it tends to be. Even at small scale,

A: No ! In some cases double conversion is the only option to make it work, but for the rest of the cases where it is not, it will mostly hurt you than help you.

this often turns out to be the case.

 

[joeschmuck]

Because doing DC --> AC in the UPS to then do AC --> DC in the PSU is stupid (especially for the efficiency) :)

I get that pint but I guess I don't understand the power distribution from a 100 kw source to what computer components. I would think that if you are sending out DC voltage over any distance that there would have to be a sense line to ensure the voltage at the end is still proper. So if you were sending out +12VDC and it traveled through 20 feet of wire, there would voltage be loss. This would be even more pronounced for lower voltages. Again, i don't work in a data center where I would suspect something like this would be used but I can't wrap my head around the distribution of critical voltage levels.

In my work we distribute DC voltages as well but our primary source voltage is 270VDC. That feeds power supplies which convert it into 5VDC, 12VDC, -12VDC, 28VDC, and even 115VAC @ 5khz. The 5VDC is distributed over very short distances though a bus bar in order to mitigate the voltage drop. The other voltages are not as critical and use normal wiring, but none of these power supply outputs run longer than 4 feet, by design of course, except for the 28VDC, this is non-critical and so we can run distances of several hundred feet just to report if a valve is open or closed.

 

[joeschmuck]

Lately some people have been doing that at a smaller scale: http://datacenterfrontier.com/google-unveils-48-volt-data-center-rack-joins-open-compute/

But some people are looking to other efficiencies, such as higher voltage for distribution, http://electronicdesign.com/power/400-v-dc-distribution-data-center-gets-real-0

Thanks for the links, I understand better now.

 

[jgreco]

I found this beautiful explanation from APC White Paper about the pros and cons of line interactive and online double conversion for ups's from 750VA-5000VA range. After giving it more consideration I am giving up the idea of getting double conversion for myself.
Big thanks to jgreco for pointing the right direction, again :)

Not a problem. It is probably worth noting that the APC paper is vaguely optimistic, and I appreciate your pessimism overall, because in the real world, I've found that the pessimist is more frequently correct for things like this.

A high quality switching power supply that is properly sized should have no problems with the cutover of an automatic transfer switch. Similarly, it should have no problems with cutting over to the inverter output of a properly functioning line interactive UPS.

Unfortunately, these things do not always work as they should. I've seen marginal switching power supplies. So for that, you might notice that I'm very conservative with power supply sizing, which leaves some extra safety margin in there. I've also seen UPS failures where the unit appears to work, but is unable to withstand the sudden slam on the inverter when power fails, and the output voltage goes all to hell.

Also for the gear that really requires uptime, redundant power supplies powered from independent power sources. This provides resiliency from a variety of failure types.

 

[jgreco]

I get that pint but I guess I don't understand the power distribution from a 100 kw source to what computer components. I would think that if you are sending out DC voltage over any distance that there would have to be a sense line to ensure the voltage at the end is still proper. So if you were sending out +12VDC and it traveled through 20 feet of wire, there would voltage be loss. This would be even more pronounced for lower voltages. Again, i don't work in a data center where I would suspect something like this would be used but I can't wrap my head around the distribution of critical voltage levels.

In my work we distribute DC voltages as well but our primary source voltage is 270VDC. That feeds power supplies which convert it into 5VDC, 12VDC, -12VDC, 28VDC, and even 115VAC @ 5khz. The 5VDC is distributed over very short distances though a bus bar in order to mitigate the voltage drop. The other voltages are not as critical and use normal wiring, but none of these power supply outputs run longer than 4 feet, by design of course, except for the 28VDC, this is non-critical and so we can run distances of several hundred feet just to report if a valve is open or closed.

Just be careful where you accidentally drop your wrenches.

 

[joeschmuck]

Hey, I've seen the results of a screwdriver hitting across the 270VDC (30 kw source), melted part of it down. It was a big screw driver too. We still have it and it works fine, but makes a point to everyone who uses it to be careful.

 

[Bidule0hm]

I get that pint but I guess I don't understand the power distribution from a 100 kw source to what computer components. I would think that if you are sending out DC voltage over any distance that there would have to be a sense line to ensure the voltage at the end is still proper. So if you were sending out +12VDC and it traveled through 20 feet of wire, there would voltage be loss. This would be even more pronounced for lower voltages. Again, i don't work in a data center where I would suspect something like this would be used but I can't wrap my head around the distribution of critical voltage levels.

In my work we distribute DC voltages as well but our primary source voltage is 270VDC. That feeds power supplies which convert it into 5VDC, 12VDC, -12VDC, 28VDC, and even 115VAC @ 5khz. The 5VDC is distributed over very short distances though a bus bar in order to mitigate the voltage drop. The other voltages are not as critical and use normal wiring, but none of these power supply outputs run longer than 4 feet, by design of course, except for the 28VDC, this is non-critical and so we can run distances of several hundred feet just to report if a valve is open or closed.

Of course you need to use high voltage like 200-300 V DC to not have huge currents and at the server end you just do a DC --> DC conversion (actually a standard PSU can already do it), still better than DC --> AC then AC --> DC ;)

 

[anodos]

Of course you need to use high voltage like 200-300 V DC to not have huge currents and at the server end you just do a DC --> DC conversion (actually a standard PSU can already do it), still better than DC --> AC then AC --> DC ;)

Hey, I like some AC/DC!

 

[jgreco]

Hey, I like some AC/DC!

 

[Ericloewe]

Hey, I like some AC/DC!

We're the guinea pigs who get the first look at the unholy mixture of what's left of AC/DC and Axl Rose.

Of course something had to go wrong, no band ever seems to do concerts on Saturdays here, much less start a tour here.

 

[joeschmuck]

I like AC/DC but Axl Rose is a different story. You can't put those together. I also like Heart, Quarterflash, Stevie Nicks, Lady Antebellum, Led Zeppelin, ZZ Top, Van Halen, Pink Floyd, Dire Straits, and a variety of others, all in moderation.

 

[BigDave]

I like AC/DC but Axl Rose is a different story. You can't put those together. I also like Heart, Quarterflash, Stevie Nicks, Lady Antebellum, Led Zeppelin, ZZ Top, Van Halen, Pink Floyd, Dire Straits, and a variety of others, all in moderation.

Used to crank up Quarterflash in the boat while bass fishing, there was something about the vibrations of the music that caused the fish
to get very aggressive on top water baits. If it had not happened to me personally, I would have never believed it :D