Can (3) 6-20r 20 amp 208v Duplex Receptacles be on a 60 amp circuit?

[Todd0x1]

Combined with fused PDU above, If NRTL labeling of 6-50 to L6-30 adapter is not counterfeit, NEC 240.21(B) may allow the tap to 6-50R on #6cu feeders per NEC Table 210(B)(3), 210.24, or perhaps >50A feeders per 210.18 Exception, 210.23(D).

Thats a welder adapter and its listing probably limits its use to that (similar to RV adapters)

 

[suemarkp]

It kind of does yet kind of doesn't. If what is plugged in isn't the utilization equipment being served it generally needs to be listed for whatever purpose it is serving.

Like I said before you can't put several receptacles next to your service panel and then run extension cords to all your utilization equipment not in that vicinity, temporary purposes would be an exception.

And that is the problem. We have labs with computer and audio/video type equipment with no NRTL listing (typically 20% of things are not NRTL listed). Many items have CE markings, but this self certification of conforming to EU standards is not an acceptable mark in the US. Many things also have the RU label (UL Recognized component). But that by itself is also no good. We also have a bunch of custom made things because you can't buy them (we do R&D work), plus items used in labs that when deployed in the field are exempt from the NEC (e.g. aircraft avionics that is typically approved under FAA test/experimental processes or military acceptance).

Many of the NRTL rules fall back to UL requirements, and the NEC basically regurgitates many of the UL standards. I find the NEC easier to look through than all the UL standards just because the scope of UL is so wide and I have much less experience reading those documents.

 

[Todd0x1]

800 Total on all shelves running the entire 30'. Google "miner container"

I would be curious how the numbers compare using 3x 84 space panels, 100x 6-50R and 100x sketchy adapter vs 10x 84 space panel & 400x 6-20R duplex.

 

[ramsy]

100x 6-50R and 100x sketchy adapter vs 10x 84 space panel & 400x 6-20R duplex.

Sketchy adapter missing fused load, or fused-plug strip, violates both NEC 240.21(B) feeder tap rule, and Equipment Listing requirements (NRTL).

Underfloor channel running length of container may be a wireway, see NEC 376. Up to 30 CCC's can fill 20% of wireway without conduit derating, and avoid tap rules if desired. All you need is wire, receptacles, and removable cover for wireway channel under floor.

 

[Todd0x1]

Sketchy adapter missing fused load, or fused-plug strip, violates both NEC 240.21(B) feeder tap rule, and Equipment Listing requirements (NRTL).

Underfloor channel running length of container may be a wireway, see NEC 376. Up to 30 CCC's can fill 20% of wireway without conduit derating, and avoid tap rules if desired. All you need is wire, receptacles, and removable cover for wireway channel under floor.

Even with 30 CCC's in wireway, youre talking about 800 CCCs so thats alot of wireway. Then I was thinking what if a prefab shop made up 4S boxes with 12/4 MC and two 6-20R duplex in each. That would be 200 runs of the 12/4 MC cable.

This power density here is insane (I have no idea how this would get cooled) something like 3kw per square foot. And what are the fault currents going to look like? Seems like the electrical gear will take up more space than the load.

I am very interested to see how this turns out.

 

[winnie]

Bitcoin mining is one of the reasons for silicon carbide microelectronics. The computers run at about 125C and are water cooled. The steam generated is used to supply district heating

 

[kwired]

Off topic to some extent, but read about people wiring these bitcoin mining operations from time to time and was wondering if those jobs pay in regular or cryptocurrency?

 

[cottora]

Frankly, I am finding out that most of these operations don't care about compliance. You cannot use standard Data Center equipment and be successful (that stuff is meant for 99.99 uptime and often has outlet level monitoring).

Google bitcoin container and you can find multiple outfits. My first concern is the safety of people. I can live with burning up miners.

That said, there are several manufacturers going through UL approvals for miner specific PDUs. These things run full blast, 24x7 (and scream).

kwired - crypto of course...

 

[ramsy]

Even with 30 CCC's in wireway, youre talking about 800 CCCs so thats alot of wireway. Then I was thinking what if a prefab shop made up 4S boxes with 12/4 MC and two 6-20R duplex in each. That would be 200 runs of the 12/4 MC cable.

This power density here is insane (I have no idea how this would get cooled) something like 3kw per square foot. And what are the fault currents going to look like? Seems like the electrical gear will take up more space than the load.

I am very interested to see how this turns out.

Design efficiency with off the shelf equipment was a noble goal.

Unfortunately, for Wireway or Art. 390 Underfloor Raceways within steal-frame channels under typical shipping container wood floors are going the wrong way, side to side, rather than end to end. https://www.supercubes.com/blog/2013...close-up-look/

Perhaps renting temporary feeder cables with Scatter Box PDU's described in NEC Article 530 Motion Picture productions, or 525 Carnivals.

Whatever wiring method is used, cost efficiency for the POCO's requirements may also include temporary cables, rather than permanent raceways.

 

[winnie]

Interesting point: if as a _business decision_ you don't care about burning up machines but instead only care about protecting lives, then the NEC really is not the tool you want to use.

While bitcoin is computer based, you can stand downtime in ways that most data centers cannot.

Instead you want to engineer this such that you build to much lower cost (and lower reliability _and_ 'safety', where safety is read as 'probability of burning up'), and then set things up so that humans are no exposed to the hazard. Then get an NRTL to bless the entire system.

We design electrical systems so that humans can safely sleep next to the wiring. But this is a _steel container_ filled with computers. Design the system like a control panel where the power must be _off_ to enter. Humans are _not_ permitted to 'enter the furnace' (and furnace it is; you have a megawatt of power going into a shipping container) when it is operating.

Design the system so that if there is a significant indication of fire the whole thing shuts down.

Design the system with residual current detection which will shut down large swaths in the event of a significant ground fault (say a circuit overheats and burns through the insulation, you don't wait for a 20A breaker to trip on overcurrent; instead you shut down the entire 200A panel).

Have your breaker panels on the _outside_ of the container so that you can easily turn off problem circuits.

Schedule downtime where the power is off for maintenance.

I bet with considered design you can come up with something that is much cheaper to build, protects human life as well as 'standard' practice, and better matches the system reliability and probability of unplanned self damage to the actual process.

-Jon

 

[cottora]

Interesting point: if as a _business decision_ you don't care about burning up machines but instead only care about protecting lives, then the NEC really is not the tool you want to use.

While bitcoin is computer based, you can stand downtime in ways that most data centers cannot.

Instead you want to engineer this such that you build to much lower cost (and lower reliability _and_ 'safety', where safety is read as 'probability of burning up'), and then set things up so that humans are no exposed to the hazard. Then get an NRTL to bless the entire system.

We design electrical systems so that humans can safely sleep next to the wiring. But this is a _steel container_ filled with computers. Design the system like a control panel where the power must be _off_ to enter. Humans are _not_ permitted to 'enter the furnace' (and furnace it is; you have a megawatt of power going into a shipping container) when it is operating.

Design the system so that if there is a significant indication of fire the whole thing shuts down.

Design the system with residual current detection which will shut down large swaths in the event of a significant ground fault (say a circuit overheats and burns through the insulation, you don't wait for a 20A breaker to trip on overcurrent; instead you shut down the entire 200A panel).

Have your breaker panels on the _outside_ of the container so that you can easily turn off problem circuits.

Schedule downtime where the power is off for maintenance.

I bet with considered design you can come up with something that is much cheaper to build, protects human life as well as 'standard' practice, and better matches the system reliability and probability of unplanned self damage to the actual process.

-Jon

This is exactly right! The thing to remember is that these miners are built to be used in a residential setting (I have one in my basement that provides heat and keeps the inlaws away). They have sophisticated protection and will shut down at the first sneeze of an issue.

My BIGGEST fear is the metal. What if a rat chews threw a wire and someone touches a hot container, shelf, etc...

 

[winnie]

This is exactly right! The thing to remember is that these miners are built to be used in a residential setting (I have one in my basement that provides heat and keeps the inlaws away). They have sophisticated protection and will shut down at the first sneeze of an issue.

My BIGGEST fear is the metal. What if a rat chews threw a wire and someone touches a hot container, shelf, etc...

My biggest concern would be fire.

Everything is metal, bolted or welded together. The container will be bonded to the supply neutral and grounded. A low current fault won't cause much change in the voltage of exposed metal. A high current fault will be transient. If you use residual current detection (same sort of tech as GFCI, but with a much higher current threshold, doesn't protect people but does protect equipment) then faults could be limited to 10's of amps prior to automatic shut down. So IMHO with a little care as to proper bonding of everything together the shock hazard is nil. Especially if you run the system that no human enters when the computers are powered. (Remember they are all networked, you know which machines have died _from the outside_ and can pull them every maintenance cycle.)

But you have _lots_ of power in a very small space. If the normal output power of one of your circuits goes to acting like a concentrated resistance heater rather than powering a set of miners (which will generate the same amount of total heat, simply in a more diffuse fashion with fans running) then you have a rather effective ignition source. Hell, the normal heat output of the miners, if it isn't dissipated by some sort of forced cooling, is enough to heat the entire _outside_ of the shipping container up to a couple hundred C.

Now I am positing that humans are excluded from this environment when it is running, and that you are designing the system with the intent that it is at high risk of catching fire but that the fire would be safely contained and there would be no more than _expected_ economic harm. Then you calculate the risk of failure/fire versus the costs of preventing that fire, and minimize your total costs. If you build a system that is half the price but has a 10% annual chance of being a total write-off, that might be a win.

 

[cottora]

My biggest concern would be fire.

Everything is metal, bolted or welded together. The container will be bonded to the supply neutral and grounded. A low current fault won't cause much change in the voltage of exposed metal. A high current fault will be transient. If you use residual current detection (same sort of tech as GFCI, but with a much higher current threshold, doesn't protect people but does protect equipment) then faults could be limited to 10's of amps prior to automatic shut down. So IMHO with a little care as to proper bonding of everything together the shock hazard is nil. Especially if you run the system that no human enters when the computers are powered. (Remember they are all networked, you know which machines have died _from the outside_ and can pull them every maintenance cycle.)

But you have _lots_ of power in a very small space. If the normal output power of one of your circuits goes to acting like a concentrated resistance heater rather than powering a set of miners (which will generate the same amount of total heat, simply in a more diffuse fashion with fans running) then you have a rather effective ignition source. Hell, the normal heat output of the miners, if it isn't dissipated by some sort of forced cooling, is enough to heat the entire _outside_ of the shipping container up to a couple hundred C.

Now I am positing that humans are excluded from this environment when it is running, and that you are designing the system with the intent that it is at high risk of catching fire but that the fire would be safely contained and there would be no more than _expected_ economic harm. Then you calculate the risk of failure/fire versus the costs of preventing that fire, and minimize your total costs. If you build a system that is half the price but has a 10% annual chance of being a total write-off, that might be a win.

We will work in the container. The miners will auto shutdown if they get to how (and we have 20 exhaust fans moving 160,000CFM). Wouldn't our panel notice any fault and trip the main breaker (it is a standard 400 amp Square D)?

 

[winnie]

We will work in the container. The miners will auto shutdown if they get to how (and we have 20 exhaust fans moving 160,000CFM). Wouldn't our panel notice any fault and trip the main breaker (it is a standard 400 amp Square D)?

Circuit breakers trip only on conditions that they are designed to detect. For example a common thermal/magnetic circuit breaker will trip on overload if extended for sufficient time, and will trip rapidly on a short circuit of sufficient magnitude. But it will not trip on a fault that doesn't exceed its trip characteristics. Your 400A breaker will not trip on a 20A current limited fault; such a fault just looks like an acceptable load.

So say you have a slightly scraped wire that is making poor contact to the metal of the container, such that a current of 20A flows. This will generate quite a lot of heat, but won't trip a breaker until the heat melts more insulation and causes a larger fault. It is even possible that the fault would 'burn clear' and never trip the breaker.

This is the reason for using ground fault detection. You can have a system that will supply 100A to a load, but will trip if say 5A flows through a fault to ground. GFCI devices take this even further and shut down with 5 or 6 mA of 'residual current'. However if your application is protecting equipment rather than people, you use a much higher trip threshold to avoid nuisance trips.

If people will be working in the container when the computers are operating, then I believe you are obliged to wire the container to standards such as the NEC in order to protect these people. I was only suggesting that looser standards would be appropriate and acceptable if you ensure that people are not inside the container when operating.

-Jon

 

[cottora]

Circuit breakers trip only on conditions that they are designed to detect. For example a common thermal/magnetic circuit breaker will trip on overload if extended for sufficient time, and will trip rapidly on a short circuit of sufficient magnitude. But it will not trip on a fault that doesn't exceed its trip characteristics. Your 400A breaker will not trip on a 20A current limited fault; such a fault just looks like an acceptable load.

So say you have a slightly scraped wire that is making poor contact to the metal of the container, such that a current of 20A flows. This will generate quite a lot of heat, but won't trip a breaker until the heat melts more insulation and causes a larger fault. It is even possible that the fault would 'burn clear' and never trip the breaker.

This is the reason for using ground fault detection. You can have a system that will supply 100A to a load, but will trip if say 5A flows through a fault to ground. GFCI devices take this even further and shut down with 5 or 6 mA of 'residual current'. However if your application is protecting equipment rather than people, you use a much higher trip threshold to avoid nuisance trips.

If people will be working in the container when the computers are operating, then I believe you are obliged to wire the container to standards such as the NEC in order to protect these people. I was only suggesting that looser standards would be appropriate and acceptable if you ensure that people are not inside the container when operating.

-Jon

Thanks! We are going to use PDUs and limit the load banks to 30 amps which will be feed with a 30 amp panel breaker into a 30 amp PDU breaker. The price savings isn't worth the risk.

 

[kwired]

So say you have a slightly scraped wire that is making poor contact to the metal of the container, such that a current of 20A flows. This will generate quite a lot of heat, but won't trip a breaker until the heat melts more insulation and causes a larger fault. It is even possible that the fault would 'burn clear' and never trip the breaker.

Thanks! We are going to use PDUs and limit the load banks to 30 amps which will be feed with a 30 amp panel breaker into a 30 amp PDU breaker. The price savings isn't worth the risk.

I have many times seen a lot of destruction on 20-30 amp 480 volt circuits yet the overcurrent device never opened. There is high enough voltage on this system to sustain arcing, but if current never goes over the overcurrent device trip curve it won't open.