OSHA didn't hint at what to do. There is not a lot of empirical data nor a lot of details yet. So we end up with problems like this:

http://www.efcog.org/wg/esh_es/events/E ... ks_July%20[Compatibility%20Mode].pdf

You can read up on the proposed formulas and how it is addressed with DC batteries here:

http://www.battcon.com/PapersFinal2012/ ... 0Flash.pdfState of the art right now is as follows:

1. IEEE 1584 has an "exemption" for <125 kVA, fed by one transfomer, <=208 V. This is meant as an AC rule though and it's not even really a rule, more of a passing comment. So one can make the mistake of applying this assumption.

2. The Doan "max power transfer" formula that shows up in 70E works like this. Assume that the "system" is purely inductive and that the arc is purely resistive. Maximum power transfer to the arc occurs when the arc resistance equals the system impedance. Thus arc power equals 1/2 of the bolted fault current multiplied by the system voltage. Then multiply by time to get energy and estimate flux at a given distance assuming a sphere (divide by 4*pi*distance*distance). This is a direct adaptation of Ralph Lee's formula. It is similarly vastly conservative.

3. The Ammerman model looked at over a hundred papers studying DC arcs to derive a much more realistic current/voltage relationship to determine arc power. This results in a smaller number than the Doan/Neal formula. This is the state of the art theoretical model. Still, concentration of the thermal energy (exponent in the AC models) and an estimate of the percentage of energy that actually makes it to the victim is not taken into account so the number is several times higher than reality.

4. Some studies have been conducted under 300 V. At these voltages, arc stability is challenging but there is not yet a predictive formula for this. Kinetrics has done some test work (commissioned by Duke as I understand it).

5. However the results that are public (as quoted above) don't tell the whole story. For instance, the most common substation battery voltage that I have seen is 125 V. Data is given for a test voltage of 130 V and 20 kA normalized to 2 seconds and 12" exposure distance. The Doan formula is 28 cal. The Ammerman formula is 25 cla. The normalized test data is 7.5 cal. But since the arc only sustained for 0.8 seconds, and even taking into account anthropometrics we are looking at an 18" working distance for a battery string. These two factors combine to reduce the "normalized" value down to 7.5 * (0.8/2) * (12*12) / (18*18) = 1.3 cal/cm^2. This is barely above 1.2 cal and below the 2 cal/cm^2 threshold recommended in 1910.269 for requiring PPE. Further, measured arc current is 6200 A vs. a theoretical predicted 7900 A (Ammerman). Theoretical values tend to settle out at around 200-300% of tested values at higher voltages and currents but at low incident energies (under 10 cal/cm^2), there is a huge amount of divergence even when the system voltage increases to 260 V. My personal conclusion from this is that the theoretical equations are always larger than measured data, and that so far I see no reason to panic and worry about safety when it comes to DC power supplies so far.

6. What's missing above, and Ammerman comes close to it, is that there is a threshold value above which arc voltages are almost constant, forcing arc currents to climb rapidly. Below that threshold, arcs are very unstable and currents are very low. This threshold forms a basis for a lower voltage "cutoff" where arc flash is simply not possible at any current up to 28 V, and takes an enormous and unrealistic amount of current up to around 70 V. These conditions are not even worth considering below several kA even when entering the "constant voltage" range. This means that most industrial DC power supplies, golf cart batteries, and similar low voltage systems need not be considered for arc potential at all.

7. I can't find a single case in the OSHA investigation logs of an arc flash incident with a major injury requiring hospitalization involving a battery or UPS system.

OK, stepping away from the public information out there this is my personal analysis for a site that has lots of DC systems that span the gamut from 12 MW DC electric excavators to electrostatic precipators to substation battery arrays:

1. Using even the theoretical calculations and the largest DC generators and motors (such as a GE MD 824) that I've found and plugging in the stall currents as well as peak power currents, I can't find a "high voltage" DC system that reaches above the 1.2 cal/cm^2 threshold either.

2. I've also tried the same effort with the existing NWL-constructed electrostatic preciipitators at my site with the theoretical calculations and a 50 kV working voltage. Again, the result is the same...can't get above 1.2 cal/cm^2.

3. Plugging in realistic values for batteries (my batteries), the theoretical equations give low cal/cm^2 values, much lower than the data from the Kinetrics (Duke) study published in Battcon referenced above.

My overall conclusion from what I've studied on DC arc flash is that 70E in particular is all wet. The tables are for voltages and currents that are not in common use anywhere and the theoretical predictions are off by a large multiple compared to the test data that's out there. 70E is trying to be "leading edge" here but the end result was that a set of recommendations was published that in no way reflects real world applications.