Effect of a short circuit

[gar]

161102-1123 EDT

The purpose of this comment is to show why electrical isolation should be used in an RS232, 422, or 485 communication path. If not isolation, then a sacrificial circuit is needed.

Consider a typical installation.
1. All equipment is fed from the same main panel.
2. All equipment has an EGC connected to its chassis.
3. All EGCs originate from the main panel.
4. All RS communication components reference to their equipment chassis.

5. Computer is in an office.
6. A CNC machine is on the shop floor.
7. For simplicity assume both are powered from 120 V. A higher voltage at the CNC just makes the problem worse.
8. Computer to CNC RSxxx communication cable is Belden 8723, but doesn't really matter.

9. At the CNC the hot 120 V line is shorted to the EGC. Note the EGC is the same size and kind of wire as the AC hot. In past times the EGC was smaller, but that even makes the problem worse.

10. Upon occurance of the short from hot to EGC the difference in voltage between the two pieces of equipment is about 1/2 the supply voltage at the shorting end.

11. For a 120 V or higher system this voltage will destroy RSxxx components, and possibly further into the equipment.

You can draw the circuit and study why. Below are scope waveforms that display a real world experiment.

Test setup:
1. Residential 50 kVA pole transformer.
2. 200 A Sq-D main panel.
3. 15 A QO breaker.
4. 50 ft roll of #14 Romex.
5. 25 A 50 mV Weston shunt, 50 A 100 mV. Connected in neutral path.
6. Rigol scope isolated from AC and EGC via Sola constant voltage transformer.

White and black are used. In other words white is being used as the EGC. White and black shorted at the far end. The actual EGC is floated. When the Romex was uncoiled an extension cord was used for extension of the voltage measuring test lead. Uncoiing the Romex produced no great effect.

Blue is voltage and red is current.

Turn on of the short is a random function.


*****
Romex in its original rolled state. Peak voltage slightly over 60 V, and peak current is slightly less than 500 A. Calculated peak current is slightly higher based on 2.5 ohms per 1000 ft. In the third plot you can see the voltage across the white and black wires in series, about 130 V.
.

.

*****
The Romex rolled out moderately straight. Extension cord used to extend voltage probe. Have not incestigated why peak voltage appears lower. The current looks like the same peak. Thus, uncoiling made no difference in current.


.
*****
This plot is the same as the 2nd plot except that the voltage is the source to the breaker, and the voltage scale has been changed. During the short the voltage measured includes voltage drop across the breaker. Looks like the main panel sees about a 50 V drop in the peak at 500 A peak load.


.

.

 

[gar]

161102-2050 EDT

The peak far point voltage on my second plot seemed incorrect at about 40 V.

Reran some experiments tonight. But I am having problems with some very much higher frequency noise than line frequency that is time synchronized with the AC line showing up in the plot tonight. Thus, no plots to show now. However the peak voltage at the Romex far midpoint is now just over 60 V as might be expected instead of the 40 V in my previous plot 2.

Also measured the peak AC voltage on the output side of the breaker and it is about 125 V. Thus, a better showing of the Romex being a resistive voltage divider of 1/2. This also indicates the large voltage drop in the breaker at the 400 A peak level, about 10 to 15 V.

.

 

[gar]

161103-2044 EDT

Worked on my experimental setup. Now I have a moderately clean plot as a replacement for my second plot of my first post.

I still have the high frrequency noise, and I dont.t know the source. This comes in and appears most dominate in the current measurement channel. When the first post's measurements were made there was no noise. The noise was not as bad today, but still present.

To minimize noise I put a 220 ohm 0.01 ufd filter at the scope input for the current channel (red plot). The voltage source that is proportional to current (the current measuring shunt) is an 0.001 ohm resistance. Very low impedance for noise to be coupled in. The following plot still shows some small noise pulses on the blue channel (voltage).

The peak remote point voltage is 60 V relative to the input side of the measuring shunt, and roughly 1/2 the voltage at the input end of the Romex. Expected result. Peak current is slightly less than 500 A. Still the same Sq-D 15 A QO breaker at the main panel.

Observe that all the plots show the current waveform ending at the source voltage zero crossing. This is where the arc from contact opening extinguishes. This is not where the breaker contacts separate. There is no way for the breaker to know where the voltage zero crossing is.

A little over 2 mS before the ending zero crossing you see a small blip in the curve. My off hand guess is this is where breaker contacts separated, and what follows this point is an arc conduction until the next current zero crossing.

In 1962 I measured the transit time for a QO 15 moving contact to to separate thru its 1/2" of travel at 4 mS. If that is reasonably valid, then the arc in my plot below extinguished before the contacts were fully srparated. Arc termination is determined by the current zero crossing, and not contact position. In the plot I believe it took about 5 mS for the QO bimetal strip to heat and unlatch the contacts.

.


.

 

[gar]

161104-2023 EDT

In my last post, numbered 3, note the irregularity in the voltage waveform during the first 5 mS, then there is about a 5 V abrupt change in the voltage curve. After this point the voltage curve looks to be very straight and smooth. In this region of a sine wave the sine curve starts to look moderately straight, but not quite so straight. Also note how smooth the curve is compared with the first 5 mS. Because of the abrupt 5 V change and the smoothness I believe this last 2 mS is the region of arcing. At the moment I don;t have an explanation for the straightness. Other experiments are needed.

The irregularity of the first part of the curve I would explain as fluctuation of the breaker's contacts contact resistance at high current.

.

 

[drcampbell]

RS-422/RS-485 transceivers that tolerate 2.5 kV do exist.
http://cds.linear.com/docs/en/datasheet/1535fb.pdf
(this particular one is pretty pricey - about $10 each)

It seems like a line item for specifying equipment: "Connection terminals shall tolerate an application of 480 vac for one cycle without damage."

 

[gar]

161112-1631 EST

drcampbell:

The LTC1535 you referenced is a nice choice. At DigiKey it is listed at 11.18/1 and 6.59/1000. A number of other manufacturers have isolated units with prices in this range or higher. Isolated interfaces are not monolithic ICs, but rather are hybrid and thus more expensive.

I really favor isolated units and I believe they are actually cost effective if you consider the costs from failures of non-isolated units. The other costs to consider are either in data errors or equipment failure and associated down time and troubleshooting.

But compare a manufacturing cost of just 0.74/1 or 0.32/1000 for a TI 8N75176BDR. This is a monolithic device with common mode of not greater than +15 and -10.

I believe it is reasonable to say that most equipment advertizing 422 or 485 do not use isolated interfaces. Thus, one has to look at the equipment specifications closely and consider what can happen in the real world.

What I have shown is that in a real world situation one can get damaging common mode voltages. Most people are not aware of this source of trouble. I believe most people believe that the EGC wire is equivalent to ground, and this is not true under fault conditions. It is also a problem just from the standpoint of noise.

.

 

[gar]

161112-1705 EST

In another thread, http://forums.mikeholt.com/showthread.php?t=180115 , I discussed measurements on the voltage drop across a switch when the switch was opened in a DC circuit with inductance and an arc occurred.

I ran a new experiment on this thread's circuit, but using a QO15 as the switch being monitored in an AC circuit with a large fault current. This thread is on an AC fault current. The experiment is the same as previously discussed in this thread, but now I am monitoring the voltage drop across the breaker before and after it opens. I expect to see some of the same results here that I saw with ordinary contact switches in a DC circuit.

The test circuit consists of the series combination of a QO20 (the power application switch, it also trips) from 120 V hot, 100 ft of #14 copper wire in a coil (50 ft white, 50 ft black) pretty much non-inductive, the QO15 being tested, a 50 A 100 mV shunt, back to neutral. The Rigol scope is isolated from the AC source via a Sola CV transformer.

Using a DC test current of about 3.5 A the voltage drops and resulting resistance calculations are:
QO15 being tested --- 0.06 V, R = 0.017 ohms
50 A 100 mV shunt --- 0.007 V, R = 0.002 ohms
100 ft #14 copper --- 0.927 V, R = 0.266 ohms
Current limit resistance --- about 11 V, R = 3 ohms
Car battery --- about 12 V

Expected peak current at 120 V, but we don't really have 120 V across this circuit, 170/about 0.3 = 560 A. There are other source impedances that reduce the voltage across the mentioned series circuit, thus actual current is lower.

Following is one plot:

.

.
Application of power to the test is random. It happens when I turn on the QO20 breaker. This occurred 2 mS before a zero crossing. Not enough heating of the trip mechanism to trip yet. Tripping occurs at about a delay of 8 mS, but this is not yet to a current zero crossing, and an arc develops for about 2 mS.

Peak current from the sine wave source occurs slightly before the contacts open. This is about 450 A. Voltage across breaker is about 5 V. A calculation from 0.017 and 480 is 8.2 V. in the ballpark.

When the contacts open the step voltage change is about 14 V. Correlates with other DC switch measurements.

Arc exthinguishes at AC current zero crossing as expected. And the are voltage increases as arc current decreases, negative resistance, as expected. No major resonance problems.

.

 

[gar]

161112-2510 EST

I have found a source for a table of ionization potentials for the elements. See

http://environmentalchemistry.com/yogi/periodic/1stionization.html

Oxygen and nitrogen, near the bottom of the table, are in the range of 13 to 15 V. Thus, there seems to be a correlation with my measurements.

.

 

[gar]

161114-0928 EST

It is interesting to note that when you look at the 18th column, inert gases (outer shell is full of electrons), in the Periodic Table that the ordering of the elements is :

Element -- Ionization Potential -- Atomic # -- Row #

He, 24.6, # 1, row 1;
Ne, 21.6, # 10, row 2;
Ar, 15.8, # 18, row 3;
Kr, 14.0, # 36, row 4;
Xe, 12.1, # 54, row 5;

and their ionization potentials follow the same order, but inversely.

Other gases:

O, 13.7, atomic # 8, row 2;
N, 14.5, atomic # 7, row 2;
H, 13.6, atomic # 1, row 1.

.

 

[wirenut1980]

161112-1631 EST

drcampbell:

The LTC1535 you referenced is a nice choice. At DigiKey it is listed at 11.18/1 and 6.59/1000. A number of other manufacturers have isolated units with prices in this range or higher. Isolated interfaces are not monolithic ICs, but rather are hybrid and thus more expensive.

I really favor isolated units and I believe they are actually cost effective if you consider the costs from failures of non-isolated units. The other costs to consider are either in data errors or equipment failure and associated down time and troubleshooting.

But compare a manufacturing cost of just 0.74/1 or 0.32/1000 for a TI 8N75176BDR. This is a monolithic device with common mode of not greater than +15 and -10.

I believe it is reasonable to say that most equipment advertizing 422 or 485 do not use isolated interfaces. Thus, one has to look at the equipment specifications closely and consider what can happen in the real world.

What I have shown is that in a real world situation one can get damaging common mode voltages. Most people are not aware of this source of trouble. I believe most people believe that the EGC wire is equivalent to ground, and this is not true under fault conditions. It is also a problem just from the standpoint of noise.

.

Thank you gar, very interesting experiment. The approximately 60 V (40 V in one plot) you measured in your first 2 plots was between which two points? The EGC at your main panel and the ground pin for the communications cable? Wouldn't there be bonding between the EGC and the RSXXX communications cable ground pin (instead of floating the EGC like in your experiment)?

 

[gar]

161116-1714 EST

wirenut1980:

I will try to clarify the problem I am trying to describe, and how my experiment relates.

Thank you gar, very interesting experiment. The approximately 60 V (40 V in one plot) you measured in your first 2 plots was between which two points? The EGC at your main panel and the ground pin for the communications cable? Wouldn't there be bonding between the EGC and the RSXXX communications cable ground pin (instead of floating the EGC like in your experiment) [/QUOTE] When I use an Apple computer quote doesn't work. Presently using Apple, Quoye works fine on Windows.

Consider two circuits coming from a main panel. Could even be on the same breaker. But for simplicity use two breakers, same phase or different doesn't matter. Each circuit has its own EGC terminated at the same point in the main panel.

At the end of each circuit its equipment's RS485 interface is referenced to its EGC. The two RS485 ports are connected to each other with a single twisted pair. No need to consider a shield, but if one existed it should be connected at only one end.

Have one end of the communication path a computer and the other end a CNC machine.

Normally there is no voltage drop along either EGC path. Thus, there is no voltage difference between the chassis of the Computer and the CNC chassis. Good communication can occur.

Now at the CNC apply a dead short between a 120 V hot wire and EGC. The EGC at the CNC end will rise to near 1/2 the peak voltage of the CNC hot wire at the main panel relative to the voltage of the EGC at the main panel. Because the EGC to the computer has no current flow the computer EGC is the same voltage as at he main panel. Thus, the two RS485 ports have the 1/2 voltage applied between them. This destroys non-isolated RS485 chips.

My experiment experimentally demonstrates this. This I will describe after dinner.

.

 

[gar]

161116-1944 EST

wirenut1980:

Now to continue with the discussion.

Essentially the computer's EGC wire has zero voltage drop along its length. No current in that EGC. Therefore, no matter where the computer is located, at the main panel (0 length for its cable), or 1000 ft away, the computer chassis is at the same potential as the neutral or ground bus at the main panel.

In most cases the EGC wire in a circuit will be of the same diameter and material, or smaller, compared to its associated current carrying conductors. Thus, for the experiment I used two wires of the same resistance, black and white of a #14 Romex copper cable.

Black and white were shorted together with an AB 300 V terminal block at one end of the cable. This shorted end is the destination end of the cable (CNC machine in my theoretical model). In my initial test the 50 ft Romex cable was left in its shipping coiled state. This meant both ends of the cable were physically within about 1 ft of each other. Another 6 ft or so of #14 wire connected from the Romex coil to the bus bars and breaker in the main panel. The 50 A 100 mV shunt was inserted in the white wire at the source end of the Romex. A twisted wire pair from the shunt goes to the B scope channel providing common for both the current and voltage scope inputs, and the current signal. A 10x voltage probe without use of its common wire was used for the voltage measurement.

The first plot had the scope voltage probe connected to the AB terminal block that shorted the destination end of the cable. Thus, measuring the mid point of a 1/2 resistive voltage divider.

We can expect to see a substantial voltage drop at the main panel from the about 500 A peak load. This is shown in the third plot where the scope voltage probe was on the main panel hot bus showing the input voltage to the breaker. This drop is about (3.5-2.6)*50 = 45 V. The breaker has about 0.017*480 = 8.2 V drop. Peak voltage without load is about 3.5*50 = 175 V. This gives us about 175-45-8 = 122 V as the input to the Romex at 480 A. Thus, 122/2 is 61 V and a good estimate of the measured 60 V at the far end mid-point.

My second plot showing 40 V is clearly an instrumentation error problem, and a result of the sloppy voltage connection I made.

Basically the first plot voltage measurement is from the main panel neutral/ground point to the CNC EGC at the CNC under a short circuit condition of hot to EGC at the CNC.

Have I provided any clarification?

.

 

[wirenut1980]

Yes, thank you :thumbsup:. There is voltage drop on the EGC from the panel to the CNC machine carrying fault current that amounts to roughly half the supply voltage. This voltage drop results in there being a voltage on that EGC with respect to "remote earth" or zero potential, which exists on the computer EGC on a different circuit. This potential difference can cause malfunction or damage to the communication cable and possibly the equipment that is connected to either end of it due to current flow caused by the potential difference. If the communication cable has an isolated interface, it would not allow current flow on the communication cable.

I'm curious for your thoughts on whether there might be a similar effect for a ground fault on the utility system since the primary and secondary neutrals of the utility are typically bonded, and the secondary neutral is bonded to the EGC at the main panel.

 

[GoldDigger]

Yes, thank you :thumbsup:. There is voltage drop on the EGC from the panel to the CNC machine carrying fault current that amounts to roughly half the supply voltage. This voltage drop results in there being a voltage on that EGC with respect to "remote earth" or zero potential, which exists on the computer EGC on a different circuit. This potential difference can cause malfunction or damage to the communication cable and possibly the equipment that is connected to either end of it due to current flow caused by the potential difference. If the communication cable has an isolated interface, it would not allow current flow on the communication cable.

I'm curious for your thoughts on whether there might be a similar effect for a ground fault on the utility system since the primary and secondary neutrals of the utility are typically bonded, and the secondary neutral is bonded to the EGC at the main panel.

It is indeed possible for the utility neutral, as delivered to the service point, to go well above remote earth potential during a utility fault.
The saving grace is that for a single building service this results in the building neutral and all building EGC to move together to that higher voltage.
Because all EGCs should originate from one point, where the service bond is located, data runs inside one building should not be harmed, unless somebody put in an isolated ground with a local ground rod for some of the equipment. :happysad:
Runs that go from one building to another are much more vulnerable to excursions of the utility neutral and local GES caused by faults.

 

[gar]

16117-2224 EST

wirenut1980:

In your post numbered 13 you made a concise description of what I have been saying.

GoldDigger has answered your question.

Now consider a home with a 20 ohm to earth ground rod, and there is a direct hit by a bolt of lightning with a 10,000 A capability.

.

 

[GoldDigger]

16117-2224 EST

wirenut1980:

In your post numbered 13 you made a concise description of what I have been saying.

GoldDigger has answered your question.

Now consider a home with a 20 ohm to earth ground rod, and there is a direct hit by a bolt of lightning with a 10,000 A capability.

.

As a supplement to the original answer, yet to be given )), I speculate that the voltage of the lightning strike will drastically decrease the effective resistance of the ground rod, but it will not be enough to save you.

 

[drcampbell]

This might be one instance where an isolated dedicated ground is valuable.
Bond all the cabinets with an EGC per usual practice.
Isolate the data devices from the cabinets.
Run a dedicated EGC from the panel to the transmitter to the receiver, over which fault current will not flow, sparing the RS-485 transceivers from excess-common-mode-voltage damage.