Balanced grounding between devices & fault

[Shmuel Davis]

I have designed a communication device that is being powered by 24V.
It has two types connectors. The first provides the power and control
signals, while the second provides communication to the outside world.
The device is contained within a metallic box that is completely
isolated.

It was found that some of the devices failed at initial power on, or
due to a later cycling of the power. Those devices were found to pull
the 24V DC rail down to < 5V DC and would draw in excess of 200mA.
There is a possibility that the device fails to handle the system
inrush current.

In my analysis we found there to be a short between the +3.3V supply
and GND. When the +3.3V LDO and the Altera gate array that was driven
by the +3.3V was replaced, the device functioned correctly.

On the input to the control signals, we are using a National
Semiconductor LM2676 3A Step-Down Voltage Regulator to step down the
voltage to +5V. We are then using a National Semiconductor LM3940 LDO
Regulator to provide us with the required +3.3V for the Altera FPGA.

All the communication signals from the outside world are floating and
are passed through transformers and Holt transceivers that are powered
by the above +3.3V supply.

For those devices that the fault was found, the shield from the cables
that provide communication to the outside world were not connected to
the GND of the device. I have two devices where this connection was
made and the fault did not occur.

Any suggestions of might be the source of the fault ? Suggestions have
been made that a GND imbalance between my device and other devices it
is connected to, may have caused damage to the components of my device
(possibly large currents into my device). However I don't understand
how this can happen when there is no connection between the other
devices and the GND of my device via the cable shield ? Can spiking
occur across a SM0805 footprint that could have caused damage to the
gate array if the voltage difference gets high enough between the two
grounds ? Also another point: In a normal connection where the cable
shield does balance the grounds between devices – why doesn't this
balancing cause large currents to flow between the devices ?

Does anyone have any other suggestions ?

 

[The other John Smith]

I have designed a communication device that is being powered by 24V.
It has two types connectors. The first provides the power and control
signals, while the second provides communication to the outside world.
The device is contained within a metallic box that is completely
isolated.

It was found that some of the devices failed at initial power on, or
due to a later cycling of the power. Those devices were found to pull
the 24V DC rail down to < 5V DC and would draw in excess of 200mA.
There is a possibility that the device fails to handle the system
inrush current.

In my analysis we found there to be a short between the +3.3V supply
and GND. When the +3.3V LDO and the Altera gate array that was driven
by the +3.3V was replaced, the device functioned correctly.

On the input to the control signals, we are using a National
Semiconductor LM2676 3A Step-Down Voltage Regulator to step down the
voltage to +5V. We are then using a National Semiconductor LM3940 LDO
Regulator to provide us with the required +3.3V for the Altera FPGA.

All the communication signals from the outside world are floating and
are passed through transformers and Holt transceivers that are powered
by the above +3.3V supply.

For those devices that the fault was found, the shield from the cables
that provide communication to the outside world were not connected to
the GND of the device. I have two devices where this connection was
made and the fault did not occur.

Any suggestions of might be the source of the fault ? Suggestions have
been made that a GND imbalance between my device and other devices it
is connected to, may have caused damage to the components of my device
(possibly large currents into my device). However I don't understand
how this can happen when there is no connection between the other
devices and the GND of my device via the cable shield ? Can spiking
occur across a SM0805 footprint that could have caused damage to the
gate array if the voltage difference gets high enough between the two
grounds ? Also another point: In a normal connection where the cable
shield does balance the grounds between devices - why doesn't this
balancing cause large currents to flow between the devices ?

Does anyone have any other suggestions ?

While you are looking for the cause of the problem, send the failed devices
to Altera for a failure analysis. They might be able to point you to a
specific pin.

Good luck.

 

[The other John Smith]

I have designed a communication device that is being powered by 24V.
It has two types connectors. The first provides the power and control
signals, while the second provides communication to the outside world.
The device is contained within a metallic box that is completely
isolated.

It was found that some of the devices failed at initial power on, or
due to a later cycling of the power. Those devices were found to pull
the 24V DC rail down to < 5V DC and would draw in excess of 200mA.
There is a possibility that the device fails to handle the system
inrush current.

In my analysis we found there to be a short between the +3.3V supply
and GND. When the +3.3V LDO and the Altera gate array that was driven
by the +3.3V was replaced, the device functioned correctly.

On the input to the control signals, we are using a National
Semiconductor LM2676 3A Step-Down Voltage Regulator to step down the
voltage to +5V. We are then using a National Semiconductor LM3940 LDO
Regulator to provide us with the required +3.3V for the Altera FPGA.

All the communication signals from the outside world are floating and
are passed through transformers and Holt transceivers that are powered
by the above +3.3V supply.

For those devices that the fault was found, the shield from the cables
that provide communication to the outside world were not connected to
the GND of the device. I have two devices where this connection was
made and the fault did not occur.

This sounds like there is a destructive EMF developing between the signal
lines and ground at power-on. Mismatched transmission line effects? Can you
add a resistor in the lines and a TVS? Also, if it works and does not fail
with the shield connected, why not leave it connected?

Any suggestions of might be the source of the fault ? Suggestions have
been made that a GND imbalance between my device and other devices it
is connected to, may have caused damage to the components of my device
(possibly large currents into my device). However I don't understand
how this can happen when there is no connection between the other
devices and the GND of my device via the cable shield ? Can spiking
occur across a SM0805 footprint that could have caused damage to the
gate array if the voltage difference gets high enough between the two
grounds ? Also another point: In a normal connection where the cable
shield does balance the grounds between devices - why doesn't this
balancing cause large currents to flow between the devices ?

It could cause large currents to flow if the source impedance of the
impalance is low enough. However, the source impedance of the imbalance
could be low enough to destroy a semiconductor without causing "large"
currents to flow. What is the current due to ground imbalance?

 

[Ken Smith]

The device is contained within a metallic box that is completely
isolated.

Imagine a walk across a carpet and touch the box. A spark jumps from my
hand to the metal box. Where does it go from there on its way back to the
earth?

 

[Shmuel Davis]

This sounds like there is a destructive EMF developing between the signal
lines and ground at power-on. Mismatched transmission line effects? Can you
add a resistor in the lines and a TVS?

How would such an EMF cause this type of damage and how would your
suggested solution solve the problem ?
What is a TVS ? What would be the connection of this resistor/TVS -
from shield to GND or some other form of connection (The signals on
the bus are a differential pair).

Also, if it works and does not fail with the shield connected, why not leave > it connected?

This was going to be one of the solutions. However, I wanted to try
and come up with some sort of theory of what was causing the fault. I
was also worried that someone had suggested that if we do short the
shield to GND then we may cause large currents to flow, doing more
damage than good (even though the devices with the jumper in it did
not show this fault !)

It could cause large currents to flow if the source impedance of the
impalance is low enough. However, the source impedance of the imbalance
could be low enough to destroy a semiconductor without causing "large"
currents to flow. What is the current due to ground imbalance?

Is there any way to measure these phenomena you mentioned above ? We
do not have access to the system that the devices were placed in so we
can't replicate the precise circumstances under which the fault
occurred (we are not able to measure the "current due to ground
imbalance" and I was only suggestion that this might be a cause of the
fault). Is there any way within an "artificial" labratory environment
for us to narrow down what the source of the fault is (and thus the
remedy ?)