Possible Latchup problem with RS485 transciever

[logjam]

I have a project where the input of an RS-485 receiver can have
voltage on it before the main power supply comes on.

My project has worked well in the past when my cable's bus voltage was
12v and I used some 12v to 5vDC Astrodyne DC-DC converters.

I am now using a cable bus voltage of 48vDC and some MeanWell DC-DC
converters.

I have experienced strange part failures since switching over to the
48vDC supplies. I think the source of the problem may either be
latchup or the regulated output of the DC-DC converter rising too
slowly. It takes about 200us to go from 0v to 6v and within 50us has
settled to 5v. None of the parts should be damaged from 6v, even
continuously. The parts rarely survive 5-10 power cycles.

I have two signals, one that is about 50% duty cycle at 1MHz, and the
other that is "on" for about 2us and "off" for about 5ms. So far I
have not experienced a problem with the RS-485 receiver connected to
the 2us/5ms signal, just the 1MHz signal. The duty of the signal
shouldn't matter with RS-485, right? Considering A will be what B is
not...

The part failure seems to be where one of the RS-485 inputs either
sinks or sources current into the transmission line. With one part I
observed a 10 ohm resistance from an input pin to ground. This part
does seem to still have some functionality since if both inputs of the
receiver have the same signal the "R" output of the receiver will
mirror its inputs, however there is still current being either sinked
or sourced depending on the specific chip and how it decided to die, I
guess...

The RS-485 devices that I have observed this with are the 75ALS180 and
the MAX3467. Neither are specifically "Hot Swappable".

From reading the Maxim IC data sheet

http://datasheets.maxim-ic.com/en/ds/MAX3465-MAX3469.pdf

I don't understand what makes the MAX3467 not hot swappable and the
3468 hot swappable. As far as I can tell, the feature only makes sure
the transmit stage doesn't activate during processor initialization.
This isn't my concern since I only receive a signal.

I've read about solutions to protect the ICs against latchup. One of
them involved series resistors. I may try to place 10k resistors in
series with the A+B signals after the termination resistor. I can't
imagine that these devices should behave like this, latchup shouldn't
occur on a device like this from the differential input pair.

Can anyone suggest what may be causing my problems?.Which one of these
ideas should I use? Pull up/down resistors? The RS-485 standard
already allows for voltages above VCC and below GND, so if I'm not
crazy I shouldn't be having issues with these parts! The only part I
have left with enough quantity to work with is the LTC485.

Sorry for the long unorganized message. Its been a very long day!
I'll be working on this project all weekend...

Thanks,
Grant

 

[Nico Coesel]

I have a project where the input of an RS-485 receiver can have
voltage on it before the main power supply comes on.

My project has worked well in the past when my cable's bus voltage was
12v and I used some 12v to 5vDC Astrodyne DC-DC converters.

I am now using a cable bus voltage of 48vDC and some MeanWell DC-DC
converters.

I have experienced strange part failures since switching over to the
48vDC supplies. I think the source of the problem may either be
latchup or the regulated output of the DC-DC converter rising too
slowly. It takes about 200us to go from 0v to 6v and within 50us has
settled to 5v. None of the parts should be damaged from 6v, even
continuously. The parts rarely survive 5-10 power cycles.

Perhaps you have problems with some inductance somewhere causing the
voltage spike. Connect a 10u to 100uf capacitor directly across the
RS485 chip. I've had a similar problem with a circuit and placing the
extra capacitor solved the problem nicely.

 

[Grant Stockly]

Perhaps you have problems with some inductance somewhere causing the

voltage spike. Connect a 10u to 100uf capacitor directly across the
RS485 chip. I've had a similar problem with a circuit and placing the
extra capacitor solved the problem nicely.

Right now I have a 1u tantalum near the ICs and (2) .1u ceramics as
close to each driver/receiver as possible. I've just swapped in the
LTC485 and I haven't been able to kill it. I almost gave myself
carpel tunnel twisting the cable on and off... : )

The cable that connects to this board contains 48vDC and the two
RS-485 pairs. I need the circuit to withstand live connections/
disconnections with any order of power/ground/485 connections.

I have been working on the issue for a few hours now, and have a lot
more information to add. (schematic, scope captures) I will update
this thread soon.

Thanks!

Grant

 

[Grant Stockly]

Thanks for the response! I have investigated the problem further this
afternoon and have some new information. I will add it below the
responses to your questions.

6V is abs max for the chip you mention below. I would definitely get rid
of that overshoot.

How long is it "OK" to have a ringing signal or transient exceed the
VCC or GND limits? Not at all? I'm primarily a digital person. Will
a zener diode on the power supply clamp fast enough to remove the
transients? I suppose a zener along with a very low ESR electrolytic
would help. Right now all I have is a 1u tantalum, and whatever is in
the output stage of those switching DC-DC power supplies.

Make sure you don't exceed the input voltage abs max. They do not
mention any subtrate diode maximum current which is kind of strange. I'd
ask them about it.
For series resistors to work you would need to know the substrate diode
ratings.

Can you post a schematic?

Yes, here it is:
http://media.diywelder.com/RS-485/Root_1.pdf
-I included 5 "load" resistors on the PCB to satisfy the minimum load
required to meet the regulation specification of the power supplies.

New information:
-The 75ALS180 ICs in their "dead" eventually reach a temperature of
130F.

-Quite frequently, but not all the time, I experience transients and/
or fairly long negative voltage periods when first making the
connection with the circuit.
http://media.diywelder.com/RS-485/poweron-close.gif
http://media.diywelder.com/RS-485/plug in transient.gif
http://media.diywelder.com/RS-485/noise on connection.gif

The negative part of the first to examples worries me more than the
positive part. The last capture was VCC during connecting the
circuit. At 100ns/div, would a zener be fast enough to protect
against these positive spikes?

-When NO power supply is connected to the PCB, only Clock A/B and/or
Line A/B, the 75ALS180 in this "dead" state (I don't have any good
ones left) will cause the VCC pin to reach 2vDC. It does this
approximately 10ms after the PCB has been connected. ALL of the
capture examples in this section are from a chip that I consider to be
dead. I will specifically state when a capture is from a good
environment.

http://media.diywelder.com/RS-485/poweron-bus disconnected-power supply disconnected and applied.gif
The above screen capture shows the 48vDC and 5v VCC voltages when the
A/B data pairs are NOT connected. The DC-DC converter turns on
approximately 350ms after 48v is present

http://media.diywelder.com/RS-485/poweron-close.gif
Sometimes, but not always, the above example occurs.

http://media.diywelder.com/RS-485/poweron-bus ramps vcc up to 2v.gif
When plugged in, before the DC-DC converter turns on, the bus ramps up
to 2v.

http://media.diywelder.com/RS-485/poweron-bus and power supply connected at the same time.gif
The above screen capture shows the 48vDC applied to the power supply,
VCC ramping up to 2v, and then about 80ms later VCC ramping up to 5v.
The DC-DC converter hadn't been disconnected for very long, as you can
see it had about 8v in its capacitors when I powered it up. This may
explain the faster turn on time compared to other examples.

http://media.diywelder.com/RS-485/p...ted-power supply disconnected and applied.gif
If I leave the bus connected, but disconnect the power supply, VCC
will remain at 2v. If 48v is applied after the bus was connected, the
above capture shows the DC/DC converters response. Doesn't look high
impedance to me! : )

GOOD -I can't manage to kill the LTC485. Its data inputs remain high
impedance no matter how poorly I try to connect and disconnect the
cable. What killed 75ALS180/MAX3467 IC after IC last night doesn't
seem to hurt this part.

http://media.diywelder.com/RS-485/LTC485 operation.gif
Here is VCC coming up and the "R" output of the LTC485 activating

http://media.diywelder.com/RS-485/LTC485 output.gif
The "R" output. This is good. The other ICs when dead do not give me
an output, they only load the lines.

http://media.diywelder.com/RS-485/1M signal drive.gif
This is the A and B lines. The B line did not have a ground clamp,
only A.

http://media.diywelder.com/RS-485/bus in and out of circuit spread apart.gif
This is the A and B lines (A on scope probe, B on probe ground) when
terminated. The faint gray trace is out of circuit, but terminated,
and the black trace is connected to the LTC485 and terminated. I'm
not sure if the signal was being offset by the scope's ground. I had
to disconnect the scope's RS-232 port from my computer (which shares a
ground with the RS-485 circuit). Before disconnecting the RS-232
cable it looked a lot worse. The scope's own ground could still be
impacting the signal presentation for all I know...

So, I know the MAXIM and TI ICs are damaged. I don't know HOW they
were damaged. It could be a latchup type failure from seeing voltage
on the RS-485 inputs before VCC came up. I hope that isn't the case
since in the "real world" a device could be connected to a network and
powered down, right?

The second possibility is that the transients and negative voltage
periods damage the 75ALS180 and MAX3467, but not the LTC485. The
LTC485 may be designed with better internal protection to such
events? The LTC485 specifies a maximum input voltage of 12v, which
would help with any possible positive voltage periods which would
otherwise damage a 5v part.

All that aside, the LTC485 data sheet specifically says "Power-Up/Down
Glitch-Free Driver Outputs Permit Live Insertion or Removal of
Transceiver" and "Driver Maintains High Impedance in Three-State or
with the Power Off". Since this is a half duplex device, driver and
receiver share the same inputs. Therefore the data sheet clearly says
that the "driver" pins can see voltage when the power is off.

The 75ALS180 data sheet says "Glitch-Free Power-Up and Power-Down
Protection", which is not quite as spelled out, but from a RS-485
point of view indicates to me that it won't disturb network traffic
during power-up and down. The MAXIM data sheet doesn't specifically
speak about power-up and power-down except when referring to their
"Hot-Swap Versions", but in that case it is only referring to a set-
reset setup inside the IC which prevents a high impedance driver
enable from accidentally disturbing the bus during power up. The
device I'm using is not a hot swap version, but I'm also not having
and will never have a problem with two devices trying to drive the
same line.

In closing, I would hope that all 3 of the RS-485 devices would be
perfectly capable of handling voltage on their inputs when the power
supply is off. If this is not the case, please tell me! It only
seems logical! : ) The common mode voltage can already exceed VCC
and go below ground, so it shouldn't care when the power is off,
right? Ground+12v and Ground-7v is always the same regardless of
where VCC is...

Is it possible, or logical, to assume that a transient or negative
voltage could cause an input to conduct power to VCC? Maybe an input
is conducting through a damaged protection diode to VCC???

I hope that I have described the problem clear enough that someone can
tell me without too much effort what my problem is! I would like to
fix the problem so that I can no longer kill 75ALS180s on demand. I
will then switch to using the LTC parts which have proven to be at
least a little better in this environment. I figure that if the
circuit will work with the suicidal parts, it will work even better
with the LTC parts. I'm not comfortable with using the LTC parts
until I'm sure the problem has been fixed. : )

Thanks for your time,
Grant

 

[Joerg]

Thanks for the response! I have investigated the problem further this
afternoon and have some new information. I will add it below the
responses to your questions.


How long is it "OK" to have a ringing signal or transient exceed the
VCC or GND limits? Not at all? I'm primarily a digital person. Will
a zener diode on the power supply clamp fast enough to remove the
transients? I suppose a zener along with a very low ESR electrolytic
would help. Right now all I have is a 1u tantalum, and whatever is in
the output stage of those switching DC-DC power supplies.

I am not sure how it all sits together but let me give some pointers anyway.

The "real" abs max is usually a few percent higher but that depends on
how conservative a manufacturer is. Once you reach "real" abs max levels
you must not exceed that at all. Never, ever.

1uF is not a lot, try more. And personally I do not use tantalums in
supply rails, ever.

Yes, here it is:
http://media.diywelder.com/RS-485/Root_1.pdf
-I included 5 "load" resistors on the PCB to satisfy the minimum load
required to meet the regulation specification of the power supplies.

There are just parts and ports, not sure where it all goes. Consider
protection measures especially if the data line is long or runs in a
noisy environment. You might need transzorbs, ferrites and the like.

New information:
-The 75ALS180 ICs in their "dead" eventually reach a temperature of
130F.

That is quite typical after an overvoltage meltdown. If the power supply
had more punch you'd probably see some smoke.

-Quite frequently, but not all the time, I experience transients and/
or fairly long negative voltage periods when first making the
connection with the circuit.
http://media.diywelder.com/RS-485/poweron-close.gif
http://media.diywelder.com/RS-485/plug in transient.gif
http://media.diywelder.com/RS-485/noise on connection.gif

Ouch!


The negative part of the first to examples worries me more than the
positive part. The last capture was VCC during connecting the
circuit. At 100ns/div, would a zener be fast enough to protect
against these positive spikes?

Yes but that won't work for production because zeners have too much
tolerance. You could concoct something around a TL431 but why not use
more bypass capacitance? It's cheap.

The negative part can be capped by a nice Schottly diode.

-When NO power supply is connected to the PCB, only Clock A/B and/or
Line A/B, the 75ALS180 in this "dead" state (I don't have any good
ones left) will cause the VCC pin to reach 2vDC. ...

That is normal. It is being fed via the substrate diodes. That can kill
chips if VCC ratchets up past abs max.

... It does this
approximately 10ms after the PCB has been connected. ALL of the
capture examples in this section are from a chip that I consider to be
dead. I will specifically state when a capture is from a good
environment.

http://media.diywelder.com/RS-485/poweron-bus disconnected-power supply disconnected and applied.gif
The above screen capture shows the 48vDC and 5v VCC voltages when the
A/B data pairs are NOT connected. The DC-DC converter turns on
approximately 350ms after 48v is present

http://media.diywelder.com/RS-485/poweron-close.gif
Sometimes, but not always, the above example occurs.

http://media.diywelder.com/RS-485/poweron-bus ramps vcc up to 2v.gif
When plugged in, before the DC-DC converter turns on, the bus ramps up
to 2v.

http://media.diywelder.com/RS-485/poweron-bus and power supply connected at the same time.gif
The above screen capture shows the 48vDC applied to the power supply,
VCC ramping up to 2v, and then about 80ms later VCC ramping up to 5v.
The DC-DC converter hadn't been disconnected for very long, as you can
see it had about 8v in its capacitors when I powered it up. This may
explain the faster turn on time compared to other examples.

http://media.diywelder.com/RS-485/p...ted-power supply disconnected and applied.gif
If I leave the bus connected, but disconnect the power supply, VCC
will remain at 2v. If 48v is applied after the bus was connected, the
above capture shows the DC/DC converters response. Doesn't look high
impedance to me! : )

GOOD -I can't manage to kill the LTC485. Its data inputs remain high
impedance no matter how poorly I try to connect and disconnect the
cable. What killed 75ALS180/MAX3467 IC after IC last night doesn't
seem to hurt this part.

LTC makes good quality chips. Still, don't push your luck there either ;-)

http://media.diywelder.com/RS-485/LTC485 operation.gif
Here is VCC coming up and the "R" output of the LTC485 activating

http://media.diywelder.com/RS-485/LTC485 output.gif
The "R" output. This is good. The other ICs when dead do not give me
an output, they only load the lines.

http://media.diywelder.com/RS-485/1M signal drive.gif
This is the A and B lines. The B line did not have a ground clamp,
only A.

That's a lot of ringing. I'd consider a little AC termination here.

http://media.diywelder.com/RS-485/bus in and out of circuit spread apart.gif
This is the A and B lines (A on scope probe, B on probe ground) when
terminated. The faint gray trace is out of circuit, but terminated,
and the black trace is connected to the LTC485 and terminated. I'm
not sure if the signal was being offset by the scope's ground. I had
to disconnect the scope's RS-232 port from my computer (which shares a
ground with the RS-485 circuit). Before disconnecting the RS-232
cable it looked a lot worse. The scope's own ground could still be
impacting the signal presentation for all I know...

So, I know the MAXIM and TI ICs are damaged. I don't know HOW they
were damaged. It could be a latchup type failure from seeing voltage
on the RS-485 inputs before VCC came up. I hope that isn't the case
since in the "real world" a device could be connected to a network and
powered down, right?

Depends on your setup but my impression is that your chips were killed
by overvoltage.

The second possibility is that the transients and negative voltage
periods damage the 75ALS180 and MAX3467, but not the LTC485. The
LTC485 may be designed with better internal protection to such
events? The LTC485 specifies a maximum input voltage of 12v, which
would help with any possible positive voltage periods which would
otherwise damage a 5v part.

I think the MAX was spec'd for 13V or so abs max but I don't know what
happens when that's exceeded.

All that aside, the LTC485 data sheet specifically says "Power-Up/Down
Glitch-Free Driver Outputs Permit Live Insertion or Removal of
Transceiver" and "Driver Maintains High Impedance in Three-State or
with the Power Off". Since this is a half duplex device, driver and
receiver share the same inputs. Therefore the data sheet clearly says
that the "driver" pins can see voltage when the power is off.

The 75ALS180 data sheet says "Glitch-Free Power-Up and Power-Down
Protection", which is not quite as spelled out, but from a RS-485
point of view indicates to me that it won't disturb network traffic
during power-up and down. The MAXIM data sheet doesn't specifically
speak about power-up and power-down except when referring to their
"Hot-Swap Versions", but in that case it is only referring to a set-
reset setup inside the IC which prevents a high impedance driver
enable from accidentally disturbing the bus during power up. The
device I'm using is not a hot swap version, but I'm also not having
and will never have a problem with two devices trying to drive the
same line.

In closing, I would hope that all 3 of the RS-485 devices would be
perfectly capable of handling voltage on their inputs when the power
supply is off. If this is not the case, please tell me! It only
seems logical! : ) The common mode voltage can already exceed VCC
and go below ground, so it shouldn't care when the power is off,
right? Ground+12v and Ground-7v is always the same regardless of
where VCC is...

Keep in mind that there always comes a point where it either breaks down
or where a substrate diode begins to conduct and then quietly feeds your
supply rail. And it doesn't know when to stop ;-)

Is it possible, or logical, to assume that a transient or negative
voltage could cause an input to conduct power to VCC? Maybe an input
is conducting through a damaged protection diode to VCC???

Yes, very possible. I'd ask the vendors at what point that will happen.
Hard to say with datasheets as skimpy as for the MAX. I'd ask. Email
often works well (but don't necessarily hold your breath with TI).

I hope that I have described the problem clear enough that someone can
tell me without too much effort what my problem is! I would like to
fix the problem so that I can no longer kill 75ALS180s on demand. I
will then switch to using the LTC parts which have proven to be at
least a little better in this environment. I figure that if the
circuit will work with the suicidal parts, it will work even better
with the LTC parts. I'm not comfortable with using the LTC parts
until I'm sure the problem has been fixed. : )

Take a look at the 75ALS180 datasheet, schematic of inputs and outputs:
http://focus.ti.com/lit/ds/symlink/sn65als180.pdf

You can see 18K and 3K and yes, those in series will feed the supply
rail but it seems two diodes in series with 1.1K should keep that down
to a few volts, depending on how much arrives on the bus lines.