PIC resets when powering fluorescent lights

[Richard Swinfen]

Need some advice on powering a bank of fluorescent lights. My
controller uses a PIC 16F876 to a 2803 Darlington array to power 10A
240V relays one of which powers the 240v AC light bank. There are
about 9 lights in the bank (my garage), ranging from 40W to 70W, each
with their own starter, amounting to about 450W operating power.
Other relays power other light banks lesser in number and operating
power, with no problem.

The problem symptom is a resetting PIC when the relay is fired. The
same happens if the lights are powered by the original light switch.
The problem cause I believe is inductive spikes during the lights
power-up cycle. I have scanned the groups and can't see this specific
problem being discussed. What is discussed has led me to try
varistors across the load and across the relay contacts both at the
relay. I think, by doing that, I am preventing the spikes from being
inducted through the relay into the DC power lines, spiking the PIC.
However, I have read that this might not prove a good solution if
spiking is regular. In any case, it doesn't work, and in fact seems
to cause the problem to occur more consistently now.

I haven't yet attempted an RC snubber - not totally convinced on what
I would be achieving with that and would prefer some specific advice
on the problem first.

If you can help, I would be most grateful.

Regards,

Richard.

 

[John Popelish]

Richard Swinfen wrote:

Need some advice on powering a bank of fluorescent lights. My
controller uses a PIC 16F876 to a 2803 Darlington array to power 10A
240V relays one of which powers the 240v AC light bank. There are
about 9 lights in the bank (my garage), ranging from 40W to 70W, each
with their own starter, amounting to about 450W operating power.
Other relays power other light banks lesser in number and operating
power, with no problem.

The problem symptom is a resetting PIC when the relay is fired. The
same happens if the lights are powered by the original light switch.
The problem cause I believe is inductive spikes during the lights
power-up cycle. I have scanned the groups and can't see this specific
problem being discussed. What is discussed has led me to try
varistors across the load and across the relay contacts both at the
relay. I think, by doing that, I am preventing the spikes from being
inducted through the relay into the DC power lines, spiking the PIC.
However, I have read that this might not prove a good solution if
spiking is regular. In any case, it doesn't work, and in fact seems
to cause the problem to occur more consistently now.

I haven't yet attempted an RC snubber - not totally convinced on what
I would be achieving with that and would prefer some specific advice
on the problem first.

If you can help, I would be most grateful.

Regards,

Richard.

Obviously, the large transient voltages from the relay power side are
getting back into the PIC. A good starting point would be a complete
schematic of the whole system, so that we can discuss coupling
mechanisms and fixes. Can you draw something up , get it converted to
GIF format and post to alt.binaries.schematics.electronic? Or email
it to me and I will post it for you.
--
John Popelish

 

[Richard Swinfen]

Good idea John - only my 'schematics' are more layout diagrams in my own
'style'! Moreover the controller device does a lot more as a slave device
for a multi-node networked home automation system, I wouldn't want to
complicate the question. I do believe, perhaps naively, that a simplified
device would have the same problem. I could knock up something that
represents the area of the circuit that deals with this, but in an attempt
to expedite an answer, I'll try to describe the rest of the relevant
circuit. The PIC, 2803 and relays are powered by a 7805 with recommended
caps on both sides (1uf & .33uf) itself powered by a 1.2A 12v regulated
supply. Another possibly relevant point is that the PIC determines
zero-crossing using a dirty-but-working 10MOhm direct live feed to RB0. I'll
check removing this doesn't fix it.

Does that help?

Cheers,

Richard.

"John Popelish" <jpopelish@rica.net> wrote in message
news:41AFAF58.52D22D85@rica.net...

Richard Swinfen wrote:

Need some advice on powering a bank of fluorescent lights. My
controller uses a PIC 16F876 to a 2803 Darlington array to power 10A
240V relays one of which powers the 240v AC light bank. There are
about 9 lights in the bank (my garage), ranging from 40W to 70W, each
with their own starter, amounting to about 450W operating power.
Other relays power other light banks lesser in number and operating
power, with no problem.

The problem symptom is a resetting PIC when the relay is fired. The
same happens if the lights are powered by the original light switch.
The problem cause I believe is inductive spikes during the lights
power-up cycle. I have scanned the groups and can't see this specific
problem being discussed. What is discussed has led me to try
varistors across the load and across the relay contacts both at the
relay. I think, by doing that, I am preventing the spikes from being
inducted through the relay into the DC power lines, spiking the PIC.
However, I have read that this might not prove a good solution if
spiking is regular. In any case, it doesn't work, and in fact seems
to cause the problem to occur more consistently now.

I haven't yet attempted an RC snubber - not totally convinced on what
I would be achieving with that and would prefer some specific advice
on the problem first.

If you can help, I would be most grateful.

Regards,

Richard.

Obviously, the large transient voltages from the relay power side are
getting back into the PIC. A good starting point would be a complete
schematic of the whole system, so that we can discuss coupling
mechanisms and fixes. Can you draw something up , get it converted to
GIF format and post to alt.binaries.schematics.electronic? Or email
it to me and I will post it for you.
--
John Popelish

 

[John Popelish]

Richard Swinfen wrote:

Good idea John - only my 'schematics' are more layout diagrams in my own
'style'! Moreover the controller device does a lot more as a slave device
for a multi-node networked home automation system, I wouldn't want to
complicate the question. I do believe, perhaps naively, that a simplified
device would have the same problem. I could knock up something that
represents the area of the circuit that deals with this, but in an attempt
to expedite an answer, I'll try to describe the rest of the relevant
circuit. The PIC, 2803 and relays are powered by a 7805 with recommended
caps on both sides (1uf & .33uf) itself powered by a 1.2A 12v regulated
supply. Another possibly relevant point is that the PIC determines
zero-crossing using a dirty-but-working 10MOhm direct live feed to RB0. I'll
check removing this doesn't fix it.

Does that help?

The networked connections can complicate things, since they are all
antennas that can feed spikes to the PIC. I big no-no is having any
connections to the PIC go outside the supply rails, even briefly. The
specs for the PIC are damage specs, not malfunction specs. You might
try adding an additional 1 Mohm resistor between the 10 Mohm resistor
and PIC, with a 4.7 volt zener at the junction of the resistors ot the
0v rail. This will eliminate any positive input voltage above 4.7
volts, and will limit the negative current ot a few tenths of a volt
across 1 Mohm.

What sort of circuit do you have connected to the reset pin of the
PIC?
Are any unused pins programmed to be inputs?

What sort of relay driver circuit connects the PIC to the lamp relays?

--
John Popelish

 

[Richard Swinfen]

"John Popelish" <jpopelish@rica.net> wrote in message
news:41AFBF00.E3A8A86C@rica.net...

Richard Swinfen wrote:

Good idea John - only my 'schematics' are more layout diagrams in my own
'style'! Moreover the controller device does a lot more as a slave
device
for a multi-node networked home automation system, I wouldn't want to
complicate the question. I do believe, perhaps naively, that a
simplified
device would have the same problem. I could knock up something that
represents the area of the circuit that deals with this, but in an
attempt
to expedite an answer, I'll try to describe the rest of the relevant
circuit. The PIC, 2803 and relays are powered by a 7805 with recommended
caps on both sides (1uf & .33uf) itself powered by a 1.2A 12v regulated
supply. Another possibly relevant point is that the PIC determines
zero-crossing using a dirty-but-working 10MOhm direct live feed to RB0.
I'll
check removing this doesn't fix it.

Does that help?

The networked connections can complicate things, since they are all
antennas that can feed spikes to the PIC. I big no-no is having any
connections to the PIC go outside the supply rails, even briefly. The
specs for the PIC are damage specs, not malfunction specs. You might
try adding an additional 1 Mohm resistor between the 10 Mohm resistor
and PIC, with a 4.7 volt zener at the junction of the resistors ot the
0v rail. This will eliminate any positive input voltage above 4.7
volts, and will limit the negative current ot a few tenths of a volt
across 1 Mohm.

What sort of circuit do you have connected to the reset pin of the
PIC?
Are any unused pins programmed to be inputs?

What sort of relay driver circuit connects the PIC to the lamp relays?

--
John Popelish

John,

The network connections are via a MAX485 which itself has a common ground
with the other nodes. I could try eliminating all the comms connections as
well, I guess. I have eliminated the live connection via the 10 Mohm with
the same results (grounding the input). All other inputs are in use - in
fact every pin is in use.

I like the idea of the 1 Mohm + zener circuit you suggest - I'll add that at
some point! I've used zeners for DC transient elimination on other
inductive loads that I drive with the 12v line.

I'm glad you brought up the reset pin. I should have mentioned that. MCLR
has a weak pull up (10k) to 5v. I know internally it is filtered for short
low pulses, but perhaps there is a weakness here? Perhaps a zener here to
prevent inverted transients? Maybe it would still be brought low and cause
a reset anyway?

The relay driver circuit is very simple. Just direct PIC->2803 connections
with direct delay driving from the 2803. Perhaps some filtering/protection
here?

You'll gather I have limited knowledge of filtering and protection, so your
advice is gratefully received.

Regards,

Richard.

 

[Richard Swinfen]

"CBarn24050" <cbarn24050@aol.com> wrote in message
news:20041203025736.11588.00001546@mb-m29.aol.com...

Hi, several things spring to mind. 1. decoupling capacitors. 2 RFI filter
on
the ac supply to the psu. 3. DC filter to the pic vcc pin, I use 100r
+01uf cap
(low pass) twice, that seems to kill everything. 4. Are you using the RB0
input
in interrupt mode? don't. 5. write your software so that it recovers from
crashes, ie use the brown out interrupt.

1. Can you elaborate?
2. I could battery power the device first to determine if this would help I
guess
3. Can you explain exactly how you do this - a snubber across the vcc and
gnd? What do you mean by 'twice'?
4. Yes I do - but this is a rather handy feature of the chip - it is
actually the only interrupt I use as the device is driven by zero-crossing
logic to ensure timing integrity for triac control. All other interrupts
are masked and polled instead.
5. This seems to be a general comment - but a good one!

I've stumbled across a discussion on EDABoard
(http://www.edaboard.com/ftopic74861.html). There seems to be some debate
on the methods for snubbing the pic power lines. As I already filter the
7805 regulator with a 1uF cap (as recommended somewhere), not sure what the
effect of adding additional filtering with a smaller cap to the +5v supply
is going to have? The 7805 sheet even suggests <100nF "might cause
instability".

Also, on this thread there is a suggestion of increasing the current to MCLR
to prevent a spike from bringing it low - seems a suspect solution and think
I'd rather prevent the spike than just handle it.

Don't get me wrong - I'm not adversed to trying things, just this device is
used daily and uninstalling it is time consuming and I want to minimise the
downtime!

Thanks,

Richard.

 

[CBarn24050]

Subject: Re: PIC resets when powering fluorescent lights
From: "Richard Swinfen" Richard_rumoveme_@Swinfen.com
Date: 03/12/2004 13:29 GMT Standard Time
Message-id: <coppng$ao4$1@newsg3.svr.pol.co.uk


. 1. decoupling capacitors.

Its very important to place decoupling capacitors at the supply pins of the PIC
0.1uF should do.

2. I could battery power the device first to determine if this would help I
You could but whats the point? Your flourescent tube needs 1500v to strike the

remnant of that spike goes everywhere, line filters are just about mandatory.

3. Can you explain exactly how you do this - a snubber across the vcc and
gnd?

No, it's not a snubber it's a low pass filter. A resistor in series with the 5v
line followed by a cap to 0v. The bigger the resistor the better it works but
you get a bigger volt drop on the line so you need to minimise the PIC vcc
current. You can put it upsteam of the regulator but it's not as effective.

What do you mean by 'twice'?

2 in series.

4. Yes I do - but this is a rather handy feature of the chip - it is
actually the only interrupt I use as the device is driven by zero-crossing
logic to ensure timing integrity for triac control. All other interrupts
are masked and polled instead.

It seems like a good way to do it but it isn't. Firstly you will get multiple
interrupts at the crossing point, any stay noise pulse will cause you false
triggering and the there is the mains "drop out" problem to contend with. A
better way is to have the timer interrupt running at say 50uS, poll the
crossing point input each time through the loop, and do some "debouncing" on
the result to give you a reliable trigger. You can also generate your time
delays and trigger pule steams at the same time.

 

[Richard Swinfen]

. 1. decoupling capacitors.

Its very important to place decoupling capacitors at the supply pins of
the PIC
0.1uF should do.

Ok - will do.

2. I could battery power the device first to determine if this would help
I
You could but whats the point? Your flourescent tube needs 1500v to strike
the
remnant of that spike goes everywhere, line filters are just about
mandatory.

Point taken - and I should. But could this actually be causing a reset?

3. Can you explain exactly how you do this - a snubber across the vcc and
gnd?

No, it's not a snubber it's a low pass filter. A resistor in series with
the 5v
line followed by a cap to 0v. The bigger the resistor the better it works
but
you get a bigger volt drop on the line so you need to minimise the PIC vcc
current. You can put it upsteam of the regulator but it's not as
effective.

Ok - sounds good - will definitely try that. Just need to determine the PIC
current req's - shouldn't be too high.

What do you mean by 'twice'?

2 in series.

Ok - guess you have good reasons for doing this.

4. Yes I do - but this is a rather handy feature of the chip - it is
actually the only interrupt I use as the device is driven by zero-crossing
logic to ensure timing integrity for triac control. All other interrupts
are masked and polled instead.

It seems like a good way to do it but it isn't. Firstly you will get
multiple
interrupts at the crossing point, any stay noise pulse will cause you
false
triggering and the there is the mains "drop out" problem to contend with.
A
better way is to have the timer interrupt running at say 50uS, poll the
crossing point input each time through the loop, and do some "debouncing"
on
the result to give you a reliable trigger. You can also generate your time
delays and trigger pule steams at the same time.

Don't get multiple interrupts - logic debounces through required processing
during interrupt (admittedly <50us worth). I also run a timer interrupt (so
I lied about RB0 being the only one!), which is used for the triac timing -
seems pretty reliable - despite heavy processing of comms and other logic,
timing of triacs' output remains highly stable.

What do think about issues around MCLR? Increasing current or more filtering
on that pin. Or are you meaning to include MCLR when you refer to 'supply
lines' above?

Very grateful for your help.

Cheers,

Richard

 

[John Popelish]

Richard Swinfen wrote:

John,

The network connections are via a MAX485 which itself has a common ground
with the other nodes. I could try eliminating all the comms connections as
well, I guess. I have eliminated the live connection via the 10 Mohm with
the same results (grounding the input). All other inputs are in use - in
fact every pin is in use.

I like the idea of the 1 Mohm + zener circuit you suggest - I'll add that at
some point! I've used zeners for DC transient elimination on other
inductive loads that I drive with the 12v line.

I'm glad you brought up the reset pin. I should have mentioned that. MCLR
has a weak pull up (10k) to 5v. I know internally it is filtered for short
low pulses, but perhaps there is a weakness here? Perhaps a zener here to
prevent inverted transients? Maybe it would still be brought low and cause
a reset anyway?

I would add a small capacitor between MCLR and the ground pin of the
PIC.

The relay driver circuit is very simple. Just direct PIC->2803 connections
with direct delay driving from the 2803. Perhaps some filtering/protection
here?

Id this what you are referring to as 2803?
http://www.allegromicro.com/sf/2801/

This sort of driver represents a connection between a relay driving
supply and the logic supply. You should arrange the two supply
connections at pin 9 to minimize the length of conductor that carries
both logic signal returns and relay currents. When there are more
than one of these chips, a true single point connection between the
two supply negatives is not possible, but you can still keep this bus
as short and heavy as possible (connecting all the pin 9s) and then
tie both supplies to that bus.

What is pin 10 connected to? I am assuming it is the relay supply
positive. There should be a bypass capacitor between the relay supply
positive and the pin 9 bus, one at each 2803 (pin 9 to pin 10) with
the relay power line going out from that point, instead of from
somewhere else along the positive supply line. That way, when a
switch pens, the current through the loop out to the relay detours
from the switch to pin 9 path to the bypass capacitor to pin 9 path.
So the transient is contained outside the logic supply.

Relay contacts bouncing while driving inductive loads like the lamp
ballasts will generate lots of sparks and high voltage spikes and also
take little bites out of the contact material. An RC snubber across
the contacts or even an MOV will absorb lots of that energy.

You'll gather I have limited knowledge of filtering and protection, so your
advice is gratefully received.

Regards,

Richard.

--
John Popelish

 

[Richard Swinfen]

I'm glad you brought up the reset pin. I should have mentioned that.
MCLR
has a weak pull up (10k) to 5v. I know internally it is filtered for
short
low pulses, but perhaps there is a weakness here? Perhaps a zener here
to
prevent inverted transients? Maybe it would still be brought low and
cause
a reset anyway?

I would add a small capacitor between MCLR and the ground pin of the
PIC.

Yep - sounds like a good idea - and actually what Microchip suggest (now
that I've read the datasheet properly!). Just got to extricate the board
from the application and modify - argh.

The relay driver circuit is very simple. Just direct PIC->2803
connections
with direct delay driving from the 2803. Perhaps some
filtering/protection
here?

Id this what you are referring to as 2803?
http://www.allegromicro.com/sf/2801/

Yep - although in this application, I only have one array - suffices for the
number of relays that need powered.

This sort of driver represents a connection between a relay driving
supply and the logic supply. You should arrange the two supply
connections at pin 9 to minimize the length of conductor that carries
both logic signal returns and relay currents. When there are more
than one of these chips, a true single point connection between the
two supply negatives is not possible, but you can still keep this bus
as short and heavy as possible (connecting all the pin 9s) and then
tie both supplies to that bus.

I do try to keep the relay power lines fairly independent (in wiring terms)
from the logic supply - however, they do both use the regulated +5v, so
there are limits!

What is pin 10 connected to? I am assuming it is the relay supply
positive. There should be a bypass capacitor between the relay supply
positive and the pin 9 bus, one at each 2803 (pin 9 to pin 10) with
the relay power line going out from that point, instead of from
somewhere else along the positive supply line. That way, when a
switch pens, the current through the loop out to the relay detours
from the switch to pin 9 path to the bypass capacitor to pin 9 path.
So the transient is contained outside the logic supply.

Pin 10 is not directly the supply +ve to the relays, but as above, they are
both derived from the 7805. More bypass capacitors sound like a good idea
though. What size for this purpose? Not sure what formula is proposed
here. 0.1uF again?

Relay contacts bouncing while driving inductive loads like the lamp
ballasts will generate lots of sparks and high voltage spikes and also
take little bites out of the contact material. An RC snubber across
the contacts or even an MOV will absorb lots of that energy.

Yes - sparking is a concern but the MOV's in place should help. Clearly more
filtering to protect the logic line is the way to go, though. I suspect,
though have yet to prove, the main problem is the weak pull-up on MCLR and
will address that with a lower resister and addition of the cap.

John Popelish

Thanks again,

Richard.