Product seems to malfunction possibly due to mains noise?

[cupoftea]

Hi,
This product and the problem is as common as the sun rising in the morning,
so we may as well speak of it here.....

Our customer sent us a trailing edge dimmed incandescent lamp.
(made by a northern UK manufacturer).
Its dimmed by customer dimming unit on restaurant wall.
Microcontroller in the lamp uses circuit to detect zero cross
and then switchs the back-to-back AC FETs to dim.

One customer sent 8 units back to the manufacturer, saying that
every 2 mins or so, the light "blips"...kind of quickly goes off
and on again. This, they said happened in the restaurant, but also
in another building when they took them there to see if it "Blipped"
there also.

Anyway, we received the faulty lamps...and We powered these units up and they work fine...no blipping. We put them in thermal
chamber at 60degC and they still didn't blip.

So we have been trying all the usual things to try and put a "disturbance" on the
mains so that we can get them to "blip". Since we cant really solve the fault unless
we first demonstrate it.

Anyway, we can't get it to "blip", but we can get it to minutely "flicker" when we apply mains
to a nearby 1kW mains isolation transformer primary (its secondary is open). When I say "flicker",
I mean, its a ridiculously "hard to notice" flickering instant but, yes, it does do it. Not
as exaggerated as the "blipping" that the customer sent us a video of though.

Anyway, we've been trying to enhance the disturbance that we get with the isolation transformer,
so we tried inrushing into a 220uF , 400V cap repeatedly, but that didn't give any blipping.
Next we will try a CLC filter in the mains, just upstream of the dimmer...and see if switching
that in momentarily instigates noise. (we will simultaneously switch off the mains, then
immediately switch in the CLC, so as to get a ring_up_spike in the mains). We will switch in
the CLC suddenly at mains peak.

Can you think of any other ways to get noise into the mains?

We need to get this thing to "blip" on us. Then we can solve it.

 

[dick_freebird]

Seems to me you should instrument up the mains feed at the
failing site and see if it's really more of a "brown-out" there,
like motor-start (A/C compressor dims my whole house for a
moment...) events that are location-unique.

If mains are dropping below your input spec then you can
pin the tail on that donkey and discuss with them, what their
real needs are - maybe just a tweak to undervoltage limit or
something?

 

[cupoftea]

Thanks, but they dont think its that, as it would affect other equipment in the vicinity, but it does not.

 

[betwixt]

I would agree with Dick_freebird, especially in a restaurant setting where A/C or even large microwave ovens are causing brief drop-outs. It isn't necessarily the voltage drop per se, it could be causing the zero crossing detector to miss a beat.

Brian.

 

[dick_freebird]

Thanks, but they dont think its that, as it would affect other equipment in the vicinity, but it does not.

Why are you taking their word for it, as it's clear they
do not know?

Every appliance has its own conducted susceptibility.
Maybe the cheapo dimmer is just the most twitchy.
Or most prone to stretch a brief event into an observable
one.

 

[hiphei]

I also highly agree for what dick_freebird wrote.. eg. some 15 years ago we had to solve issue with one specific tv type which typically rebooted itself when eg. lights were being turned on in the next room.. there were multiple cases in different types of households and always everything else worked just fine.. it turned out to be really sensitive main microcontroller watchdog that was able to most likely capacitively detect certain events in mains network and it responded by rebooting the firmware..
Also having being doing audio installation repairs and fault finding in the past in restaurants and bars i can confirm they are like a pandoras boxes of earth currents..

 

[cupoftea]

Yes, thanks, "earth currents"...yes...and it looks like the use of common mode filters is the only way to combat these things.
Y caps to enclosure as always.
Not Y capping an enclosure is suicide IMHO.
Obviously directly connecting to enclosure is ok but not usually desirable.

We who throw away our Y caps...will burn in the fires of hell.

 

[hiphei]

Another way is to combat the electrical network at the customers site, but in restaurants typically they have been built and rebuilt many times on top of previous install and no one has clear total view of the cabling routes, earth routes, ground loops, etc. and it's usually very hard to find someone capable of sorting that out..

 

[danadakk]

Suspicion of pin(s) on board rebooting processor....? Terminated in relatively
high Z so susceptible to pickup ? Like a reset pin with a cap on it (that looks
open at line freq) and a high valued R that likes to generate offsets with
leakage and reduce noise margin at pin.......

Regards, Dana.

 

[dick_freebird]

It's true that a RC reset pin will creep toward loss of noise margin as the capacitor develops any leakage, and the R sets the bar. What follows a false-trigger can take a while to get done.

 

[cupoftea]

We have now been hitting our equipment with transients from a transient generator...so please may i rekindle the thread from there?.....i know it is "not goood" to start new threads on the same subby.

Anyway.....

Hi,
We were doing Burst testing on an offline SMPS with and on some auxiliary circuitry that's with it.
The kit comprised of a 24W offline flyback SMPS ("24v PSU") in an earthed metal enclosure. The 24v from
this goes through a 30cm cable to another earthed metal "SWITCH enclosure".

The "switch enclosure"
comprises semiconductor load switchs in the AC mains line.
So the "switch enclosure" recieves cables from...
1...a control unit, which commands the fet switchs on and off
2....mains L,N & E
3...24V from the "24v psu" as described.

The "switch enclosure" has a very small offline SMPS in it, which merely provides bias power
to the AC switchs' gate drives, and also to the mains zero cross cctry.

The "switch enclosure" has no common mode choke or y caps whatsoever.
The only mains filter it has is the 10uF electrolytic cap for the wee flyback
which is as described.

The "24v psu" has no y cap connected from HVDC+ to secondary ground. (which as you
know, it really should do, as this is academically correct)
The "24v psu" has a y cap from
a)...sec ground to earth
b)...prim HVDC- to earth.

This setup malfunctions temporarily every time we hit it with a 2kv mains burst transient.
The transient burst is at 1kHz, and it as 90 degrees. (from a schaffner BEST kit")
We do many tests with the transient coupled to L,N,E,L&E,N&E,L&N&E,L&N.

When 2n2 y caps are added L-E and N-E at the mains input to the "switch enclosure", then there
are less failures from being hit with the transient.....but it still fails when the transient
is coupled to L or L&N.
....So therefore i added a 68n X2 capacitor to the mains input of the "switch enclosure"....but that
makes it worse!!!...more susceptible to malfunction after transient.

Also, adding a y cap from HVDC+ to sec gnd of the "24v PSU" doesn't improve things.
However, disconnecting the y caps from the "24v psu", and adding them from L-E and N-E instead, and also adding another y cap
from HVDC+ to sec gnd makes it perform without malfunction.

So, how do you tell , from the failure when the transient is applied...whether its a diff mode or common mode problem?
(eg supposing it fails after an "L&N" coupled transient, but not after an L&N&E coupled transient)
I suspect its always a common mode problem no matter what. But what do you think?