Relay chattering problem

[moffy]

It is possible for the 16000uf cap to make a noise especially if it is dried out a bit.

 

[Quantised]

When I tried to listen on each component with a stethoscope I could not define where the noise came from, seems to be everywhere.

Replaced some components and setup all the pots again for correct supply voltages etc and now the noise does not reappear like it did before - so waiting to see when it does - sometimes 1 or 2 days after power up! None of the components replaced were faulty though so the mystery continues.

Thought it might be the SG3524 but tried to see how that worked ....... a week in intensive care after wading through that data sheet!

 

[moffy]

If you get a full conduction cycle when the rest are minimal this can give a VERY LARGE current pulse. The magnetic field can move stuff around.

 

[BobW]

Alec_t briefly mentioned lack of hysteresis in the circuit, and that was also the first thing that I noticed. Since you don't show details of what is connected to the + and - inputs of the comparator, it's impossible to tell exactly how the circuit was intended to handle situations when the input signal(s) vary when close to the trip point. Since you infer that the original circuit was tweaked until it worked, then it may have always been borderline, and some component value has now drifted far enough to cause problems. In my opinion, in this kind of circuit, there needs to be some hysteresis to prevent relay chattering. I would connect a large value resistor between the comparator output and the + input. As a starting value I'd try a resistor that's about 20 times the impedance of the source that's feeding the + input.

 

[Quantised]

moffy - I also think the large pulse creates a big flux that sounds like a dull click, you hear it whenever the pulse is seen on the scope. For sure it is not the relay. A little piezo disc connected to the scope and stuck on various places with bluetack is a great tool I found out.

BobW the full circuit is shown a few posts after the original and it does not have hysteresis but I do not think it is the relay anymore more to do with a bug pulse as the triac fires. When I said the original was tweaked it is because several resistors have piggy backs to change the values or soldered to the underside of the board, done at the factory.

The circuit that controls the error amplifier (no circuit yet as i have not decoded the board) is a 339 with hysteresis and when the system gets down flipping the 339 the hysteresis cuts in and that is where the pulses start. So as you and Alec_t say the hysteresis may be an issue as the unit keeps jumping either side when it sees a pulse.

Maybe either increase it or remove it, as far as I can see it is not needed. Also thinking about putting a cap in the error input line to smooth it a bit but not sure where to put it or what size.

 

[alec_t]

I would expect removing any hysteresis to make things worse rather than better. But who knows .....

 

[Quantised]

The circuit for the voltage control is attached.

When the charger is on the drop across R1 is greater than the preset and pin14 goes low, this in turn pulls pin2 low and adds in some hysteresis to the voltage divider detecting the bat voltage. When the set voltage is reached the current slowly drops until the whole circuit flips to a low voltage charge strategy, about 0.6v below the higher charge level.

It appears that this low level voltage charge rate is around 12.9v so close to the batteries rest voltage and all the pins on the 339 are just hovering around the flip point. When the charger tries to come on at such a low charge rate everything is so delicately balanced I think the charge pulses cause the circuit to flip and hence the sudden large pulse...... a guess of course.

So I was thinking of clipping R45 out of the circuit completely and doing away with the low charge mode as I do not need it

Or to put cap on one side (or the other) of R27 to smooth the error amp input signal, not sure what size or which side as the cap may discharge into the amp?

Any thoughts?

 

[alec_t]

There seem to be a few probs with the circuit:-
The transformer secondary is shorting the battery .
Is pin 5 connected only to the 10u cap?
1k in series with 560k is pointless. Or is a connection point missing?

pin14 goes low, this in turn pulls pin2 low

Not pin 2?

 

[Quantised]

Well spotted the xfrmr short, was a bit lazy and forgot to add the bridge and smoothing cap....as my mouse hand is in plaster at the moment.

Also changed the diagram to show the pin5 and 10u cap connection better as they are connected to the 1K, so that is why I reckon if pin14 is low then it flips pin2 low as well.

Have shown a possible position (in blue circle) of a proposed cap that I think will cut off fc at 3Hz and may smooth pulses from entering the error amp - though my resistor network cap filter theory is well rusty.

OR add the cap and also clip R45 out of the circuit.

Alec - are you a rugby fan? 7's coming soon.

 

[alec_t]

OR add the cap and also clip R45 out of the circuit.

I think the cap would be more effective the other side of R27. Against that, R27 would then no longer protect the comparator input from cap discharge. Removing R45 would shift the voltage trip point significantly. How would you then prevent over-charging?

 

[eTech]

Hi

I don't know if you can try this test....but if you disconnect the charger, and let it run on battery only,
does the problem still exist? Just curious....

eT

 

[Quantised]

Removing R45 would shift the voltage trip point significantly. How would you then prevent over-charging?

That is a good point and I had'nt thought about that because R45 is part of the divider network when in charge mode ie 3K//30K so the 3K resistor should be replaced with a 2.5K. Or I could pull pin5 low all the time but that still leaves a 339 in circuit that may flip.

The 1k pot has been relocated to the case so the charge voltage can be adjusted from 13 up to 16v and we monitor it very closely and turn it down to float level when fully charged so the auto float function is not needed. would get a lot of glue out of the way and simplify.

If we charge to 16v (equalise) the amps never fall low enough to trip the charger into float, so we dial it down to say 13.4v at which point the current is low enough to trip into float mode and that means it falls to 12.8 volts and that is when the pulses start.

As the batteries are our core power source we want to get this one right.

 

[Quantised]

I don't know if you can try this test....but if you disconnect the charger, and let it run on battery only,
does the problem still exist? Just curious....eT

As the problem is caused by spurious large pulses from the triac firing the charger when the AC is off the problem no longer exists.
When it happens one can feel the pulse just by holding the battery cables size 00 or 70mmsq so a chunky power pulse, maybe 50 or 80 amps.

 

[tvtech]

Ahh, I love alec_t. Seriously.

I remember a few Years ago a CRT gave me a whole heap of crap. Certain things did not make sense. I thought I would ask here and lo and behold...alec stuck his neck out for me..not a tvtech but he knows his stuff.

So we discuss what could be wrong. We throw ideas back and forth, We think. We discuss. We reason. We probably had the best discussion ever as in how to resolve the problem. End of the day, the set was scrapped. But that whole discussion was worth Gold to me...

So, there is a good Guy......who knows his stuff and never brags about his knowledge......just walks the talk.

I just had to say that because nobody else will.

You need to know that Alec. You are appreciated and not taken for granted.

Regards,
tvtech

 

[eTech]

When it happens one can feel the pulse just by holding the battery cables size 00 or 70mmsq so a chunky power pulse, maybe 50 or 80 amps.

Back in the day, we used to call that REAL solid state...

we used to have logic made up of literally hundreds of relays on relay racks that ran off of trickle charged batteries. The relays would sometimes chatter intermittently caused by ripple from the charger. We solved this by installing a big DC choke in the DC supply line...

Anyway... Just wondering

eT

 

[eTech]

well said tvtech....and I agree

eT

 

[OlPhart]

OMG, "The main components are a large transformer driven by a triac from AC"
As a newby observer of this thread.... You're running a "large transformer" from a TRIAC ????

Anybody heard of back EMF commutation on the triac?
If it's what I'm smelling, the transformer is "loping" i.e. rolling conductive & not based on phase.
Try running it from a power supply wo/triac & transformer.

Triacs are Notoriously Goofy on inductive loads without Serious back EMF control.
Oh, the sound you hear may be the magnetic "jump" of the transformer getting energy at seemingly random points in time/phase.
Been gone a while... but I'm back for now... G.H.<<<)))

 

[alec_t]

Aw shucks, TV. Now I'm blushing .
Quantised, are you sure the cap at pin 8 of the 339 is 688pF? Strange value. It gives poor filtering of any spikes on the 7mΩ current-sense resistor. Simulation shows 680nF gives better filtering (if the pin 9 trimmer is tweaked accordingly).

 

[Quantised]

tvtech - I have not known him as long as you but he certainly been guiding the research on this project towards a solution, kinda makes you think about what is going on a bit more.

eTech - I think the pulse comes from the triacs and so affects the xfrmr on the AC side first and then the DC side so choke may help but wont eliminate the pulse. It is the actual pulse I want to get rid of. Probably be a big choke that one can wrap 00awg wire around.

OLPhart - Yeah I know its a big xfrmr run off a triac but this thing has been delivering up to 120 amps into 12v batteries for 25 years now. It is simple, has no processors or SMT or multilayer boards and can - eventually - be repaired by NEF's (non electronic folk) like me. This must be fix #10 and all the latest kit one buys lasts about a month or max a year so really love it.

It has snubber circuits on the triac and these maybe outdated now by modern methods but it seems to work ok most of the time, had to swap the MAC223 for BTA26 after it blew a few years ago so maybe the snubber is not suited to the BTA26???? Do you want the triac circuit posted.

Alec - Pretty sure it is 688pf but would it matter if the R45 is clipped as the whole circuit would just float. I could put say a 1u tant in there though as I have one of those in the scrap box.

 

[alec_t]

would it matter if the R45 is clipped

I assume by 'clipped' you mean open-circuited rather than bypassed? I guess try it and see, and it wouldn't hurt to stick a tant in there first, though, to see if that fixes things. But if you're going to be manually monitoring the end-of-bulk-charge anyway, isn't the whole circuit redundant? Or have I missed something?