Why did this TRIAC dimmer burn?

Question

I have built an ATtiny controlled 230 VAC dimmer which has unfortunately burned. I would like to fix the design so it is reliable.

Files – Fritzing Schema · Arduino Code

The dimmer was used to dim halogen bulbs (28 W each, 5 in parallel, i.e. approximately 150 W in total) and produced smoke after 2 months of usage. The bulbs were still on, but could not be dimmed anymore. Physically, the distance between dimmer and bulbs was approximately 10 m, with another approximately 5 m of wire to the last lamp.

The circuit is using a three-quadrant triac BTA316-600D. (As the triac driver MOC3052 is connected over G and MT2, only quadrants I and III should be used.) I have added an RC snubber with R11, C2 and an inductor L1 because the triac fired once when connecting it to AC. This is based on advice from AN437 and AN-08-06.

Optically, it looks like the snubber resistor burned and melted part of the film capacitor.

I wonder if the circuit failed because L1 does not use an air coil (as AN437 states that “Limit high dI T /dt with a non saturable inductor of a few μH in series with the load”), or because the current over R11 can be much higher than I expected … Or something else, because a second dimmer board has been in use for at least twice as long without any issues so far.

What has to be fixed on this schema so the board works reliably?

Circuit

Board

Burned part

Power measurement

This is the voltage measured over a 0.5 Ω resistor in series with a 28 W halogen bulb, cold start. Peak is around 1.75 V (cold filament) which is reduced to 0.29 V (hot filament). (The absolute values seem wrong because it would not fit for a 28 W lamp, but the relative factor should be okay …)

Improved schema

R4 is now a 470 Ω resistor.

@MarkoBuršič If you have specific suggestions for improvements, please add them. — Simon A. Eugster, Aug 18 '20 at 04:40
@SimonA.Eugster Nicely built. 230 VAC on 'dot board' scaares me a little - but I've done it :-) ||. Learn from the knockers (where they add value) and ignore their rudeness. | Changing the occasional horizontal lines to vertical and parallel ones will largely render the diagram acceptable enough. (C1 D4 R11 ...) The track through U4 should be rerouted. L1 upper terminating on an unrelated track doesn't look good and similar may occasionally cause unintended routing errors. — Russell McMahon, Aug 18 '20 at 07:04
@MarkoBuršič Creative comment welcomed. Uncreative slanging unwelcome. Adding a few pointers would have helped. — Russell McMahon, Aug 18 '20 at 07:05
@jsotola Creative comment welcomed. Uncreative slanging unwelcome. Adding a few pointers would have helped. — Russell McMahon, Aug 18 '20 at 07:05
The dot board does not meet the creepage requirement for 230 Vac. I would not dare to use it for any prolonged period of time. What’s the load? What’s the saturation current of your inductor? — winny, Aug 18 '20 at 08:07
Schematic guidelines: https://electronics.stackexchange.com/questions/28251/rules-and-guidelines-for-drawing-good-schematics — winny, Aug 18 '20 at 08:08
@RussellMcMahon Thanks for the input! I have re-drawn the diagram – is it better now? — Simon A. Eugster, Aug 18 '20 at 15:49
@winny Thanks for the link! Points mentioned should be considered. Regarding creepage, I am aware of that – but how does it make sense that the TRIAC and optocopuler pins have 2.54 mm spacing too? Load is 150 W (a bit more with cold filament). No idea about saturation current, the [data sheet](https://cdn-reichelt.de/documents/datenblatt/B400/L-4RCC.pdf) does not list it? — Simon A. Eugster, Aug 18 '20 at 15:58
Much better! Classic issue, but components often have material class II and cheap dimmer PCB is material class III, at which point you pre-form the legs to give more spacing on the PCB. 6 A rated current, so you are not saturating it. That would have been my best guess, saturated inductor, too high dV/dt and the snubber failing. What’s the voltage rating on the burned resistor? — winny, Aug 18 '20 at 16:16
Wait, cold filament you say. Did it fail during startup by any chance? — winny, Aug 18 '20 at 16:24
Creepage issues aside, you also have no sag/surge protection. Adding MOV's across the line and from each line to PE will help with smaller surges. I see no fuse either... — user187594, Aug 18 '20 at 16:26
@winny Probably [those china resistors](https://www.ebay.com/itm/64-values-1280pcs-1-ohm-10M-ohm-1-4W-Metal-Film-Resistors-Assortment-Kit-/261374861457) rated 250 V … which would not be enough. Regarding cold filament, I have not been there so I cannot tell. Added a halogen lamp measurement I did last year. They have 136 Ω cold resistance. — Simon A. Eugster, Aug 19 '20 at 12:25
@Stiddily Just wondering, is surge protection not something that usually happens on the mains already? Where would you place a fuse, between triac and load? — Simon A. Eugster, Aug 19 '20 at 13:15
With 5 of them in parallel, you are probably saturating the inductor during inrush. Rod inductors have soft saturation characteristics, so it’s hard to say for certain it will produce enough dV/dt to break your triac and/or snubber. — winny, Aug 19 '20 at 14:31
@SimonA.Eugster it should be, but it really depends on your location, both geographically and the building you're running this in. For anything Hi-Rel that's not a chance you should take. Sags are more common, and a dip from 120V to 80V can increase your current draw enough to damage components. Fuses should ideally be inline with both your hot and neutral lines, but just the hot line will suffice. The surges I'm talking about are measured in microseconds. The power grid is a very finicky thing. — user187594, Aug 20 '20 at 14:28
@winny Will replace the inductor with a more suitable one then! Something like [this one](https://www.mouser.ch/ProductDetail/Pulse-Electronics/PM4343332NLT)? — Simon A. Eugster, Aug 21 '20 at 08:48
You can otherwise just cheat and implement a soft start to ramp up the firing angle to limit the peak current. — winny, Aug 21 '20 at 09:05
@winny That is in fact already happening, the dimmer only soft starts/stops with the current code. Thinking about cold filament, that should really do it because the low voltage at the beginning of the soft start would not allow for high currents anyway! — Simon A. Eugster, Aug 21 '20 at 11:23

score 3 · Accepted Answer · answered Aug 18 '20 at 17:57

3

The value of R4 is too high (10k). You need 5mA to turn on the triac. At the switching limit with 10k (+ -50V), the system can oscillate, causing the triac and R11 to overheat. Replacing the R4 will solve the problem. (R4=360...680 Ohm)

answered Aug 18 '20 at 17:57

csabahu

2,156
4
5

I indeed forgot to consider that the triac should also be fired at lower voltages! Regarding the resistor value, assuming 50 °C ambient temperature, the MOC3052 would allow 200 mW power dissipation, and with Vtm = 2.5 V the max current through the MOC terminals would be 80 mA, requiring R4 > 2.8 kΩ at 230 V – or did I get something wrong? – Simon A. Eugster Aug 19 '20 at 12:55
1

Current only flows through the optocoupler, until the triac turns on (< 20us in every 10ms). Overall power is very [small](https://drive.google.com/file/d/1r_kCBwAhzqVL9XeOeOV7uM3pa-qxMga_/view?usp=sharing). – csabahu Aug 19 '20 at 14:14
Ah right, the gate only needs to be charged! (Cannot access the document; can you enable sharing?) – Simon A. Eugster Aug 20 '20 at 11:50
@Simon A. Eugster Maybe the [link](https://drive.google.com/file/d/1r_kCBwAhzqVL9XeOeOV7uM3pa-qxMga_/view?usp=sharing) will be good... – csabahu Aug 20 '20 at 15:28
Thanks, yes, that works! The document looks interesting, will study it. Regarding your answer, how/where is oscillation happening? – Simon A. Eugster Aug 21 '20 at 08:39
1

But with highly inductive load, **or very low RMS current load**, the current passing through the load may be too close to Triac latching current IL and can lead to oscillations at turn-on. (Page 16/27.) – csabahu Aug 21 '20 at 08:50
Ah, I see! The latching current is quite low at 30 mA, so this could happen only when the AC voltage is around 1 V (cold filament with R = 170 Ω each bulb) or 11 V (hot filament) with 5 bulbs in parallel … right? – Simon A. Eugster Aug 21 '20 at 20:48
1

Yes alright. However, the optocoupler is also a triac, meaning the same applies if it is operating at very low current. – csabahu Aug 23 '20 at 09:33
Ok, so the optotriac oscillates because the current over it is too low. Its Ift is 10 mA, with the 10 kΩ R4 the current is below Ift until 100 V. Can this cause oscillation over the snubber without the main triac oscillating? – Simon A. Eugster Aug 25 '20 at 14:56
When the optotriac oscillates, it controls the large triac and the whole system will be unstable. The oscillating triac appears to be fully on for part of the period. – csabahu Aug 25 '20 at 15:41
The sensitivity of the triacs is quite different, so in the same circuit, one sample is good and the other is bad. That's why we use the optotriacs with higher currents. – csabahu Aug 25 '20 at 15:55