Dimming Problem

[P. Short]

As some of you may know, I've been working on a problem that was first noticed during the Beta test phase of Renard-C. I've done more testing, though, and I now think that the problem is more related to the SSR design than to the controller. I have seen it occur with the Renard64, and other people seem to have seen it as well.

The basic problem, which has appeared in various guises, is that the lights appear to flash (or twinkle) at low dimmer levels (especially when ramping from 0% to 100% and back in intensity). It isn't extremely consistent, and seems to affect incandescent lights. It seems to appear mostly with small 4W-7W nightlight bulbs (without the rest of the nightlight controller), and hasn't been reported (as far as I know) with mini-light strings.

At first I had trouble getting the problem to appear here (and when it did, it would go away as soon as I put a scope on it). Eventually I was able to get the problem to appear fairly consistently by using SSRs with just the optos (omitting the triacs), using nightlight bulbs.

Eventually I was able to get a scope picture of what is going on. In the photo below, the top trace is the signal driving the opto-isolator (a narrow pulse towards the end of the AC cycle), and the bottom trace is the current through the light (and opto-coupler). As can be seen, the triac turns on towards the end of one half-cycle, and then (on its own) continues through the next half-cycle. This is enough to cause the light to appear to flash or twinkle.

I do not think that the problem is with the digital signal, I think that the problem is with the triac inside the opto-coupler. My suspicion is that something happens inside the opto when it is fired that late in the cycle. Maybe it doesn't turn on completely, therefore is in a high-dissipation mode, and the resulting local hot-spot causes it to turn on by itself in the next half-cycle.

The test that I am doing (omitting the power triac altogether) is somewhat unusual. However, I think that the problem that I'm seeing with great frequency is the same problem that others were seeing (with lower frequency) using power triacs.

One solution that I have been trying is to simply avoid driving the triac late in the cycle. This has the disadvantage that the light never goes totally off, with significant visible filament glow. Or, if the control pulse is completely suppressed with very low dimmer levels, the lights appear to jump (from completely off to a moderate filament glow).

So I'm open to any suggestions (including ignoring the problem, and telling people to use higher-power lights).

--

Phil

[ErnieHorning]

The datasheet specifies the minimum input current to guarantee that all TRIAC’s will turn on. Maybe it’s possible for a TRIAC to be too sensitive, so backing off on the current would help.

Maybe it’s not enough current and the gate is only partially turning on and fully turning on, on the next cycle.


The only time that I had a similar symptom was when I ran full wave DC through a long Cat5 cable and the cable acted as a capacitor. This raised the zero crossing level and caused false triggering. The LED’s in the optos would turn on and draw just enough current to pull the zero cross down and they would turn off.

Edit: If the problem is in the opto, it will probably be worse at higher temperatures; try the hair drier test.

[Mudsculpter]

Subscribing to thread.

[P. Short]

The hair-drier test does make the problem worse.

--

Phil

[Sean Bowf]

Does a change in the input resistor cause any change in the symptoms?

I think it is an interesting problem...but if it is not occuring with a string of mini-lights, is it really a player? I guess I am wondering, in what application of the Renard/SSR will this cause a problem?

Sean

[P. Short]

That's one thing that I intend to try, see if over-driving the input helps the situation.

Some people are making decorations using lots of individuallly controlled, low-wattage 110 VAC bulbs. For example, someone is building a 3' diameter Star using this type of design.

--

Phil

[Sean Bowf]

That's one thing that I intend to try, see if over-driving the input helps the situation.


I was actually thinking the other direction. I figured by under driving it (within reason) that it would allow it to more easily shut off. This comes to mind, because if I recall correctly, you were unsure about the 750 ohm resistor I use as an input, which would mean to me that you are using a lower value one. I know anything from 5 to 50 mA is acceptable by the datasheet...but am wondering if shooting for 5mA would allow it to turn off more easily, since it would be barely on.

What value input resistor are you using now?

On another note, if heat makes the problem more prevalent, then wouldnt the cold of winter make it less prevalent?

Sean

[P. Short]

Right now I'm using a 680 Ohm resistor with MOC3023 optos. I'm thinking that if I increase the drive to the opto that it is less likely to end up in a partially turned-on state, which might result in extra power dissipation and higher internal temperature. When the triac is totally off, it dissipates very little power (very little current), and it also dissipates relative little power when it is turned (only 1V or so across the terminals). I don't know it this is even possible, but if it ends up in a state where it is partially turned on (passing current, but with a higher than normal voltage drop) then it would dissipate a lot more power (either over the whole device or at local hot-spots within the part).

And yes, I would think that winter temperatures may make the problem less common.

--

Phil

[P. Short]

Tripling the current with a MOC3023 doesn't seem to help much.

--

Phil

[ErnieHorning]

Ok, I’ve been racking my brain on this because it didn’t make sense, but I think it’s starting to.

I’ve been staring at your scope image for awhile and finally noticed the current surge. Then it occurred to me – with the low current and short wires, could the inductance of the coiled filament be causing false triggering? Maybe adding a resistor somewhere around 33 to 51 ohms in series with a .01µ capacitor across the TRIAC would alleviate this problem.

[P. Short]

I've been thinking along those lines as well. The other thing to try is to use the SSR circuit normally intended for use with inductive loads (although that normally goes between the opto and the triac, which isn't applicable to my current test setup).

--

Phil

[A Marchini]

I've been thinking along those lines as well. The other thing to try is to use the SSR circuit normally intended for use with inductive loads (although that normally goes between the opto and the triac, which isn't applicable to my current test setup).

--

Phil

This could be tested by creating a group of normal resistors (in parallel to equate to the wattage dissipation) and see if a purely resistive load doesn't exhibit the problem.

Tony

[P. Short]

The snubber (.01 uF/47 Ohm) seems to improve things. It's hard to be sure, though, since the hair-dryer is a pretty blunt instrument, and the weather has become cooler here in the last day.

--

Phil

[P. Short]

That snubber doesn't seem to be the fix, it was flickering when I went to look at it last night (no hair-dryer this time).

--

Phil

[DynamoBen]

I don't know if this will help with this specific issue but I thought it was worth sharing.

In my previous life I dealt with a lot of SCR dimming. With this style of dimming there is a generally accepted rule that a SCR dimmer needs a minimum load to properly function. If the SCR was loaded too lightly it won't turn off properly or will flicker. What were used to do to test for this was to add a "ghost load" by tying a 100W light bulb to that output. If this fixed the issue then the load on that output needed to be increased. We generally suggested increasing the overall load by adding a power resistor to that output; you could however leave the 100W and place it where it won’t be seen.

I’ve been doing Christmas shows for over a decade which means I often have to control Christmas light via dimmers. I can’t count the number of times I’ve run into the above issue(s) because there wasn’t enough load for the dimmers to work properly. The quick work around in the past has been adding a mechanics trouble light the offending circuit. Last year I place a 100W lamp in a service hallway and ran an extension cord to the power strip the lights were plugged into.

[P. Short]

Thanks, Ben. That is some comfort that this is not a unique problem to this circuit (I'm not being sarcastic). What it really sounds like is that the current capabilities of the SSR circuit must be somewhat matched to the load (not too big, and not too small).

Actually, having that problem with back-to-back SCRs would surprise me a little bit. In that case, each SCR would have a half-cycle to recover before the next time that it could potentially trigger. Or did those designs use some other technique (like full-wave-bridge rectifiers in front of the SCR)?

Right now I am still somewhat fixated on this problem, in part because I still don't understand what is going on in the triacs. I would really like to understand it well enough to know what is behind the 'rule of thumb'. While it probably won't help me to understand the device physics, I'm thinking about building some SSR boards based on back-to-back SCRs and an H11D1 opto-coupler just to experiment a little bit.

--

Phil

[DynamoBen]

I have an email out to a friend of mine; he can speak in detail about the physics and more importantly the exact reason why. I will post once he responds, but if memory serves it’s similar to the reason you need to you a pull up/down resistor on a switch input. This additional resistor tied high or low keep the line from floating and causing random button hits. In the case of an SCR if there isn’t enough current passing through it will “float” thus not changing state properly or even flickering.

In regards to a back to back design I recently stumbled across schematics for a commercial dimmer that used it. I’ve included the link below; they discontinued the line and released all the drawings. The left side of the page has a header “Legacy Products” there you will find a number of different service manuals with schematics and technical info about the product.

http://www.leprecon.com/support.php?id=31

[DynamoBen]

Here is the info from my friend about why this happens:


If you are using a SCR dimmer the issue is that SCRs need a certain minimum amount of current to stay in conduction. Since you are working with a linear (or resistive) load you may end up in a situation where your low load may not get the dimmer to fire each half cycle but most likely you will not have a problem with it not wanting to turn off. What this means is that you will set it to a level and it will either be at that level or flicker off.

Now the larger the SCR (in current rating) the larger the hold current it will require. And of course this changes if you are not running a linear load but that’s a discussion for another day.

A lot of people with low loads like using triacs for dimmers instead of SCRs because their hold current is less but their current ratings are less as well. The other issue is that are much less tolerant to short circuit (ie infinite current) spikes. So do be careful. Shoebox dimmers typically use triacs as they are much lower in cost, handle low wattage loads well but then also usually need to be fused as they will react faster to a short than a magnetic breaker.

[P. Short]

The problem that I was seeing (based on the scope) is the opposite of what your friend was mentioning. I'm seeing spurious triggering of the triac in the opto, which occurs sporatically on the half-cycle after the triac triggers relatively late in the cycle. When set at a particular level (especially when hot) the light flashes on, and does not appear to be flickering off.

Also, I'm a bit confused by the comment about holding current. In this case the trigger is not a brief pulse, but a level that remains on until almost the end of the cycle. So the hold current should not be relevant here.

I'm not sure how to fix the problem that LRNeener is encountering with his setup. The proposals that I have in mind would be to use smaller triacs (like the Z010 series) for the small loads, or to go to a much more complicated scheme using back-to-back SCRs, which should not be susceptible to false-triggering. In either case, this would involve a new SSR board design, which is not very desirable.

--

Phil

[DynamoBen]

Also, I'm a bit confused by the comment about holding current. In this case the trigger is not a brief pulse, but a level that remains on until almost the end of the cycle. So the hold current should not be relevant here.


Like I said before this may not pertain to what you are dealing with but I thought the information might be of value.

[ErnieHorning]

Also, I'm a bit confused by the comment about holding current. In this case the trigger is not a brief pulse, but a level that remains on until almost the end of the cycle. So the hold current should not be relevant here.
Actually it might. You release the gate just before it passes through zero, which means that the holding current applies at the start of the next half cycle. If it turns on by itself, because of lack of current, it will already be on before you expect to turn it on.

[A Marchini]

Also, I'm a bit confused by the comment about holding current. In this case the trigger is not a brief pulse, but a level that remains on until almost the end of the cycle. So the hold current should not be relevant here.
Actually it might. You release the gate just before it passes through zero, which means that the holding current applies at the start of the next half cycle. If it turns on by itself, because of lack of current, it will already be on before you expect to turn it on.

Was there ever a resolution on this? Perhaps off forum? I was just curious if there was a particular fix in mind or if the fix is just to ...load it up.

Tony M.

[P. Short]

So far, load it up. The other thing under consideration is to use a lower power, possibly more sensitive triac for small loads.

--

Phil

[fkostyun]

this also describes problems I am having with my 40a ssrs - they need at least ..3a on them to fire properly. (a 50watt bulb works perfectly)

[xmus]

I might be off track here, I haven't read every line of this thread (just discovered it) but this is why i added the "lockout" feature to the Renard code back in April. If the dimming level is below some preset level, no pulse is sent what-so-ever.

In tradition renard code a pulse is always sent, and the pulse for zero is a pulse that is just slightly before the ZC such that the SSR triggers essentially just before it shuts off, which results in the "nearly off state". But theoretically, no matter how close you get to the ZC (that is, just before the ZC) you are never perfectly absolutely off, the SSR will latch for some (even if very short) amount of time. This tiny amount of current caused single C7 bulbs (small loads) to flash occasionally during dimming. My personal theory is this is due to skew and signal integrity, and is worsened the closer to ZC you try to make the pulse in the off state. That is, the closer you get, the more likely skew/SI will push that signal to the other size of ZC, causing it to latch for a full cycle (full intensity on, instead of full intensity off). I captured wave forms similar to Phil's at some point back on picdimmer.com, i have no idea if they are still there.

Anyway, bottom line, i think in the off state, there should be no pulse sent. And the threshold for the off state should be programmable, this is what is happening in the XMUS code today. If you do this, you avoid flashing of low current channels....

I should note, when a channel is fully loaded, this type of flashing is not even noticed....

To be really redundant, if you send a pulse just before the ZC and call it "OFF", it isn't actually 100% fully off, it is just "really close" to being off, and the closer you get to being "really really close to off", the more chance you have of being totally on sometimes.. If you go the other way, and make it "less off, but almost off" so you don't see the flash occasionally on light loads, then you cant dimm all the way to fully off.


Hope that makes some sense. And I hope I'm not completely out in the weeds on this topic....

Phil did a great job of turning the lockout hack into really good code.
EDIT
here is the code, at the bottom, read the .asm, search for "lockout"
http://www.ritzfam.com/ChristmasMadness/Renardx16SSR/Release/