Three phase phase-angle control problems

[TOLiS_GR]

Hello all, I recently completed a stand-alone phase angle controller to control a heating element. My circuit is a "textbook" circuit using an LTV814 for the zero cross detection and a MOC3021 to control a TRIAC. Mains Line connection goes to the TRIAC, then to the resistive element and from the element to Neutral.

For my tests I used an atmel mcu that does the required delaying to control the heating element. While connected to the load and using a panel mounted ammeter as well as an ammeter clamp the circuit seems to work flawlessly and with proper linearity.

When I try replicating the circuit for a three-phase heating element I get some erratic behavior while in the lower settings. My ammeters seem to jump around and on some occasions I can see my lamps flickering! I built three identical circuits, I use a transformer on one of the phases to power the logic part of the circuits and otherwise they are standalone circuits. Each Line (L1-L2-L3) goes to a separate TRIAC, there to a different heating element, and then all of them connect to Neutral. I also use a resistor on the TRIAC to avoid false triggering.

Now, my question is.. am I missing something serious? Unfortunately I don't have a multi-channel oscilloscope or a low-voltage three-phase power supply, but when I tested my circuits on the bench to verify them as single phase (before moving to high voltage) they worked fine.

I've tried all three circuits with all different L1-L2-L3 connections to them and I always get that strange erratic behavior. When using them one by one with any of the mains line they work perfect.

Any ideas on what I'm doing wrong?

Thanks in advance

[Andy Watson]

but when I tested my circuits on the bench to verify them as single phase (before moving to high voltage) they worked fine.

What voltage did you use to test them on the bench? Have you accounted for the fact that the LTV814, with 100k series resistor, will take a significant amount of voltage before they start to operate? I.e. they are not detecting zero crossing, but some, significant time later. This extra delay could easily push a late trigger event into the early part of the next cycle.

[TOLiS_GR]

On the bench I verified the working principle using mains voltage and a single channel oscilloscope. First tests were with a 60Watt / 230V lamp.

Next test was with 5KW heating element, using a panel ammeter and a clamp ammeter (used both to be sure of the results as the panel ammeter is too sensitive and fast, however both were right).

Now, according to what my processor "sees", the zero cross pulse is from 900 to 1200uS. True zero cross should be in the middle if both LTV814 leds and my resistors etc. were 100% identical, so taking that in account I decided to control my power from 5 to 95%.

During the higher voltage / load tests (3x 5KW resistive heating elements), all three control modules with any element and any phase work flawlessly individually (e.g. module one, element one, phase one, the rest OFF and every possible variation like module three, element two, phase 3). However, when all modules are on, and all elements are connected (each to a different phase), they don't work properly. Not in the lowest settings or the high ones, they work correctly there. It's somewhere in the 20-40% things get weird. One phase gets fully open, the other bounces.

I'll try (tomorrow probably) using a snubber r/c for the MOC3021 as there may be some inductance from the elements (they're located in the same device, one besides the other).

Next test would be to completely isolate each module (they now share the 5V power supply, which however seems to be clean and without oscillations etc).

Any ideas? Maybe I'm missing something obvious concerning three-phase control? I've read every single application note and datasheet I could find but I may have missed something..

[Andy Watson]

How good is the neutral? Is it being "pulled" away from from its assumed neutral position?
Are the supplies noisy? I think I would use the LTV814 to synchronise a clock on the processor - effectively making a phase-locked loop - this would give you some noise immunity.

[tautech]

On the bench I verified the working principle using mains voltage and a single channel oscilloscope. First tests were with a 60Watt / 230V lamp.

Next test was with 5KW heating element, using a panel ammeter and a clamp ammeter (used both to be sure of the results as the panel ammeter is too sensitive and fast, however both were right).

Now, according to what my processor "sees", the zero cross pulse is from 900 to 1200uS. True zero cross should be in the middle if both LTV814 leds and my resistors etc. were 100% identical, so taking that in account I decided to control my power from 5 to 95%.

During the higher voltage / load tests (3x 5KW resistive heating elements), all three control modules with any element and any phase work flawlessly individually (e.g. module one, element one, phase one, the rest OFF and every possible variation like module three, element two, phase 3). However, when all modules are on, and all elements are connected (each to a different phase), they don't work properly. Not in the lowest settings or the high ones, they work correctly there. It's somewhere in the 20-40% things get weird. One phase gets fully open, the other bounces.

I'll try (tomorrow probably) using a snubber r/c for the MOC3021 as there may be some inductance from the elements (they're located in the same device, one besides the other).

Next test would be to completely isolate each module (they now share the 5V power supply, which however seems to be clean and without oscillations etc).

Any ideas? Maybe I'm missing something obvious concerning three-phase control? I've read every single application note and datasheet I could find but I may have missed something..

You say you've got 3 phase control problems but that's not quite correct as 3 phase use is normally between each of the 3 phases and not between phases and neutral.

You've stated you're using a single neutral return for the 3 x 230VAC supplies, I'd suspect that wiring is not heavy enough (low impedance) and if at your control circuits there's an elevated neutral voltage (should be close to zero) along with the switching noise at other than full power setting there'll be corruption of your Triac control.

[TOLiS_GR]

Andy, I'm using the LTV to measure the zero cross pulse and thus calculate the theoritical "real" zero point, adding some timing "space" for safety. I'm also using it to calibrate my processor's speed using the 50Hz mains frequency - this ensures that my timing is pretty accurate. Indeed it is some kind of pll.

Now, what both you and tautech say about my neutral.. If my neutral wasn't good enough, wouldn't the problem be all-around? e.g. it's not happening while the control is at 50%, switching the triac at the highest peak where I would get the worst switching noise etc. I'll check tomorrow I guess (the three-phase testing isn't done here as I don't have the facilities for such loads here)

Tautech, while searching I found that one of the three phase connection "scheme" is the 4wire Y connection, three loads one on each phase with a single neutral line. If my connection is not called three-phase any suggestions on how I could search (google, e.g.) for better results?

Also, any ideas on how to improve my wiring are welcome. Would you suggest removing the neutral completely? Delta connection? Is phase-angle control feasible using Delta connection and three stand-alone controllers?

Thank you both for helping!

[Ian.M]

Its likely that the switching spikes from one controller are causing false zero crossing detection for the controller that is on a phase lagging the first one.  That in turn will disturb the next etc. 

[tautech]

Tautech, while searching I found that one of the three phase connection "scheme" is the 4wire Y connection, three loads one on each phase with a single neutral line. If my connection is not called three-phase any suggestions on how I could search (google, e.g.) for better results?

Also, any ideas on how to improve my wiring are welcome. Would you suggest removing the neutral completely? Delta connection? Is phase-angle control feasible using Delta connection and three stand-alone controllers?

Thank you both for helping!

For your application WYE with a neutral is probably the best.
Here's some reading to help understand what's the problem.
http://static.schneider-electric.us/docs/electrical%20distribution/low%20voltage%20transformers/harmonic%20mitigating/0104ed9501r896.pdf
http://www.saunderselectric.com/pdf/a-primer-on-power-harmonics.pdf

While you think you've implemented a balanced load system, because it's Triac controlled over 3 phases the switching and the resultant unbalanced return paths are f***ing the job up for you.

[TOLiS_GR]

So, from what I understand, I may have an improvement if I user better neutral wiring but the main reason is that the unbalanced load messes up my controllers!

I've seen here http://www.nuwaveproducts.com/ssrman-1p.htm that their product can be connected either like mine or in an inside-delta. They also use a separate transformer for each controller. Apart from providing the zero cross detection do you think this can also provide enough isolation between loads? Or would it be better to try and noise-proof my zero cross detector and triac as my "logic" side seems to be fine in the low voltage side and when the loads mess up the return paths it's the high voltage side that gets weird.

Here I see that they have a few different delta connecting diagrams http://ccipower.com/support/resources/technical-reference/load-configurations and not any Y connection.

I guess I'll have to convert my circuits for delta connection and see what happens, this would also eliminate the neutral line.

Thanks again for the ideas/suggestions, it seems that most of the stuff I can find online are either about three-phase inverters or motor controllers.

[Circlotron]

Not in the lowest settings or the high ones, they work correctly there. It's somewhere in the 20-40% things get weird.

The greatest RF interference is generated when the triac fires at the top of the AC voltage waveform. At the lowest and highest setting it will be a lot quieter. Put an AM radio nearby and have a listen. Maybe you could try put a powdered iron toroid in series with the loads right near each triac. This will slow down the load current jumping from zero to maximum in ~1uS as it is now. There are toroids available specifically for this purpose.

[TOLiS_GR]

I'll try that too!! Thanks! Going to get separate transformers now to try isolating each module and remove the zero cross from the line, see if that and/or an inside delta connection helps.

I'll also try adding a snubber to the triacs just in case 

any suggestions on the toroids? I see they're cheap enough but any "directions" on what/how to choose?

[Apollyon25_]

If its just for a heating element, why don't you simply rectify 3ph then PWM the dc output to the load?
If its for proper three phase control (for VSD etc) then you'd be better to look at proper 3ph vector control i.e. buy a cheap VSD or, from memory, ST have a 3ph motor controller eval kit...

Triac's went out of fashion for this kinda stuff over 20 years ago - Because they suck.

[Circlotron]

Or depending on the thermal time constant of the heating elements you could just switch them on at the voltage zero crossing for an even number of half-cycles periodically.

[Siwastaja]

If its just for a heating element, why don't you simply rectify 3ph then PWM the dc output to the load?

Simply? Because of extra losses and extra complexity required to unnecessarily rectify the AC, whereas triacs work with AC directly?

If its for proper three phase control (for VSD etc) then you'd be better to look at proper 3ph vector control i.e. buy a cheap VSD or, from memory, ST have a 3ph motor controller eval kit...

Motor drive to drive heating elements? Really?

Triac's went out of fashion for this kinda stuff over 20 years ago - Because they suck.

Um, no? Triacs are used for exactly this all the time. Sure, they are not used in VFDs anymore.

And I'm rather sure that the false triggering control problem is in the design (maybe layout-related EMI) and would most probably affect other types of semiconductor switches, too; the cause needs to be rectified. Redesigning the thing using unnecessary, wrong and overly complex structures that are not designed for the task will most likely make the problem worse; at least that won't solve it.

[TOLiS_GR]

TRIACs and discrete back-to-back SCRs are still well used for heating elements control.

As far as I researched my options are the following

- phase-angle control, which produces too much noise, however I may be able to fix that in software (e.g. ignore false triggering based on previous period length)
- burst control, which seems to be the recommended control method for nichrome heating elements (which I didn't know I had 'till yesterday)
- multiple heating elements and combination of them (switched just on/off at zero cross depending on the needed temperature)

what was also unknown to me was the option to use a triangle/delta connection without neutral. Now I'm preparing my circuits for inside delta and 400V, but I'll need to change the resistors too so it will take some time.

My ONLY concern know - since I/we/you people here helped me realize it's a noise problem due to imbalanced circuit, is the legalities. I had NO problems on other devices near/around my test-setup using phase-angle control on 15KW load apart from the time the false triggering happens. If I manage to make my triggering system immune I think it would be ok. On the other hand, while burst control doesn't produce any noise, I've been told that lights on the same line might have flickering, which is legally ok, but customer (parents, wife, people leaving on the same home ) unacceptable!

[tautech]

TRIACs and discrete back-to-back SCRs are still well used for heating elements control.

As far as I researched my options are the following

- phase-angle control, which produces too much noise, however I may be able to fix that in software (e.g. ignore false triggering based on previous period length)
- burst control, which seems to be the recommended control method for nichrome heating elements (which I didn't know I had 'till yesterday)
- multiple heating elements and combination of them (switched just on/off at zero cross depending on the needed temperature)

what was also unknown to me was the option to use a triangle/delta connection without neutral. Now I'm preparing my circuits for inside delta and 400V, but I'll need to change the resistors too so it will take some time.

My ONLY concern know - since I/we/you people here helped me realize it's a noise problem due to imbalanced circuit, is the legalities. I had NO problems on other devices near/around my test-setup using phase-angle control on 15KW load apart from the time the false triggering happens. If I manage to make my triggering system immune I think it would be ok. On the other hand, while burst control doesn't produce any noise, I've been told that lights on the same line might have flickering, which is legally ok, but customer (parents, wife, people leaving on the same home ) unacceptable!

The normal configuration for 2 SCRs is in parallel with polarity reversed WRT each other. It has the advantage or halving the duty cycle of each device allowing smaller devices to be selected.

Personally I'd imagine investigation of snubbers should be explored to solve your problems.

[TOLiS_GR]

I tried using snubbers on the TRIACs with no effect. I also tried connecting three analog dimmers and they are ok, I guess due to their working principle they are immune to the switching noise.

From what I understand now the problem is false triggering on the zero cross detector due to noise. Maybe if I remove the Neutral and change my connection scheme it will improve, but I don't have the means to take RF measurements.

Anyway, I just finished adding code to try and filter / ignore false zero cross signals, and now working on a code that does burst fire control. Gonna test them in a couple of days and see the outcome.

[TOLiS_GR]

So people.. removing the Neutral did nothing to solve my problem and I decided making a simple on/off control for 5 of my heating elements and a phase-angle control for 1 in order to provide fine trimming.

However a friend of mine showed my 2 devices that seem to work really fine (according to the amperometers they load each phase nearly identically). Further inspection showed zero-cross optocouplers (which makes me think they use some kind of burst control). What confuses me now is the wiring and "theory" of control. Both have six elements, one uses six triacs the other uses seven.

Anyone knows something helpful?

[sarepairman2]

why would a hobbiest ever need more then 100 miliamps??

[Andy Watson]

If you examine the first circuit it is simply a delta connected configuration of three heaters with a duplicated delta connection of three heaters. The second circuit is more interesting. By appropriate choice of switching sequence the heaters can appear as two lots of delta connected heaters (as in the first diagram), or as one delta connected heater with resistance = 2R. Very cunning!

But neither configuration solves your triggering problem - you need a robust timebase.

[tautech]

Further inspection of your schematic in the OP shows you have NO local decoupling on the IC's.
Surely you've not overlooked something that basic?

[TOLiS_GR]

tautech, decoupling and protection stuff was omitted to make the schematic simpler

sarepairman2, cause the hobbyist took it personally to control properly a 15KW three-phase load, and why not actually? and how could someone learn without experimenting, testing, asking?

Andy thanks for the info. The first circuit gave me that idea too, the second got me confused but what you say makes things a little clearer! I've studied the circuits of both these devices and they don't have anything more/less than mine (nothing more filter-wise etc). Main difference is in the optocouplers, MOC3083. I've also noticed that the main cpu doesn't have zero-cross-detection in all phases (the main board gets only two lines), would that help understanding the controlling scheme?

[Apollyon25_]

After many years in the power electronics design industry VSD/UPS/PFC etc IGBTs simply replaced SCRs.
Large MW inverters were getting these (SCRs) phased out when I left. That's not to say they don't have uses. Or are not in use.

Siwastaja asked "motor drive to drive a heating control?"
He seemed to miss my question - why are you controlling 3 phases? When you can work with DC - especially for a resistive load. The phases will balance, you wont have control issues... you have loss in the rectifier, sure, I never said you wont, but in the resistive load context who cares?

My suggestion was "if you want a proper 3ph control algorithm(s), in a readily accessible platform then you can use a 3ph motor eval kit" - i.e. add suitable power control elements.

Don't like my suggestions, fine, don't use them. But "contributors" can keep their sarcasm.

[TOLiS_GR]

Well, first things first, I had to start from somewhere.. the product I had to "play" with was using triacs. Each and every single phase dimming circuit (the way I started with high power electronics) was using triacs. My project worked perfectly and the next step was a three phase circuit. That's how it evolved.

I'm not stuck with triacs but since I do have another product that's using triacs and works finely in three phase, why not trying to investigate/understand/educate myself on that?

I'm not keen on using rectified AC to control the resistive load mainly due to losses and the thought of it being sub-optimal (I hope I make myself clear enough, english isn't my mother language). Still it would be another option if I didn't know for a fact that it can be done the triac way.

Now, do you have any examples of using a 3 phase control algorithm for a motor working properly with a resistive load? I mean, do you know for a fact that this works? I will investigate it later, but I'm still trying to figure out how the other system I have works. The wiring is the one posted in my previous post with 7 triacs. It's using zero-cross optoisolators (MOC3083), snubberless triacs, and the controlling mcu has NO zero cross detection at all.

[Apollyon25_]

No problem, its a logical increase in project scope to go from single phase to 3 phase.
There is no ideal solution however as it is for a resistive load, I'd simply rectify, then PWM.

However a lot of the problems you are having may be because your algorithm for control isn't setup for 3-phase.
As I mentioned before, ST (and likely others - Semikron had an awesome development platform back in the day) have a motor control eval kit available, so you could take a look through their source code. Its been a while since I have done power electronics but any half decent algorithm wont care what the load is.

"Proper" 3 phase control is non-trivial at the best of times.

A 15kW resistive load wasn't it? A motor controller dynamic brake (a simple resistive dump, prior to regenerative braking becoming the norm) would limit the DC bus link to below 800Vdc (to protect the bus caps) and this was often only a single low side IGBT driven with PWM - at 800Vdc, 15kW is only ~19A, though this isn't usually a complete bus discharge and only run transiently as part of the speed/torque control loops.

As with all mains (single or 3phase) projects, the voltages/currents involved are all potentially lethal, and any device failure is likely to be catastrophic. So be careful.