Phase controller problem 5

[itsmoked]

So. I get called to look at a 'desktop furnace' that has a problem. It's a 'tabletop' model they tell me. I think to myself, "What can the problem be with some Nichrome wire, a PID controller, and a 'solid state relay'"? Silly me.

I get there and I don't even recognize the 7 foot tall 19" rack enclosure, three feet away, is the power supply.

Turns out the thing's 20kW.

It has the PID and that's all of what I expected.

The PID puts out 0-5VDC to a HALMAR phase controller (PC).
The PC dices up the supposed 240V into back-end sine-waves that feed a HUGE transformer that fills the bottom of the rack.

That turns the three phases into two phases of lower voltage DC at up to 200A (100%). These two outputs are crammed thru two banks of 4 elements thru fat braided straps off of bus bar.

Can someone explain why they didn't just run elements at LINE voltage but instead ran it thru this massive transformer?




It just happens that the current is twice the power setting percentage which is nice.

The problem:
When the power setting gets up to about 60% something shifts.
For example if you turn it up to 75% 150A are running thru the elements. After about 5~7 seconds they drop by about 40A. Just spontaneously drops. Simultaneously the two currents to the heaters stop being nearly identical. One might drop to 30% of the other. This manifests itself as two different voltages at the bus bars. One bank of elements may stop glowing red. If you switch the cables feeding the two banks of heaters the problem follows.

Also notable is that the ever-present humming gets an angry tinge to it. The front panel breaker gets hot after a few minutes instead of just above room temperature after hours at 100%.

This all sounds to me like one or more of the six SCRs are cutting out or back. They said they "checked them" but I know for a fact they don't have the chops to do it correctly.

Have you seen this type behavior on a phase controller? Do you recall what the problem was?

I'm going to start with the Gunnar SCR tests but it would be nice to hear any other ideas. I've pulled the PC and have it at my office. I can hook it to 3ph 240 eventually to run live tests.



Keith Cress
kcress -

http://www.flaminsystems.com

 

[ScottyUK]

Could be a gate drive problem - the transformer must introduce an amount of inductance into the circuit which may be preventing the thyristor firing correctly, although it seems peculiar for it to suddenly start. You should be able to safely watch the gate drive signals on the low side of the drive transformer and see if there are any mis-firing problems, and relate them to power circuit behaviour - i.e. look for gate pulses which aren't producing the expected behaviour in the high current path. Check conditions are right for commutation before assuming that a gate hasn't fired correctly.

Other thoughts:
Possibly the element heating chages the LR time constant sufficiently for the circuit behaviour to change?
The angry noise could be when the transformer core starts to saturate if you have firing assymmetry and a net DC component. That should be easy enough to see.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[MacGyverS2000]

The "angry tinge" sounds an awful lot like instant PWM frequency shift into the audio range... I'm with Scotty on checking the gate signal.

Incidentally, I'm currently working on a true desktop-sized heating block... six heater cartridges down-spec'd to 60W each @ 48V, independent PWM drive on each with 6 parallel PIDs. After a few hours of working with them, you can tell what their duty cycle is from the tone (1.2kHz drive), and the scope becomes a bit redundant.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[jraef]

The reason why they put the power through a step-down transformer is to handle the high watts without having to handle the high current in the SCR controller. It's very common on large furnace systems.

To do that, the SCR controller must be Phase Angle controlled. As such, all of the firing elements need to be precise (not that they don't in other schemes, this is just more critical). If you have a failing gate, you might be getting an incomplete fire at a specific phase angle. So it could be a bad gate on an SCR or a bad firing board, or both. They probably only checked the SCR voltage drop across the stacks and didn't see an anomaly. That problem would not show up like that, you need to check the gates. You typically want to see between 5 and 90 ohms across the gates in each SCR; less than 5 or greater than 90 means it's bad. The gate wires should unplug somewhere to get at it. If all the SCR gates are good, then it's the board and it's not likely worth doing a board level repair yourself. I know you could Keith, but an entire replacement board is probably under $500, about 15 minutes of your time!

Halmar used to be part of Robicon. Siemens bought Robicon a few years ago but did NOT buy Halmar. I don't know what happened to them though, I think they may have gone bankrupt. There is a company in Rialto (SoCal) called Nyco Systems that does repairs on Halmar SCR controllers though and last I heard, they still had parts like firing boards available.

"Dear future generations: Please accept our apologies. We were rolling drunk on petroleum."
— Kilgore Trout (via Kurt Vonnegut)
For the best use of Eng-Tips, please click here -> faq731-376

 

[itsmoked]

Thanks for the great info all around. I'll keep you guys in the loop.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[LionelHutz]

I'd guess the gate drive of one SCR is cutting out which leads to a 1/2 wave current in one phase and the circuit breaker probably doesn't like the DC component. You should be able to see the change in the output waveform or current to figure out which phase. Just watch you use an isolated scope probe on the gate leads if you do scope it.

You won't likely find an intermittent SCR. They typically either will work or have failed.

 

[itsmoked]

You won't likely find an intermittent SCR. They typically either will work or have failed.

An excellent point.

You're suggesting that I put a scope across the actual SCR's gate. Ground side to the cathode and tip to the gate. And the other ground to the cathode and tip to the anode and confirm triggering with respect to the trigger pulse. (?)

Fear not. Yes my scope has a completely isolated front end and my probes are CAT2 600V rated.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[ScottyUK]

Keith,

Put a current probe on the thyristor if you have one. Put one scope probe across the thyristor A-K and the other G-K (assuming a 4-channel scope). You need forward voltage and current has to be available to flow forward to achieve successful commutation. You might find that a failed commutation is a symptom of a problem with a thyristor earlier in the firing sequence rather than with the one which fails to commutate.

My opinion is that once you've proven the pulse transformer is ok then you could probably watch the gate drive from the 'safe' side.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[itsmoked]

I don't own a current probe. But! I have numerous current coils for measuring sub 50A currents. Seems I'm just looking for 'current' and not every kink in it. Do you think they'd be fast enough?



Turns out my office has the usual 3ph with the 'funny leg'. Is that going to cause me problems testing this thing?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[waross]

Watch your ground reference. The phase angles from the neutral to the phases on the "funny leg" systems are 90 degrees, 90 degrees and 180 degrees rather than 120 degrees, 120 degrees and 120 degrees.
The line to neutral voltages of the funny neutral/ground systems are 120 Volts. 208 Volts and 120 Volts instead of 120 Volts 120 Volts and 120 Volts.
I hope that your equipment is generating an artificial internal neutral reference rather than using the grid neutral for a neutral reference.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[LionelHutz]

Keith;

Just checking the 60hz current will show you if a SCR is dropping out. You'll be missing the positive or negative current on one phase.

I'm not familiar with that controller but if it uses pulse transformers for the gates then it's fairly easy to check by only measuring the gate to cathode voltage. The card will produce gate pulses, with a zero voltage dead time between each pulse. When the SCR commutates-on, the SCR itself will produce a gate voltage so the gate driving pulses will no longer return to zero voltage. The gate pulses could go from about 0.5V to 2V and the SCR would likely produce something around 0.5V to 0.8V itself while conducting current.

 

[itsmoked]

Will do Bill.

I'm still scratching my head about what a phase controller feeding a transformer thinks of non-symmetric phases.

I tested all six SCRs with a load and none were stuck on and all triggered correctly. So Mr. Hutz it seems your prophecy panning out.

Now I have to power it up with whatever load I can scrounge up and hope it misbehaves in the same manner as I witnessed in the field.

This Halmar is really a work of art. It's so nicely built with every detail making sense. There are test points for everything.

It indeed has six transformers to drive the three dual SCR units. It also has three transformers followed by three bridge rectifiers to provide the +/-15VDC the whole thing runs on.

It seems they tap off these to nail down the sync so they know when to fire the gates. (This is why I wonder about Bill's funny 3ph.)

Sounds like the current is the thing to watch. It has current coils on every output leg. They use them for current protection so they rectify them and filter them for examination of some threshold being exceeded. Hopefully the pre-rectified test points will work to allow me to monitor the SCR current.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

You could use a low-ohm resistor instead of the current transducer. Put something in series with the output to get around 1 V drop at rated current. Your scope input is floating, so you can easily measure the voltage drop across the resistor. And that will tell you a lot.

Some thyristors have a problem getting up to holding current before the gating pulse is over. Not a problem with resistive loads, but definitely so with inductive loads.

An easy test would be to put a set of incandescent lamps (if you can find them these days :-( ) parallel to the transformer primary and see if the systems gets quieter. If so, you could either look for defective snubbers on the primary side of transformer (same duty as the lamps) or look for abnormally short gating pulses.



Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[waross]

How about rolling the connections on the transformer primary as a quick and dirty test for a transformer fault?

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[itsmoked]

This thing is thrashing me so thanks for any suggestions.
Bill rolling it didn't seem to change anything.

It only malfunctions for about 15 seconds every hour or so.
This is preventing me from studying the problem while it occurs.

What I have been able to see is that the currents go way out of balance. Like from 33A,33A,33A to maybe 22A, 33A, 43A during an episode.

I made a half-wave test lamp and discovered that it extinguishes completely during a fit on channel three only. (See sketch)

I was also able to catch the gate drive as seen across the SCRs during an event. (See sketch)

First: Do you agree that for the lamp to go out, as hooked up, it would mean that SCR was stuck ON during a period it would normally be off?

And secondly any suggestions as to what component in the drive circuitry might cause this come-and-go problem?

Please note my sketch shows the drive hardware on probably the opposite channel that's misbehaving.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[itsmoked]

Here's a theory:
This is looking like a stuck-on issue. But the SCRs are fine.
The abnormal drive picture doesn't look like an extended trigger pulse.
This controller feeds a transformer.
The problem ONLY occurs at power settings above 50%.

I am suspecting the issue is a dVdt problem - failure to commutate. This means the snubber is not getting it done. Perhaps the MOV is worn out?

Any agreement here?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[ScottyUK]

Don't forget that 'stuck on' is in reality premature triggering. It could be failure to recover blocking capability but after a half-cycle of reverse bias that would be really unusual.

Could you have noise getting in to the gate drive electronics further back in the light current section of the controller, perhaps as a result of switching on a different phase? That might explain why the problem only occurs above a certain setting - a combination of the firing delay angle of the other phase being suitable to cause false triggering of the pair you're having problems with, and possibly a more powerful interfering source due to higher current.

Can you have a look to see if the false triggering happens to coincide with switching on a different phase?


----------------------------------

If we learn from our mistakes I'm getting a great education!