Mains clearance in a TRIAC driver

[gxti]

I'm making a TRIAC driver board for use with my reflow toaster oven project, which means mains voltages. Yay! IPC-2221 says that U.S. mains voltages require 0.4mm clearance if coated with solder mask. A few questions:

Can I run overlapping tracks on both sides of the board? I don't know how thick the PCB is but it's a standard, 2-layer FR4 PCB. Since the 200 mil "tracks" (fills, actually) running from the screw terminal to the TRIAC are sort of, erm, wide, they have to either overlap or taper at the ends.

How much clearance do I need between the high-voltage section and the low-voltage section? I've kept them at least as far apart as the opto-isolators are wide but I'm curious what the standards say.

Here's the schematic: http://partiallystapled.com/~gxti/circuits/2011/01/19-driver-schem.png
And the PCB: http://partiallystapled.com/~gxti/circuits/2011/01/19-driver.png

I haven't finished checking clearances yet but any comments about the circuit or the PCB are welcome. The top of the schematic is the TRIAC driver of course, and the bottom is a zero-crossing detector. I'm especially curious whether I really need two capacitors and two resistors for the detector, I got that from a Fairchild app note but there weren't any voltage or safety ratings in the example so I had to assume the worst.

[arcom]

Standard FR4 PCB is 1,6mm so yes, you can run tracks on both side of the board. However, there are few things you might consider:
1. use an optocoupler with integrated zero-crossing detector (such as MOC3041 or similar) if you don't need the actual zero-cross indication
2. move the triac as close as possible to the connector, use narrower tracks and make them thicker with a layer of solder
3. add at least an RC filter for snubbing of the triac (check out the datasheet for MOC3041)

As for the separation between high and low voltage sections, use around 6-8mm (width of the optocoupler).
For the zero-cross detector you can use only a single resistor instead of two caps and two 1K resistors but you will need a high power rating resistor (10K/5W for 120V AC mains or 22K/7W for 240V AC).

[gxti]

Thanks for the advice. The reason I'm not using a zero-crossing opto-TRIAC is that I want to PWM the output -- the detector is for keeping the pulses in sync with the AC waveform while still allowing me to delay them to get the varying levels of power I want. I also opted not to use a snubber since the load is always purely resistive, and the connector is where it is because that big pink thing is a rather large heatsink :-) That said, I do intend to move it to the right-hand side in an attempt to squeeze this onto a smaller PCB.

Why do you recommend narrower tracks and solder coating? I thought the conductance of solder was relatively bad, but I guess that would matter less since it would be much thicker than the copper. Another problem is that the tracks would no longer be soldermask-coated so I would need the full 3.2mm clearance required by IPC-2221.

[Zero999]

Thanks for the advice. The reason I'm not using a zero-crossing opto-TRIAC is that I want to PWM the output -- the detector is for keeping the pulses in sync with the AC waveform while still allowing me to delay them to get the varying levels of power I want.

I think you're confusing PWM (also known as burst control) with phase control.

You can use a zero crossing TRIAC for PWM, it can't be used for phase control.

There's no need to use phase control for a heating appliance, you're better off with PWM as it's easier on the TRIAC and produces less EMI.

I also opted not to use a snubber since the load is always purely resistive, and the connector is where it is because that big pink thing is a rather large heatsink :-) That said, I do intend to move it to the right-hand side in an attempt to squeeze this onto a smaller PCB.

Yes, no snubber is required for a resistive load.

Why do you recommend narrower tracks and solder coating? I thought the conductance of solder was relatively bad, but I guess that would matter less since it would be much thicker than the copper.

Yes solder is a poor conductor so I think that's bad advice: use the correct thickness PCB traces and you'll be fine.

Another problem is that the tracks would no longer be soldermask-coated so I would need the full 3.2mm clearance required by IPC-2221.

What voltage is the mains in your area?

I think you want more clearance than that between the DC and mains side.

Why have you used a separate opto-coupler for zero crossing? I think you're better off using the AC from the secondary side of the transformer.

I hope you're aware you need to account for the phase shift generated by the capacitors you've used to limit the current to the LEDs. In a purely capacitive circuit (which nearly to what you have) the current leads the voltage by 90^o so one LED will turn on and the other turn off as the mains voltage passes its peak, which will briefly turn 0Q1 off for a few hundred µs.

[arcom]

OK, just out of curiosity - what's the power rating of the heater?
I don't see what's the problem in using solder to make the traces thicker. We are not talking about 1km of PCB trace; just a few centimeters. If it works for car amplifiers where the currents are around 40-60A (and with longer PCB traces) then I really don't see a problem with this application. You don't have to leave entire trace exposed - just leave the middle of the trace exposed.

Fine, snubber is not needed. I kind of missed the heater part and assumed a more general use of the triac driver. My mistake.

[gxti]

I think you're confusing PWM (also known as burst control) with phase control.

You're right, phase control is what I was originally after. I presume PWM in this case would be firing or not firing entire half-cycles, in which case a ZC TRIAC would work fine. Since you encouraged me, I will now go this route, but when would phase control be a good idea, if ever?

What voltage is the mains in your area?

I think you want more clearance than that between the DC and mains side.

120V. I was referring to mains-to-mains clearance, not low-voltage clearance, according to IPC-2221 rules at 170V peak-to-peak. The low-voltage side is still off in its own little corner no matter what I do to the high-voltage tracks. It's irrelevant anyway since you agree about keeping the mains tracks coated.

Why have you used a separate opto-coupler for zero crossing? I think you're better off using the AC from the secondary side of the transformer.

Mostly because the Fairchild appnote suggested it.

I hope you're aware you need to account for the phase shift generated by the capacitors you've used to limit the current to the LEDs. In a purely capacitive circuit (which nearly to what you have) the current leads the voltage by 90^o so one LED will turn on and the other turn off as the mains voltage passes its peak, which will briefly turn 0Q1 off for a few hundred µs.

Certainly, I was going to set up a crude PLL in the microcontroller to account for all the various phase shifts, including that, delays in the opto-TRIAC and TRIAC, etc. Also irrelevant since I might as well use ZC anyway.

Thanks a bunch for the input!

[Chasm]

Solder does not really help because it's specific resistance is much higher than copper.
"Much" is a factor of about 8 for Sn60Pb, which means that the solder has to be 8 times as thick as the copper track to have the same resistance (and thus as a result halve the overall resistance of the track).

It is more efficient to use wider tracks, keep tracks as short as possible so that you may accept higher self heating, user thicker base material, or if necessary solder copper wire on the PCB.

[gxti]

OK, just out of curiosity - what's the power rating of the heater?

The heater is rated at 10A, I haven't had a chance to measure it in action since my Fluke is fused at 10A, and I don't really feel like risking a fuse, and I don't have anything handy (that I know of) to measure it directly. I'll probably buy a clamp ammeter attachment for the Fluke one of these days...

I'm targeting the driver to handle 15A to account for inrush and so I can use it/sell it/give it away for use with other ovens.

I don't see what's the problem in using solder to make the traces thicker.

I don't see a problem either, but I also don't know why I should do it, and unlike regular copper traces I don't have a calculator handy to tell me how big the trace needs to be. I don't really mind running a fat trace over 1 inch on a very simple board, especially if it means less exposed high-voltage track.

Fine, snubber is not needed. I kind of missed the heater part and assumed a more general use of the triac driver. My mistake.

No problem, a lot of people probably ask about building general-purpose SSRs and would need a snubber. I'm cheap and don't feel like buying big resistors and capacitors if I can avoid it.

[Zero999]

You're right, phase control is what I was originally after. I presume PWM in this case would be firing or not firing entire half-cycles, in which case a ZC TRIAC would work fine. Since you encouraged me, I will now go this route, but when would phase control be a good idea, if ever?

PWM can only be used for loads with a long time constant such a heater or a large universal motor driving a huge flywheel. You couldn't use PWM to drive a lamp because the frequency would need to be lower than the mains frequency so the lamp would flicker, same for a motor which would vibrate horribly.

120V. I was referring to mains-to-mains clearance, not low-voltage clearance, according to IPC-2221 rules at 170V peak-to-peak. The low-voltage side is still off in its own little corner no matter what I do to the high-voltage tracks. It's irrelevant anyway since you agree about keeping the mains tracks coated.

Well technically mains is also low voltage but I know what you mean, you're talking about line to line separation as opposed to line to SELV separation.

The heater is rated at 10A, I haven't had a chance to measure it in action since my Fluke is fused at 10A, and I don't really feel like risking a fuse, and I don't have anything handy (that I know of) to measure it directly. I'll probably buy a clamp ammeter attachment for the Fluke one of these days...

Is 10A just the fuse or is it the proper rating? does it have a power rating?

The inrush current shouldn't be too much as it's a low temperature heater rather than a high temperature incandescent light.

Anyway, you probably need a heat sink on the TRIAC which will either need to be connected to earth/ground or well insulated from the case.

[gxti]

Is 10A just the fuse or is it the proper rating? does it have a power rating?

It says 1300W on the box. And yeah, I'm definitely putting a beefy (3K/W) heatsink on there. I followed Dave's video on thermal management and worked out 25C + (1.2 (JtoC) + 0.4 (Sil Pad) + 3 K/W (heatsink)) * 20W = 117 C at full-cycle 15A operation. The TRIAC datasheet has an absolute maximum junction temperature of 125 C which is cutting it close, so perhaps I should consider splitting the TRIAC or lowering my target rating.

[PetrosA]

Out of curiosity, what's this going to cost to build? I'm wondering because a 2000W dimmer, which is basically what you're building, is both massive and expensive. Rotary types cost about $70 and have ~3"x4" of heat sink while the slide type are more like $140 and have ~4.5"x4" of heatsink.

[Zero999]

A 2kW dimmer shouldn't be that expensive and he's not building a dimmer which uses phase control, he's decided to opt for PWM which will be cheaper because it doesn't require large RFI suppression inductors and is easier on the TRIAC.

I Googled for a suitable TRIAC, found the MAC15A6G and its datasheet.
http://pdf1.alldatasheet.net/datasheet-pdf/view/172093/ONSEMI/MAC15A6G.html

If the current is 10A and its conducting both for the full cycle, the power dissipation is about 11.5W (see graph on datasheet).

The thermal resistance from junction to case is is 2^oC and the maximum junction temperature is 125^oC.

If the ambient temperature is 40^oC the maximum allowable thermal resistance between the junction and ambient is (125-40)/11.5 = 7.4^oC/W

Assuming an extra 2^oC/W for the thermal pad and a safety margin, a 3^oC/W heatsink will be fine and is not very expensive.

Also note my point about PWM being easier on the TRIAC and dissipating less power than phase control: compare the dissipation at half the power setting, with PWM at 50% duty the power dissipation is 5.75W, with phase control and a 90^o firing angle, the power dissipation is about 9.25W. This is because when the TRIAC is triggered at 90^o lots of energy is dissipated to the power dissipation is higher even though the RMS current is the same as with PWM when the duty cycle is 50%. Using PWM, there are no switching losses, only conduction losses.