SparkFun Forums

[reklipz]

No, clarifying AC in the subject line was not meant as an insult to your intelligence, but instead to make it easier for someone searching to find this thread, as AC is too short to work.

Anywho, I'm wondering how wide traces should be on a board that I'm designing. I'm using a triac to switch 120VAC, at something around 12.5Amps (1500 watts). I'm pretty sure that what I have on there right now isn't going to cut it (i think like 50mils? ).

I found an online calculator for this, and given my inputs with a trace thickness of 35um (google says this is standard thickness), the output was something between 130mils and 230mils, depending on the calculator used.

Also, I noticed that temperature also played a role. The board is a reflow toaster controller, and I plan to put it in the oven itself (not in the heating part, but in the empty space next to it behind the controls), so it will probably get hotter than if it were external. I plan on putting in a fan or two for ventilation of the area, but this is only if it is needed.

Also, there will be DC components on the board, how much of a clearance would be needed around the AC components to keep things working?

-Nate

[bigglez]

I'm wondering how wide traces should be on a board that I'm designing. I'm using a triac to switch 120VAC, at something around 12.5Amps (1500 watts). I'm pretty sure that what I have on there right now isn't going to cut it (i think like 50mils?

Greetings Nate,

I use this calculator.

I found an online calculator for this, and given my inputs with a trace thickness of 35um (google says this is standard thickness), the output was something between 130mils and 230mils, depending on the calculator used.

The result I get for 2oz external copper at 13.5A is 366mil (9.3mm).
Assuming a 40degC ambient and allowing a 20degC rise.

The short answer is make the high current traces as wide
as practical. Perhaps a polygon fill is better? Also I noticed
while reverse engineering other projects that PCBs discolour
due to high temps more at the junctions than in the middle
of a trace run. High current circuits often have problems
at soldered joints and connectors. If the traces must go
from one side to the other use multiple vias with large drills.

Also, there will be DC components on the board, how much of a clearance would be needed around the AC components to keep things working?

The required clearance from any primary (ac mains) traces
to other curcuits or to ground is approximately 6mm (236mil)
for agency approvals (VDE 0631, UL, etc.) Here's a good
summary.
As a practical matter you can reduce this as soldermask is
an insulator - you wont get approval but you can have peace
of mind.

Some other tricks are to cut slots in the PCB or cover the
high voltage points with a conformal coating or RTV.

I think the US Mil standard is 3mil per volt (so 510mil for
120Vac 170V peak). Commercial PCBs (for TV and CRT
monitors) were designed to 1mil per volt (back to 170mil
for ac mains).

Another approach is to export the high current connections
to a second PCB set up just for this function.

Comments Welcome!

[winston]

Personally, I'd be inclined to use point-to-point wiring for high current/high voltage parts (if they are reasonably straightforward) rather than on a PCB, keeping all that stuff separate. That's just a gut feel, though (indeed, would experienced people recommend that approach?)

[reklipz]

Well, this is how things look currently:

I have yet to add a transformer, so its more than likely to change. Grid is 100mils, each of the main traces are 300mils wide.

Comments?

[Philba]

the one issue I'd be concerned about is the contact area on the pin going to the heater connector. it looks like it's about 25 mil X 100 mil. I don't know if that's an issue or not - if the chip is rated > 12.5A then it must be ok.

By the way, if you tin the high current traces, you will increase the thickness by a significant amount and thus increase the carrying capacity of the trace. I have not idea how to calculate that, though.

[reklipz]

if the chip is rated > 12.5A then it must be ok.

The chip is rated for 35Amps, so yeah, lets hope that'll work.

By the way, if you tin the high current traces, you will increase the thickness by a significant amount and thus increase the carrying capacity of the trace. I have not idea how to calculate that, though.

Sounds like a plan, I just have to make sure to leave the solder mask off of the traces?

[hooziewhatsit]

Another thing I've done, in addition to making the traces as wide as possible on all four layers, is to leave the solder mask off the bottom layer, and solder in the wire trimmed from resistor leads.

I'll put two of those legs covered in solder on each trace. Then I'll cover all of it with a conformal coating spray.

[bigglez]

Well, this is how things look currently:
.....
I have yet to add a transformer, so its more than likely to change. Grid is 100mils, each of the main traces are 300mils wide.

Greetings Nate,

Bad News. Your design will *not* work at all. Here's why:

(1) The Alternistor is a trade name for a Triac. When conducting
the forward voltage drop is approximately 1.2V. At 13.5Amps
you need to dissipate 26Watts of energy. A heat sink is needed!

(2) The footprint looks like a D2PAK, certainly a SMT. The only
thermal contact is through the pins and the tab. There is not
enough area to dump the heat.

(3) The gate signal is milliamps and low voltage, so it does not
require a fat trace (although having one there help with
power dissipation).

(4) The driver for the Alternistor is an opto-coupler and
resistor. The resistor should have a 1/2W rating, you seem
to have a tiny SMT part, possibly an 0805.

(5) The clearance around the opto-coupler needs to be
6mm from low to high side. An SMT part with 150mil
body can't do that. Take a look at the DIP06 package
and don't run any traces under the body. (There are
modified TH packages that have 6mm clearance - the
leadframe extends out from the body).

(6) You do not need four terminals for a two node circuit.
Unless you need the common terminal for another circuit
(not shown) you can use a two terminal part.

Here's a project I did with multiple Triacs and opto-
couplers:



Notice that the power devices are bolted to a metal bracket,
which is bolted to the metal cabinet. This eight channel
IO card is rated to 5A, so only 10W (approx) is dumped.
In your case you have 26W or more to dump, and at an
elevated temp (inside an oven, right?).

You can read more about this projecthere.

Comments Welcome!

[/img]

[reklipz]

Wow, thanks for the awesome post.

It looks like I have quite a bit of reworking to do then.

As for why there are four points instead of the two, I also am placing a transformer on board. I could do with only 3 points, and connect the load to neutral off board, I figured it'd be easier to deal with the connections this way.

So, it seems the SMT triac (a D2PAK indeed) is out of the question, I'll have to order some in the TO-220 package. I'll also have to find some scrap hunk of aluminum, and prolly want to put a fan or two in there.

The optocoupler is the a through hole 6 pin dip. I noticed they also had a surface mount dip type, but the through hole version would be "better", no?

The resistor is indeed a SMD 0805, and I posted a question about that somewhere as well, but it seems that even though 1/2W SMDs do exist, its just a wiser idea to go with tried and true through hole.

You're power traces don't look nearly as wide as mine do, is that true or am I missing something?

I'm a little confused as to what is meant by the following:

The clearance around the opto-coupler needs to be
6mm from low to high side. An SMT part with 150mil
body can't do that. Take a look at the DIP06 package
and don't run any traces under the body. (There are
modified TH packages that have 6mm clearance - the
leadframe extends out from the body).

Again, I really appreciate you're taking time to help, I'm not sure what I would do without you guys (other than trial and error of course, which becomes expensive, and, probably begins dangerously, at least in this case).

I'll try to come up with another design that has all of the desired components (transformer, proper resistor, and different triac package if needed).

-Nate

[bigglez]

So, it seems the SMT triac (a D2PAK indeed) is out of the question, I'll have to order some in the TO-220 package. I'll also have to find some scrap hunk of aluminum, and prolly want to put a fan or two in there.

Greetings Nate,

If you want to roll your own SSR, using a triac and optocoupler,
use a TO-3 device and put it on a metal surface (heatsink) not
a PCB at 13.5A 120V.

I read that SFE made an oven controller and they gave up on
running the AC120V through the PCB (IIRC).

The easy way out is to use a commercial SSR and run your
low voltage control to it. A heatsink may still be required.
You can continue to put the AC120V transformer on the
PCB - I do that quite a lot in low power single PCB projects.

The resistor is indeed a SMD 0805, and I posted a question about that somewhere as well, but it seems that even though 1/2W SMDs do exist, its just a wiser idea to go with tried and true through hole.

When the Triac is turned on the voltage drop across the
resistor is nearly 170V (peak AC less diac and triac gate
drops). A 1/2W resistor with 400mil body can take this voltage.
An 0805 is only good to 150V, but under continuous high
voltage the value will shift. Derate these parts!

You're power traces don't look nearly as wide as mine do, is that true or am I missing something?

They aren't. My PCB is fused at 5Amps, so using the calculator
for 2oz cu. and 30degC rise I got 72.7mil. I used 70mil.
The duty-cycle of my project is quite low so 5A continuous is
unusual. I have tested the PCB design to 24.5A peak...

I'm a little confused as to what is meant by the following:

The clearance around the opto-coupler needs to be
6mm from low to high side.

Download this datasheet. Notice that the devices are available in
SMT (256mil body), SOIC (256mil body) and VDE
approved 6mm creepage leadframe. The latter has the
SOIC body but extended spacing to 400mil. With it a
design can pass the 6mm creepage spec. (The open
space from copper pad to copper pad under the device).

I'm not seeking VDE approvals so I stuck with SOIC
opto-couplers and 200mil creepage under the device.

Comments Welcome!

[reklipz]

Hmm, alright, thanks for another excellent post.

I wonder, I have some relays that could handle this current / voltage, should I just give in and use one of them? They've even got quick connect lugs IIRC.

I was hoping to use the SSR route, as it just seems cooler, but if it's really this much of a pain to route, I may as well just give in and use a relay.

I'll still try my hand once more at the opto+triac design, and i'll upload that here, but it's gonna have to wait till the weekend probably.

-Nate

[bigglez]

I wonder, I have some relays that could handle this current / voltage, should I just give in and use one of them? They've even got quick connect lugs IIRC.

Greetings Nate,

Depends upon your desire to get the project up and running.
A relay in the hand is worth two SSRs and a PCB in the future...

Relays are not without their problems. Over time the contacts
will burn out. They are mechanical and therefore noisy.
They will not allow phase-angle control which would be a
huge plus in a tightly controlled feedback system.

I was hoping to use the SSR route, as it just seems cooler, but if it's really this much of a pain to route, I may as well just give in and use a relay.

Many SSRs are intended for chassis mounting with descrete
wiring. You can still put the rest of the controller on a PCB.
Even if you continue down the path of a Triac, you can use
the TO-3 chassis mount variety in place of a packaged SSR or
mechanical relay.

Comments Welcome!

[n1ist]

Just because the data sheets spec it, watch out for ridiculous power ratings - I found some FETs where the manufacturer keeps claiming that they are rated at 120 amps continuous in a TO-220 package, even after we sent them a video of what really happens when you try to run 120A thru it.

Always take the lead size, contact area, and package dissipation into account.

[reklipz]

you can use the TO-3 chassis mount variety in place of a packaged SSR or mechanical relay.

Teccor offers a 25A alternistor in a similar package:


According to the chart, and what you said in a previous post, I'll need to dissipate anwhere from 15 to 25 Watts. The maximum case temp for the amount of current I'll be conducting (with some margin) is 100C.

Is anything special needed for the heatsink, or will a good hunk of aluminum with some fins work?

[bigglez]

Teccor offers a 25A alternistor in a similar package:

According to the chart, and what you said in a previous post, I'll need to dissipate anwhere from 15 to 25 Watts. The maximum case temp for the amount of current I'll be conducting (with some margin) is 100C.

Is anything special needed for the heatsink, or will a good hunk of aluminum with some fins work?

Greetings Nate,

A larger contact area from the device to the heatsink
reduces the thermal resistance and lowers the junction
temperature. Running the device hotter works against you
as it lowers the device performance and will shorten
the time before failure.

The device you show is a TO218 (with isolated die), and
it has a thermal rating of 0.95 degC/w.

The average power for a 13.5A load is 15W (I said 26W
earlier and in error). So the die will be 0.95 * 15 = 14.25
degrees hotter than the case. You should not run the
case above 100degC according to the datasheet.

To get an idea of what this means, take a 15W light
bulb, turn it on, and touch the glass with your palm.
Your heatsink has to lower this energy level so that
you can safely touch it.

A heatsink that keeps the case below 100degC in a
40degC environment (your oven area in a 25deg room)
needs a thermal resistance of (100 - 40)/14.25 = 4.2 degC/W
or less. It would be wise to derate this for safety.

There are heatsinks for the TO218 that meet this spec.
4.2 * 0.8 (20% safety margin) gives us a starting point.
Look for a heatsink with 3.3degC per watt or smaller.

Here's one. Notice the curves indicate a 20degC rise for
14Watts *with a 700 cu ft/min air flow*. The fan in a PC
"silver box" PSU is rated to about 25 cu ft/min flow.

Alternatively, you can use a larger heatsink with no
forced air. This saves the cost of a fan and improves
reliability and lowers noise.

Here's an extrusion that will work. Notice the
equivalent surface area is 37 sq. inches! (approx
6 x 6 inch area).

If it were me I'd head in the direction of a TO3
device as it has two mounting holes and faston
crimp terminals. The device would be off the PCB,
and probably on a finned heatsing like the extrusion
above. The idea is that when it comes to heatsinks
more is better.

Another approach, not discussed yet, is to wire the
heater in smaller segments so the 13.5A is spread over
two or more circuits. Smaller PCB mounted TO-220
or even surface mounted devices could be used.

The same power is dissipated in the Triacs, but its
not all in one tiny device.

Comments Welcome!