High Current Triac Circuits

[normad]

Hi,
Im developing a triac circuit for a high current switching application. And i was able to find this triac TIC263 which can support 25Arms continuous current.
The triac comes in a SOT-93 package. but im confused as what to use as the base of my circuit. I cant use PCB because the tracks would have to be massive to support 25Arms and i cant wire them because i cant find a suitable base for the SOT-93 package, with screw-in terminals.
How do you go about placing high current SOT-93 devices in a circuit?

thanks in advance

 

[MrAl]

Hello there,

To use that product in a single PC board application requires some thoughtful parts placement and PC board design.

The first thing you will note is that the device has a mounting tab. That means it gets bolted to a heat sink. The heat sink will have it's own mounting holes which will be mounted to the chassis in some way, either directly or insulated depending on the triac (isolated tab or non isolated). Once the heatsink mounting method is figured out, the PC board design is configured to accept the three pins of the triac and the board is to be mounted right under the triac mounted in a position that agrees with the triac heatsink mounting (the pins must stick through the board holes properly). The assembly is such that the triac is soldered into the board and bolted onto the heatsink.

The layout of the PC board is such that the two higher current pins (not the gate) are located as close as possible to the power source board entry point and load leads entry points. This means the copper clad trace runs will be as short as possible. They will also be much wider than a normal signal trace (like the gate trace) but they dont have to be extremely wide because the PC board is designed with short traces for the main power runs, and that means less heating in the board traces even though there is higher than usual current flow through them.

You can look up trace width specs on the web and see what is recommended, but try to find a guide that also takes the total trace run into account too. For example, if an 1/8 inch wide trace 10 inches long can take 10 amps without a problem, then say a 1/16 inch trace only 2 inches long can probably take 10 amsp too, but that's because the run length is shorter. Note that shorter run wire can be of smaller diameter too. Note also that the copper clad thickness comes into play here too, because that affects the cross sectional area of the conductor which affects max possible current.

If you have to experiment with this you can try making a board with several trace widths of the exact length (or longer) that you will need. For example, make one 1/16 inch wide, another 3/32 inch wide, another 1/8 inch wide, etc., and wire them externally in series. Run 25 amps through them and see what happens. You might be able to throw a little distilled water on the top of the traces to see which ones heat up the most, or come up with your own idea for this.

To get an idea with a calculation if you cant find anything else, you can base the temperature rise roughly on the rule of thumb of 1 square inch of space per watt makes the trace temperature rise by 60 degrees C. You probably want to stay lower than that though like maybe 1/2 watt per square inch.

A 1 oz per square foot copper clad board trace 0.1 inches wide and 1 inch long has resistance close to 0.005 Ohms. Halving the "oz" doubles the resistance, doubling the "oz" halves the resistance. So a 0.5 oz board will have 0.010 Ohms for that same trace size. Also doubling the trace width halves the resistance but also increases surface area. Doubling the length doubles the resistance but also increase the surface area. I'll verify this later today but that seems ok.

Using this guide and with a trace 0.1 inches wide and 1 inch long, it has 0.005 Ohms and surface area 0.1 square inches, so with 25 amps the power is around 3 watts,

I'll be taking a look on the web too.
I found this:
https://www.4pcb.com/trace-width-calculator.html

Dont know how accurate it is though.

 

[KeepItSimpleStupid]

Well, you forgot two important points:

Thermal washer: https://www.evita.lt/en/det-17960-silicone-washer-for-sot93-top3.html
Mounting kit: https://uk.rs-online.com/web/p/heatsink-mounting-accessories/0298500/

The tab is live.

 

[ronsimpson]

My last high current project: I used traces on both sides in parallel, and removed the solder mask and built up solder over the copper.

Don't use 'thermals' on the high current area.

 

[MrAl]

The heat sink will have it's own mounting holes which will be mounted to the chassis in some way, either directly or insulated depending on the triac (isolated tab or non isolated). Once the heatsink mounting method is figured out

Hello,

Or insulate the heat sink itself from the chassis and note that i mentioned the triac (isolated tab or non isolated).
But thanks for making this more clear.

 

[MrAl]

My last high current project: I used traces on both sides in parallel, and removed the solder mask and built up solder over the copper.

Don't use 'thermals' on the high current area.

Hi,

That is a method that is often used but once you read up on the electrical conductivity of standard solder you wont want to do it that way anymore. An alternate method is to solder heavy gauge solid copper wire to the trace making sure the ends are well soldered to the trace and tack soldering along the length.

 

[normad]

thanks a lot guys. really appreciate the descriptive replies.

Mr.Al i found that link on trace width calculation and according to the the trace needs to be 1-2mm thick for it to be of acceptable width(2mm). My first idea was to cut a stensil with a copper sheet for the circuit. But considering how expensive copper sheets are that might not be feasible. I like the idea about soldering copper wire on to the trace. Will have to put a lot of effor to make it look nice though.
what would really be nice was if there was a base for this with screw in terminals so that one can use wires directly instead of pcb traces.

 

[KeepItSimpleStupid]

You can solder direct to the pins and use heat shrink if this is a one of.

 

[ronsimpson]

That is a method that is often used but once you read up on the electrical conductivity of standard solder you wont want to do it that way anymore.

We made huge quantities of power supplies by building up solder on the high current traces. Many of these boards have very thin copper. So thin you can't believe it. The solder is not a thin layer but is mounded up. I know solder has high resistance. Very thin copper also has high resistance. The solder mounded up also helps get more surface area to remove heat.

 

[MrAl]

Hi again,

normad:
Another technique is to simply make heavy gauge 'jumpers' for the board and use drill holes for the ends just like any other jumper. To the hole for the triac lead would be small (because the triac lead is relatively small diameter) but right next to that on the PC board would be a larger hole for the heavy gauge jumper, and that jumper would act to connect the triac lead to whatever it had to be connected to on the PC board via another larger diameter hole for the other end of the jumper. This means you dont even need a trace anymore, just a heavy jumper of the right gauge. For a normal sized PC board #12 AWG wire would work. This wire would have insulation though not bare.

ron:
Well if you mound it up it might work, but consider that the conductivity is much lower than copper so it has to be pretty thick and that means a lot of solder gets used up. Back when i calculated the required thickness of the solder i moved to using heavy gauge solid copper wire because it works much better.
But what kind of solder did you use for those traces?

 

[ronsimpson]

Well if you mound it up it might work, but consider that the conductivity is much lower than copper so it has to be pretty thick and that means a lot of solder gets used up. Back when i calculated the required thickness of the solder i moved to using heavy gauge solid copper wire because it works much better.
But what kind of solder did you use for those traces?

I have to assume the most low cost solder one can find. Lets assume the solder is 5x higher in resistance.
The copper is deposited on the board not etched off. It is thinner than anything you can get in the US.
Assume; the solder is 5x the thickness of the copper. (Probably much thicker)
The solder resistance is equal to the copper thickness.
The two resistors are in parallel. The total resistance is 1/2.

Most people are thinking 1oz copper and 1oz solder and the resistance changes very little. I agree with your thoughts.

 

[normad]

thanks a lot for the ideas guys

im thinking of getting a 1.5mm copper sheet which would cost about 50$ around here.. then getting it cut to match the traces using a CNC plasma cutter. Ive used that for a lot of other things with aluminum so i dont see why it wont work with copper. then if i paste it on to a PVC board it would have the finish and quality of a printed circuit. except i can use it for currents of around 30A!

any downsides of this method?

 

[MrAl]

thanks a lot for the ideas guys

im thinking of getting a 1.5mm copper sheet which would cost about 50$ around here.. then getting it cut to match the traces using a CNC plasma cutter. Ive used that for a lot of other things with aluminum so i dont see why it wont work with copper. then if i paste it on to a PVC board it would have the finish and quality of a printed circuit. except i can use it for currents of around 30A!

any downsides of this method?

Hi,

Why use something that expensive when it is only a couple runs that have to be heavy gauge? Doesnt make sense to me, but it's up to you of course.

 

[MrAl]

I have to assume the most low cost solder one can find. Lets assume the solder is 5x higher in resistance.
The copper is deposited on the board not etched off. It is thinner than anything you can get in the US.
Assume; the solder is 5x the thickness of the copper. (Probably much thicker)
The solder resistance is equal to the copper thickness.
The two resistors are in parallel. The total resistance is 1/2.

Most people are thinking 1oz copper and 1oz solder and the resistance changes very little. I agree with your thoughts.

Hi again,

Yes you are probably right, but not me

Actually common solder resistivity is about 9 times higher than copper. That means to *equal* the copper we need a layer of solder that is 9 times as thick as the copper, but that's not saying that much either because we might need 5 times that to get the required total resistance. Just twice as much might not be good enough, but if it is there are other considerations too. So for 0.00135 inch thick copper we need to put a layer of solder that is about 0.012 inches thick to *equal* the copper resistance. IF that's all we need that's great perhaps, but if we needed 5 times that we'd have to dump on a layer that is 0.06 inches thick, and that's not something we would want to do using wave soldering.

Another catch is that is at *room temperature*. As the temperature increases, the resistivity increases. And the third catch is that the total surface area only goes up a little with the volumetric increase (adding more material) so that means we loose part of the effect of surface area cooling meaning the temperature rises higher than if it was just a flat sheet (one layer of copper).

So the three to five factors that have to be considered are:
1. Resistivity 9 times higher than copper.
2. Total thickness requirement might be large requiring a large amount of solder.
3. Surface area to volumetric ratio does not increase much as we pile on solder.
4. #3 leads to adding more solder to make up for the temperature increase caused by lower surface area to volume ratio.
5. Reliability: how well does piled up solder hold up over vibration and years of use, and how hard is it to get right the first try.

Silver solder resistivity is somewhere around 7 times higher than copper so that's a little better but more expensive.

All of the above led me to do it with heavy gauge copper wire with end holes drilled in the board to accomadate the wire. This allows easy adjustment of the wire diameter for different current levels too. It's also easier to get right without having to do a perfect soldering "pile up" job.

On applications that only need a little less resistance it may be worthwhile, but a better design would use another method.

But there is yet more more consideration here. The THERMAL conductivity of solder vs copper. The thermal conductivity of solder is lower than copper, so in effect it can even prevent the proper cooling of the copper underneath because it can act as a pseudo thermal insulator compared to the copper. If the thermal conductivity was too much lower it would definitely not act like more copper on top because the copper on the bottom would get hotter than it would without the solder even with less current flow though it, and that would again make the resistance rise.
A test would be to measure the resistances and try to estimate the temperature rise with various current levels.

I reluctantly add still yet another effect to consider, when working with significant frequencies (we've only really talked about DC effects so far). That's the skin effect.
The skin effect is the effect where the current tends to distribute itself around the outside of the conductor. Simply put, the copper will get about 1/2 the current and the solder about 1/2 the current that would normally be present near the surface of the copper alone. That puts the resistance at just a little over half of what it would be with pure copper, effectively making the AC resistance actually higher than lower. I dont think solder is very magnetically active, so we dont have to worry much about an increase due to that unlike if we used nickle plating instead. This all gets a little complicated though as we also have the proximity effect which may alter the current distribution in a manner that could favor the solder coating slightly better than with just the skin effect alone, but probably not enough to make a huge difference.
For DC currents this wont be an issue though.

 

[ronsimpson]

You have thought about this for a long time.
I read several advertisement for solder that is 4 to 5x resistance. I believe you have 9x solder. I also found solder that is very close to copper.

My prototypes, using heavy copper, worked fine.
The first China prototypes had arias where the copper lifted off the board from too much heat. (too thin copper)
By removing the solder mask and mounding up the solder, the traces ran cool enough to go into production.
We do not have machines to add wire to the top of a trace. We do have solder wave machines that add solder to any open areas.
I do not know the exact numbers. That is a production engineering problem.

MrAl, You have good thoughts. Thank you.

 

[MrAl]

Hi,

Solder with resistivity close to copper? That sounds very good. Expensive or no?

Yes i have thought about this in depth, and years ago i ran through the calculations and found some startling results that i never expected that's why i end up mentioning this now and then. There are a lot of issues to be addressed and that's what i like to tell people.

If your prototype worked well then that's great. I also understand that regular jumpers are not very compatible with insertion machines although that could change in the future. So there we want to avoid jumpers of any kind if possible. Solder over copper is also good to prevent corrosion.

I looked over the web quite a bit and can not find any info about this. I am sure somebody must have tested this thoroughly by now but that's part of what bothers me about this for the moment: we dont have any solid data to tell us about the feasibility of using solder to increase the current rating of a trace although there must be some out there somewhere.