14

I know this is in some way a different version of the prior question In i.e. a TO-220 package, why is mounting the (pin connected) tab to case safe?

But I am not looking for ideas what to do to mount a TO-220 (in my case an SCR that has the anode connected to the tab. That's the input with the highest potential from ground!)

What I am asking is why in the world would they have designed it that way? These things even have a hole to screw it to a heat sink and you'd want the best contact to the heat sink, not layers of plastic in between.

Now I am certainly not the first one discovering that this is rather annoying. So why did the engineers so unanimously design things that way? Why not just leave it disconnected?

Gunther Schadow
  • 1,441
  • 7
  • 22
  • 2
    Your subject line is not a good match for your question. Commenting on the subject - others have noted that the substrate tends to be connected to the tab. In negative voltage regulators the most negative connection is Vin-, whereas in positive voltage regulators it is ground. This fact is reflected in the differing device pinouts. – Russell McMahon Nov 14 '19 at 03:56
  • Two words: cost and cooling. – winny Nov 14 '19 at 10:40

4 Answers4

25

Why not just leave it disconnected?

Because of the way the IC is made. It starts off as a (relatively) large piece of doped silicon, and the features are deposited / etched / implanted into one side. That leaves the back side of the die electrically connected to one node of the circuit. Which node it is depends on what circuit is being made. Generally for things like ICs, it's the most negative terminal. It is whatever makes the circuit the most straightforward to design.

For a power device, the designer also wants the lowest thermal resistance from the active region to the case. The most critical part of this path is the narrowest directly under the die, which is why a direct connection is used.

It's far better if an insulating layer is needed to put it between tab and heatsink, where the area available for heat conduction is orders of magnitude larger than the area under the die. And remember not all applications will need this insulator, some can use an isolated heatsink.

The bottom line is cost/performance ratio. An insulator in the package would increase cost and decrease thermal performance. There are niche applications that would pay that premium, and there are some isolated tab power electronics out there to prove that if you look for them. But most equipment manufacturers will do the extra bit of mechanical design to get the best overall cost and performance.

Neil_UK
  • 158,152
  • 3
  • 173
  • 387
  • 13
    And sometimes, one can ... re-design the circuit so that no mica insulation is required, at all. I've seen a re-design of a power supply using ground return regulation for no better reason than it allowed the BJT's tab to be mounted directly to the heat sink for better thermal conduction. So, while the *standard* topology might present dangerous voltages to the TO220 tab, sometimes a re-design with an eye to making it a safe potential can kill two birds with one stone and also allow the manufacturers to deal with their own product design concerns, too. Imagination is good. – jonk Nov 13 '19 at 05:40
  • 4
    I always wondered why the power transistors used in switch mode supplies were *live* (which seems the worst possible outcome to me) - I assumed the same as you describe, so it's nice to see my assumption was correct. – Ralph Bolton Nov 13 '19 at 15:06
  • 3
    @jonk now that's a pretty cool idea. Instead of connecting the load of the cathode of the SCR, I can put it into the ground return. Except, well, if its gate is closed, the potential on the heat sink goes up again, so in my specific application it wouldn't help. But at least I understand now why this is the way it is. – Gunther Schadow Nov 13 '19 at 18:17
9

(in my case an SCR that has the anode connected to the tab. That's the input with the highest potential from ground!)

What I am asking is why in the world would they have designed it that way?

Because the anode is at the bottom of the die, which is bonded to the tab. Most power transistors (bipolar and FET) and some regulators are constructed in a similar manner, where the collector or drain is connected to the tab. Unfortunately that means the heat sink will usually be above ground, unless an insulated mounting kit is used.

Construction of a thyristor (SCR)

Enter image description here

It is possible to insulate the die from the tab, but this increases thermal resistance which makes it more difficult to cool.

Another advantage of having a direct connection is that it can be used as a terminal as well as a heat sink. Some packages (for example, DPAK) which are designed to have the tab soldered to the PCB, often don't have a separate lead for the collector or drain.

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
Bruce Abbott
  • 55,540
  • 1
  • 47
  • 89
  • 5
    Just to add, connecting the case tab to one of the "power" terminals can be quite advantageous in compact high power designs, where one has to parallel together several devices. One would connect all tabs of the devices to the heatsink, and use this both as a heatsink and as a common bus bar. – Richard the Spacecat Nov 13 '19 at 10:07
5

A transistor or any other semiconductor device is made from silicon. Initially you get a slice of silicon. Then you induce some N and P zones on one face of this slice. There you will connect bondings (welded tiny wires), that connect the silicon chip to leads. The opposite face of this slice you will need to solder to the case base - the tab in case of TO-220 or "exposed pad" of some SMD chips.

This solder holds the chip in place and guarantees the lowest possible thermal resistance. But, on the other hand, it creates parasitic electrical connection to the chip's substrate. There is no way to insert some insulation between the chip and the tab - the chip area is too small and when you will try to insert any insulation layer, you will dramatically increase the thermal resistance "junction-case".

In fact, you always want to insert insulation to connections with the maximal contact area. Thermal resistance is roughly proportional to H/S where H is the thickness of the insulating layer and S is the contact area.

Dave Tweed
  • 168,369
  • 17
  • 228
  • 393
Владимир Тактакишвили
  • 151
  • 3
2

I am not looking for ideas what to do to mount a TO-220 that has the anode connected to the tab

There are things like heatsink insulator pads for cases like yours.

enter image description here

Of course there is a penalty in thermal resistance compared to a direct connection to the heatsink, but having a layer of insulation inside the IC would result in a similar penalty as well. Having the insulation as an optional external component enables you to have a design with direct connection between the IC and the heatsink and keeping the heatsink insulated, if you need to minimize the thermal resistance.

Dmitry Grigoryev
  • 25,576
  • 5
  • 45
  • 106
  • 1
    "would result in a similar penalty as well." - I believe that penalty inside would be greater, because area there is smaller. – Mołot Nov 14 '19 at 10:18
  • 1
    @Mołot Depends on how it's done. You could sandwich an insulator layer between two copper tabs for instance. – Dmitry Grigoryev Nov 14 '19 at 11:44