Used SPA20N60C3 transistors—something strange

These are the third batch of transistors I've received. When I started preparing the review, I thought I'd give them a quick check and put together a short review, but things went completely differently from what I'd planned...

To stay up to date with new posts on the blog, I recommend following me on Instagram. I'll post notifications about all new posts there, and maybe write about something interesting too: link to my account

You can also subscribe to updates and new articles on the Telegram channel - https://t.me/KirichBlog

First, a few words about the store. I bought the transistors from the store linked here, but the link was different originally. I picked them out and bought them, then for some reason the store closed, and the seller redirected me to another one. Now the store is back, but the links have changed, and the old one leads to a blank page. What happened and why? I have no idea.

So, this is about transistors marked SPA20N60C3. Their datasheet is easy to find online.
These transistors are fairly common, for example in this or this power supply. Although, for example, when I repaired the power supply from this article, it turned out they also use SPA17N80C3, which has a lower current rating and a higher voltage rating.


But you need to be careful, because sellers sometimes shorten the marking to 20N60C3, and there is also an HGTG20N60C3 with that marking. That one is actually an IGBT transistor.


I bought 10 of them: first, to try them out, and second, because I don't need them that often.


They clearly look used. Just like with the previous ones, no complaints here.


Before testing, I noticed two things: the middle lead had broken off one transistor, and the markings were slightly different. Eight have the marking shown on the left, while a couple have the different marking shown on the right.


First, let's check the gate capacitance. As you can see from the datasheet, it's measured with 25 V between drain and source.


For the test, we need an adjustable power supply set to 25 V, with a current limit just to be safe. I set it to 100 mA.
And, of course, a capacitance meter.


In the comments on my previous review, someone asked me to explain how to measure capacitance. Here's how—it's very simple.
Connect the positive lead from the power supply to the middle lead—the drain.
Connect the negative lead to the right lead—the source.
Connect one capacitance meter probe to the left lead—the gate.
Connect the other meter probe to the right lead—the source.

That's it. Apply voltage and read the capacitance. But be careful not to accidentally short the gate to the drain, or you could burn out the capacitance meter.


The capacitance readings with no voltage (but with the power supply connected) and with voltage applied differ: 6219 pF and 2334 pF.


I tested all ten transistors, measuring gate capacitance and on-resistance at 10 V. One transistor with a capacitance of around 2000 pF seemed a little odd, and there was some variation in resistance.
I also checked how the on-resistance changes with gate voltage for samples 1, 5, and 9.
You can see that the resistance drops noticeably at just 6 V, then changes very little.


The next test was supposed to be quick and easy, but as it turned out, it took ages and led to some strange conclusions.
As you can see from the datasheet, the measurement should be made with 10 V on the gate and a current of 13.1 A. Nothing complicated.


I applied 5 V of power voltage with the current limited to 13.1 A, then applied 10 V to the gate, and... the voltage across the conducting transistor first dropped to about 1.8 V, but then started rising very rapidly. So rapidly, in fact, that I had trouble photographing the initial reading. It literally took just seconds to go from the initial 1.7 V to almost 4 V, as shown in the photo.


Okay, I don't get it. Something's not right. For the tests that follow, I’m reducing the current to 10 A and the power circuit voltage to 4.5 V to make the test safer.
Let’s start it up and see. It begins with transistor number 1.
The same roughly 1.7–1.8 V at the very beginning, for just a fraction of a second, then a sharp rise until it hits the voltage limit set on the power supply. The current starts dropping accordingly.
The datasheet shows that the transistor’s resistance is relatively temperature-dependent: 0.19 Ω at 25 degrees and 0.43 Ω at 150.

So only about 3–5 seconds pass between the first and second photos. If I keep the transistor like this for around 10–12 seconds, it reaches about 120 degrees.


Let’s take transistor number 5. Exactly the same thing happens.


Transistor number 9 behaves exactly the same, too. Well, at least there’s some consistency, I thought. My first theory was that it was a counterfeit, with a small die that simply couldn’t get rid of the heat fast enough.
Judging by the results, the temperature starts to rise, and so does the resistance. As the resistance rises, more power is dissipated in the die, and so on. Eventually, the die heats up to 150 degrees or more, and the resistance climbs to 0.42–0.45 Ω. That seems to match the datasheet, but it all happens awfully fast...


Just out of curiosity, I then tried transistor number 10, which was paired with transistor number 9, judging by their appearance and the traces of thermal paste.
And here came my second surprise: this one behaved similarly, but not quite the same. First, the voltage rose several times more slowly. Then the rise slowed down even further, and above 3.8 V it barely crept up at all. In other words, it had reached thermal equilibrium, and the temperature and resistance were changing simply as the heatsink warmed up. By the way, yes, the heatsink finally started to warm up here. Before that, it just hadn’t had time to.

My theory was that the previous transistors had small dies that simply couldn’t transfer heat to the substrate fast enough, and that the transistors themselves were counterfeit. But transistor number 9 was puzzling: it looked just like number 10.

You can also see that transistors 5 and 9 had noticeably higher resistance than number 10. That might explain something, if not for transistor number 1, which behaved the same way but had the same resistance as number 10...



All right, I test transistors 3, 8, 2, and 6, and get the same result as with the first three: rapid, runaway overheating.


I open the datasheet again and see that the transistor’s maximum current is 13.1 A at a case temperature of 100 degrees, but we’re clearly exceeding that.
Note (1) says: “Limited only by maximum temperature.”


I’ll try to check my theory about the small die, so I open up transistor number 3, which also overheated quickly, just like the others.


The first thing I notice is the massive copper substrate. Interestingly, the flange is clearly different from the usual TO-220 package, with a notch at the top. The substrate is thick, 1–1.1 mm, so there shouldn’t be any issues there either.


But this is where I’m surprised again: the die is definitely not small. It takes up almost all the area allocated for it on the substrate.


At this point I realize the problem isn’t the die area or the substrate, so I go back to the datasheet to look up the die-to-case thermal resistance.
And yes, transistors come in both standard and fully insulated packages, so this parameter differs between them, while the datasheet and test specifications are shared.
The full-pack version is rated at 3.6 degrees per watt, while the standard version is rated at 0.6—a sixfold difference. Let’s do the math: the current was 10 A, and let’s say the resistance was 0.2 Ω, giving 20 W dissipated in the die. Let’s say the air temperature was 30 degrees (the temperature of the warm heatsink). That gives us 30 + (20 × 3.6) = 102 degrees. But since the temperature was rising, the resistance was rising too. So at 100 degrees, let’s say the resistance was 0.3 Ω, which means the power was already 30 W, and the temperature would be 30 + (30 × 3.6) = 138 degrees. Entirely plausible.



Conclusions.
Personally, I think these are genuine transistors, but testing them at the specified current of 13.1 A isn’t valid because poor die-to-case heat transfer causes thermal runaway in the die. But the tenth unit makes things even more puzzling: it behaved quite differently, so the question is, why...?
Based on everything I can see, it looks like the 13.1 A test current is specified for transistors in the standard package. In other words, the die can handle 20 A, but the severely degraded die-to-case heat transfer significantly reduces the power it can dissipate.

That’s all for today. I hope you found this useful, and I’d also be very glad to hear your comments and thoughts about these transistors.