In this final section, I'll be comparing the HotRod package to the traditional QFN package. This comparison is being done using the same silicon in both packages, and the performance is evaluated. The part used for this evaluation, again, is a TPS 54824. And it was put both in the released HotRod package and a prototype QFN package.

For this performance comparison, I'll be comparing the power loss of the IC, the effective theta JA, and the temperature rise-- the total thermal performance of the IC-- and the switch node characteristics. As a QFN package will have bond wire, so there will be much more ringing on the switching node. Also, the bond wires add DC resistance. So we expect to see more power loss in the QFN package.

Here's the power loss comparison between the HotRod and QFN package. So if you look over to the curve on the right, we have the efficiency in dark gray and blue and the power loss in light gray and red. So the HotRod, if you look at the full 8 amp load, it has about 88% efficiency. And the QFN and has about 87% efficiency. So there's a 1% improvement in efficiency using the HotRod package. And that 1% improvement efficiency translates to about 0.2 watt difference in power dissipation.

In this power dissipation, it's for the entire buck converter. So this includes the dissipation in the inductor. The other thing I should mention with these measurements, is it's taken with a pre-released version of the IC. Before we released the silicon, we did make some improvements to the efficiency. So in the released version of the IC, you'll get a little bit better efficiency than what's shown in this curve.

Here are some estimations that show why the QFN package has higher power loss. And it's due to the bond wires. So if we go through the calculations of the resistance of the bond wires, we'll see that the bond wires add about 1 milliohm of resistance to the switch pin, 1 milliohm of resistance to the VN pin, and about 0.7 milliohm to the P ground pin. So this increases the high side already assigned by about 2 milliohms in the low side already assigned by about 1.7 milliohms. And again, these measurements at 8 amp show that the QFN package has about 0.2 watts more power loss than the HotRod package.

Here's the effective theta J comparison between the HotRod and QFN package. This test was done using the method where there's dissipation in the IC only. So that's a little bit more controlled experiment. On the left, we have the HotRod Rev A EVM that gets about 25 degrees C per watt. And in the middle, we have the QFN EVM that had about 21.4 degrees C per watt.

These two are used for the main comparison, because the main difference is the QFN package has the thermal pad with vias in it. And both parts only have about 1 via on each side adjacent to the package. So you can see here that the HotRod EVN does have about 3.6 degrees C per watt [INAUDIBLE] theta JA.

But I need to mention that with the HotRod EVM, if we have a little bit better placed vias over on the right, we can get very similar theta JA to the QFN. It's not quite as good as a QFN, but it's pretty close. It's only 0.9 degrees C per watt difference. Again, I'm comparing the first two mainly because it is a good comparison to show what the effect is of that adding the thermal pad to the IC.

Next, I looked at what the temperature rise of the IC was when it was operating at a power loss. So if you remember from two slides ago, the QFN had out about 0.2 watts more of power loss. But then, if you look at the case temperature rise of the two ICs-- the HotRod on the left and the QFN on the right-- the QFN actually ends up with a little bit lower case temperature. It's about 0.8 degrees C lower than the HotRod EVM.

So even though QFN has a little bit more power loss, it's able to get very similar thermal performance to the HotRod part. And again, this has taken on the pre-released version in the IC. So in the final version, it actually will operate about 4 degrees C lower than what these measurements show.

So again, adding the thermal pad to the QFN package allows it to get about the same thermal performance as the HotRod, even though it has more power dissipation. I also need to mention that if we look at the Rev B EVM-- I don't have it here-- but the Rev B EVM, the case temperature of the HotRod would be lower than the QFN. The extra power dissipation in QFN EVM brings its case temperature above the HotRod Rev B EVM.

Next, I look at what the switch node behavior was between the HotRod and the QFN package. So with the QFN package, due to the bond wires, there's extra inductance in the switching paths. We would expect this to cause more ringing at a lower frequency when there's more inductance.

And in fact, we do see that. It might not be as big of an effect as you would expect. But we do see that the QFN has about 1.2 volts more overshoot on the switching node than the HotRod part. The other thing you can see, if you look at the settling time of the ringing, the QFN takes longer to settle out due to the lower ringing frequency. And this is again a result of the added inductance due to the bond wires.

So overall, the HotRod EVM, we would expect it to have lower noise than the QFN. Because it has less ringing on the switching node. The lower ringing in the HotRod part can also enable us to get lower switching loss in a HotRod part. So when there's less inductance and lower ringing, we can drive the MOSFET harder to get the same overshoot at the switching node. When we drive the MOSFET harder, there's lower switching loss and higher efficiency.

So the HotRod not only gives us lower conduction loss than the QFN, but can also get lower switching loss than the QFN. So overall, the HotRod gives a really good efficiency improvement over the QFN.

I also looked at the falling edge of the switching node to see how the HotRod compares to the QFN. The difference wasn't as big here. You can see that there is slightly less undershoot on the HotRod part but not that much. And then, if you look at the period of the ringing, it's less than three nanoseconds for the HotRod and a bit closer to three nanosecond for the QFN.

So again, the effect of the inductance is slightly more undershoot with the QFN and a slightly lower ringing frequency for the QFN. Again, it just kind of correlates to what we would expect to see.

This table summarizes all the results of the comparison done between the performance of the HotRod and the QFN package. So again, if we look at the IC power loss when it's working as a buck converter-- here's the actual conditions it was tested in-- the HotRod part as 0.2 watts lower power dissipation than the QFN. Next, if we look at the thermal resistance of the theta JA, the QFN does have better thermal resistance due to the thermal pad that's in the middle. This is a really effective way of getting head out of the IC and into the PCB.

So if you look at three Rev A EVM, it's about 3.6 degrees C per watt higher thermal resistance. But then, if we look at the rev B EVM, it's a little bit closer to the QFN at about 0.9 degrees C per watt. This kind of shows how important it is to have some thermal vias with a HotRod part, so you can get the best thermal performance out of it.

Now, if we look at the case temperature rise, which is really what you care about in the end, even if the HotRod has a higher theta JA, as long as the temperature rise is lower, that's good. So if we look at the rev A EVM, it was higher than the QFN by only about 0.8 degrees C. But then the Rev B EVM was about 3 degrees C cooler than the QFN. So the Rev B EVM and the HotRod, overall, the HotRod gets better temperature rise than the QFN.

And in the next two, they're comparing the switch node characteristics. So again, the HotRod, there's less overshoot on the switching node, about 1.2 volts difference. And due to the inductance of the QFN, the ringing frequency is lower. Because it depends on the capacitance and the inductance. So it causes a longer settling time with the QFN. And for the HotRod, the settling time is about 28 nanoseconds. So we expect this to translate to lower noise with the HotRod package.

The other thing to point out here is the HotRod allows us put a larger piece of silicon in the same package size. So the HotRod package is a 3.5 by 3.5 millimeter package. But to get the same silicon in a QFN package, we had to put in a 4.0 by 4.0 millimeter package.

So to summarize, the HotRod package has better efficiency, switch node ringing, and a smaller size than the QFN package. So we're putting these parts in the HotRod package for good reason. This is the way to maximize the performance of the buck converter. The QFN, it can have better thermal performance, because it has the thermal pad where you can put thermal vias. This is just something that our typical HotRod packages don't have.

But with a HotRod part, you can get QFN-like thermal performance if you have well placement of thermal vias near the IC. And this is shown by the difference between our Rev A EVM and our Rev B EVM. So even though the HotRod gets better performance than the QFN, there are customers who are hesitant to change this new package technology. But really, I would hope people to consider this new package technology, because it does give better performance overall.

And if you really want to maximize your performance, you need to use the best package that's available for the part. We have also seen some applications where there's low-cost PCB fab design rules. And they're able to put thermal vias in the pad. But they can't put vias close enough to the IC to get as good thermal performance with the HotRod.

In this case, the QFN might be a little bit better. But the HotRod can get pretty similar performance to the QFN as shown with a comparison between our Rev A HotRod EVM and this QFN package with the thermal pads and the via. We have also seen a requirement in an application where the pins need to be all the same size for the package. So the HotRod, almost all of them, they don't have the same size pins. So I guess in this case, you might need to use a QFN over the HotRod package.

The last thing I need to mention is that the comparison done here is somewhat specific to the TPS 54824 family of parts. And that's because most HotRod packages are unique designs. And they're designed specifically for the silicon.

So the HotRod package, it's thermal performance can vary a lot. But we're always doing our best to get the best thermal performance of the HotRod packages we design. So I think in most cases, it will be pretty similar, where the HotRod will definitely get lower power dissipation. It may have a little bit higher thermal resistance. But the overall case temperature rise would pretty similar between the HotRod and the QFN.

But the package technology and the silicon process are tied closely together. So as our silicon process is improving, the package technology also has to improve so that we can get more current out of a smaller package. And these new silicone processes are also reducing power loss. So as we reduce power loss, the higher thermal resistance of the HotRod package isn't as critical as it was before.

So when you're trying to get better thermal performance or a lower case temperature to rise, you can do one of two things. You can reduce the power loss, or you can improve the thermal resistance of the package. Both of them will translate to a lower case temperature rise. Ideally, we will do both. But to get the best performance when it comes to efficiency, switch node ringing, and everything, the HotRod package is usually the best choice.

And as I mentioned, the HotRod package is also evolving, so that we can get more current out of these smaller packages, where we're putting smaller silicon inside these packages. So again, the HotRod, overall, it's better performance than the QFN. And this is why all of our new parts are typically using this HotRod package.