LM317 KCT vs KCS (thermal performance of KCS is better but not sure why/how)

[aiq25]

I was recently looking at the thermal performance of LM317KCT and I was looking at the RthetaJC. It looks like the KCT version is 4.2 degC/W while the KCS is 0.1 degC/W. I do not understand the difference because they are both TO-220. I have some samples of the KCT version. Is the KCS really only 0.1 degC/W? Also the datasheet doesn't mention (from what I can) there difference between KCT and KCS versions.

I guess I just don't understand why the thermal performance is different if the specs and packaging is the same. If anyone has any information and can share I would really appreciate it.

[magic]

There are a few variations of TO220, differing in thermal plate thickness and other details. The relevant datasheet(s) should show drawings and exact dimensions.

0.1C/W is suspiciously low for TO220 and may be a typo.

[aiq25]

KCS has a thicker and (very slightly) larger copper heat spreader in it.
Presumably, to take advantage of the larger heat spreader, the die is also larger.

0.1K/W is genuine, but you can't get that performance.

You can use a magic material (pyrolytic graphite) to boost lateral thermal conductivity to more than 20 times of copper, but if you spend that money, you'd be better off with a switcher.

So the bottom line, unless the heat spreader is very thick (comparable to die size), all Rjc values are just not practical.

Thanks. This is something I'm just experiementing with. I understand the 0.1 number is not realistic but just for comparsion it is a lot higher than the 4.2. I think I might try these two IC's out under the same conditions and see the thermal performance. I'm making just a simple power supply with the LM317 and there is a switch mode supply before hand as a tracking regulator.

[FriedLogic]

I would have thought that a value of 0.1°C/W at least is a likely a typo, but there are other interesting changes in the data sheets. The previous version from 2015  had 3°C/W for the KCS, but also had 1.84°C/W for the KTT, which is down to 1.1°C/W in the 2016 data sheet. The junction to ambient numbers are also lower.
The TI LM317A data sheet also has much lower values than the old NS data sheet or the 2018 ST data sheet.
How much of this is to do with differences in the parts, and how much to do with the testing is another question.

[magic]

1.1°C/W I could perhaps believe, it's in line with TO220 power trannies I have seen. Some FETs get it down to 0.2~0.3, but we are talking parts which cost 8 dollars more than LM317, not 8 cents

I still think it's a typo.

[duak]

I'm assuming this is the TI part.  I would bet it's a typo too.  I remember reading datasheets from shady companies written to promise 200% more performance than you'll ever get.  Maybe TI has developed a material with negative thermal resistance.

Building on Blueskull's replies, it looks like the LM317 dice could be packaged in either of two leadframes - look at the differences between the two options on the mechanical drawing pages.  One is TI's and the other is National's?

I ran across an app note from TI on defining and measuring the thermal chc's: http://www.ti.com/lit/pdf/spra953  No surprises here.

[aiq25]

I'm assuming this is the TI part.  I would bet it's a typo too.  I remember reading datasheets from shady companies written to promise 200% more performance than you'll ever get.  Maybe TI has developed a material with negative thermal resistance.

Building on Blueskull's replies, it looks like the LM317 dice could be packaged in either of two leadframes - look at the differences between the two options on the mechanical drawing pages.  One is TI's and the other is National's?

I ran across an app note from TI on defining and measuring the thermal chc's: http://www.ti.com/lit/pdf/spra953  No surprises here.

Yes, this is the TI part. Thanks for the app note, will look at it.