4
\$\begingroup\$

I'm considering to use UA78M33CKVURG3 in my design, so as with any linear regulator I have to make sure it will be able to handle the current draw in the worst case scenario. To calculate what the temperature increase I'm using this article as a guide, which requires the Case Thermal Resistance value for the calculation, which is omitted in the datasheet for my package of choice.

enter image description here enter image description here

Why are some values missing? Is it because they are close to 0 or smaller than 1?

\$\endgroup\$
1
  • \$\begingroup\$ I'd suggest that if an educated guess isn't good enough in practice then you are probably running the part too close to be safe in real world applications. Rtca is liable to be dominant except in very low dissipation applications. \$\endgroup\$
    – Russell McMahon
    Mar 5, 2023 at 9:41

2 Answers 2

5
\$\begingroup\$

Why are some values missing? Is it because they are close to 0 or smaller than 1?

They're probably missing because they have not been characterized.

You have to realize this is literally a semiconductor design from 1976 (!!!), so compared to modern regulators, it's … not very good, nor very cheap.

The people that buy it in a modern package probably do so because they have some sort of existing circuit design that's been proven to work, and they need to reproduce it for a newer respin of their device, as unchanged as possible. That can happen to you in a regulation-heavy environment such as aerospace.

Allowing myself some more speculation:

These are the same people that need to thermally qualify their products anyway, so that the measurements for the KVU package variant, which quite possibly only exists because some larger customer asked TI for exactly that chip in that package, simply never got any priority, and then never happened.

\$\endgroup\$
5
  • \$\begingroup\$ Makes sense, didn't noticed it was from 1976! I'm considering this part because it has the lowest price in single quantity on mouser at 0.6€. Is it reasonable to use the KTP value (1.4) in my calculations? They have very similar packages, so I assume similar °C/W value. \$\endgroup\$
    – StefanoN
    Mar 4, 2023 at 12:04
  • 2
    \$\begingroup\$ Ha! I've been in that situation, especially as a student, where +20€ on a BOM seemed like a big deal. When spinning a single board of anything, don't go and save 30ct on your regulators; your whole design working only on the second attempt will instantly erase all benefit you had from the cheaper parts. Here, the problem, taking a wild guess, is that you have a high input voltage, right? \$\endgroup\$ Mar 4, 2023 at 12:07
  • 1
    \$\begingroup\$ (it might be reasonable to do that, indeed, but I simply wouldn't do it. Too much risk for too little gain) \$\endgroup\$ Mar 4, 2023 at 12:08
  • \$\begingroup\$ makes sense, imagine a client asking you about this and replying with "I assumed". Indeed I have 12V in 3V3 out @ 500mA worst case scenario. I think I will go with one of those linear-compatible all-in-one-package switching regulators mentioned at the end of the DigiKey article I mentioned, like K7803-500R3, or the 'naked' kind like K7803-500R3-LB to save some money \$\endgroup\$
    – StefanoN
    Mar 4, 2023 at 12:29
  • \$\begingroup\$ yeah, but if you had a client, you would be paid for your design time, and you would not care whether your single-unit prototype was 30 ct more expensive, either ;) Shaving off cents on your components makes sense for large runs, not prototypes, generally. \$\endgroup\$ Mar 4, 2023 at 12:31
8
\$\begingroup\$

It's an SMT package so junction-to-case thermal resistance is not going to have much effect on junction-to-ambient under normal conditions. As Marcus points out, this is a mature part and was designed before things like aluminum core boards were easily available. The amount of copper on a normal FR4 board will have a much bigger effect. With SMT parts, junction-to-ambient thermal resistance is very much dependent on assumptions about the board copper etc., so 10% effects are not very important.

If you want an approximate number, the LM78Mxx series has a better characterized set of numbers, and the number for the TO-252 is approximately 4.4°C/W. That datasheet also gives thermal impedances to the top of the package.

There is absolutely nothing wrong with the 78Mxx series, they are used in mass volume and are an extremely reliable and stable part if you don't mind the headroom and the 5mA quiescent consumption. Hence the low cost and enormous volumes available off the shelf. The 78xx series, on the other hand, does not fill much of a need since higher currents should be using switching power supplies.

The main issue with the 78M33 in particular is that the headroom required is too high to reliably use 5V input, at least at higher output currents. Hence, if that describes your application, you might be better off using another jellybean part such as the LM or AMS1117, which isn't quite Low Drop Out, more like medium. But like most LDO-ish regulators they are rather fussy about the output capacitor value and ESR and can oscillate under some conditions of load current, input voltage, temperature etc. if those are not within the proper range.

Some datasheets for older parts do not adequately warn the designer about use of ceramic capacitors- if the datasheet does not explicitly address use of ceramic capacitors on the output, assume it will need a resistor in series if a ceramic capacitor is used (a few ohms in order to degrade the ESR to more like that of a tantalum electrolytic capacitor).

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.