EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: radensb on September 06, 2020, 11:28:28 pm
-
There are many topics on this, and I thought I had a good understanding - until I got questioned on a recent design that I though was fine.
I am designing a board for automotive application which will live behind the dashboard in the cabin. I am using a TI UA78M33IKVURG3 (https://www.digikey.com/product-detail/en/texas-instruments/UA78M33IKVURG3/296-41035-1-ND/5178912) linear regulator to drop the 14.5V (running voltage) to 3.3V. Initially I used a SZ1SMA5920BT3G (https://www.digikey.com/product-detail/en/on-semiconductor/SZ1SMA5920BT3G/SZ1SMA5920BT3GOSCT-ND/8538959) 1.5W 6.2V Zener in a SMA package to "pre-drop" the voltage to 8.35V for the regular. I did a thermal tests and got the following after an hour of runtime:
- Ambient Temp: 28C
- Temp of Zener with 100mA load: 70C
- Temp of regulator tab: 55C
The temp of the zener worried me a bit (a lot of heat for its size), so I shorted it out so that the regulator saw the full 14.5V. As a result, the regulator tab temperature rose to 65C, however I felt like this was ok because the regulator has a MUCH smaller Junction-to-ambient value of 30.3 C/W vs the zeners 250C/W(!!). I got feedback that the 65C that I am measuring its too high and that I should only tolerate a 15C temp rise. The regulator has a max recommended junction temperature of 125C and an absolute max at 150C.
Running the math, I get:
(14.5-3.3V)*0.1A = 1.12W power dissipation. 1.12W*30.3C/W = ~34C temp rise, which is only a few degrees off from what I am measuring. The regulator is an SMD part and the tab is soldered to the PCB GND plane (top layer of a 2 layer board), so my assumption was that the 30.3C/W was a worst case temp rise rate as the part would actually be using the board to sink some of that heat.
So, an operating temp of 65C is still well under the max recommended rating of 125C, right?
So my new plan was to reverse the zener to forward bias direction and allow it to drop 0.7V so the regulator would see ~13.8V and operate at ~65C.
Does this seem reasonable or am I missing something??
Thanks!
-
So my new plan was to reverse the zener to forward bias direction and allow it to drop 0.7V so the regulator would see ~13.8V and operate at ~65C.
Is it even worth it, though? The zener will only be dissipating 0.7V*0.1A = 0.07W -- just a tiny fraction of the 1.1W total.
I found this PCB thermal analysis calculator:
https://www.heatsinkcalculator.com/pcb-thermal-resistance-calculator.html (https://www.heatsinkcalculator.com/pcb-thermal-resistance-calculator.html)
1 oz. copper has a thickness of 0.0347mm. Plug in your PCB dimensions, click on the device tab and plug in the power dissipation, probably keep the \$\theta_{jc}\$ at 1.3 even though the datasheet doesn't list anything for the KVU package, and see what you get.
Another thing to consider is to use the TO-220 package as opposed to the TO-252. This will allow you to easily use a conventional heatsink if you feel you need it.
-
I would use some Zener at least for to catch possible high-voltage transients. With higher rated voltage than 8V - may be 9-10 V rated.
-
Hi
You mentioned that this is an automotive application, dashboard in fact.
The test you did shows that the regulator temperature is at 65°C.
Have you considered the environment that automotive has to work in?
Rule of thumb for circuits in enclosures is that ambient in enclosure is 10°C above outside enclosure.
Worste case for automotive is 50°C worstecase, consider car driving through Death valley.
Does a case temperature of 65°C sound reasonable now?
-
Even in the UK and similar temperate latitudes, ambient temperatures can hit 40 deg C, and in southern England typically do so at a few locations for a few days every few years. If you design to a 28 deg C ambient, expect a rash of summer failures across central and southern Europe, and North America south of the Great Lakes!
The cabin of a black car parked on a black asphalt car park on a hot still clear summer day will get considerably hotter than the nearby ambient measured away from the asphalt paved area, maybe as much as 40 deg C hotter, so your electronics could well be stating off in 80 deg C ambient before it dissipates a single watt! When the owner returns to their vehicle the electronics may have to survive functioning at that temperature for several minutes before the cabin cools sufficiently to remove the thermal stress. Meanwhile the owner is standing next to the running car waiting for conditions inside to become habitable . . . :popcorn:
-
So my new plan was to reverse the zener to forward bias direction and allow it to drop 0.7V so the regulator would see ~13.8V and operate at ~65C.
Is it even worth it, though? The zener will only be dissipating 0.7V*0.1A = 0.07W -- just a tiny fraction of the 1.1W total.
Yeah, your right. At that point, the diode would be there just for reverse voltage protection. I figured that its voltage drop would only help matters with power dissipated by the reg. Thanks for the resources. Ill check that calculator out! I cant go with a through hole part and adding a heat sink is undesirable. I was looking at the surface mount style heat sinks, but it adds height that I wanted to avoid.
I would use some Zener at least for to catch possible high-voltage transients. With higher rated voltage than 8V - may be 9-10 V rated.
Yes, I am actually doing that. I have a 19V TVS shunt right at the voltage input pin of a connector to GND. The car has a running voltage of 14.5V, so it souldnt be conducting anything unless one of those nasty alternator spikes comes through.
Hi
You mentioned that this is an automotive application, dashboard in fact.
The test you did shows that the regulator temperature is at 65°C.
Have you considered the environment that automotive has to work in?
Rule of thumb for circuits in enclosures is that ambient in enclosure is 10°C above outside enclosure.
Worste case for automotive is 50°C worstecase, consider car driving through Death valley.
Does a case temperature of 65°C sound reasonable now?
Even in the UK and similar temperate latitudes, ambient temperatures can hit 40 deg C, and in southern England typically do so for a few days every few years. If you design to a 28 deg C ambient, expect a rash of summer failures across central and southern Europe, and North America south of the Great Lakes!
The cabin of a black car parked on a black asphalt car park on a hot still clear summer day will get considerably hotter than the nearby ambient measured away from the asphalt paved area, maybe as much as 40 deg C hotter, so your electronics could well be stating off in 80 deg C ambient before it dissipates a single watt! When the owner returns to their vehicle the electronics may have to survive functioning at that temperature for several minutes before the cabin cools sufficiently to remove the thermal stress. Meanwhile the owner is standing next to the running car waiting for conditions inside to become habitable . . . :popcorn:
So, I did think about this and this is where I thought I was ok. Because the datasheet states that the max recommended junction temp is 125C and I am measuring a 36C temp rise at the tab (which is as close to the junction temp as I can get), I figured that I could still operate under the recommended temp of 125C as long as the area behind the dash never exceeded 89C while the regulator was regulating (which seemed like a very safe bet). Even if the cabin did get that hot while sitting, the driver would quickly get it cooled down to a tolerable temperature, thus if the regulator did power up in an ambient 89C cabin, it would not be operating in that environment for long (windows rolled down and AC blasting!). I figured that 40C behind the dash was a good upper bound while driving as it is still in the temperature controlled cabin. 40C + 36C = 76C under normal operating conditions, which is ~50C under the max recommended temperature and ~75C under the absolute max temperature (150C). So when people talk about hot to the touch needing additional heat sinking, I get confused because that's way under the limit.
Thanks for the help!
-
You've already calculated that the junction runs 34 deg C hotter than the tab in your application, and measured a temperature of 36 deg C above ambient at the tab. That's a Tj of 70 deg C above ambient. To keep Tj under 125 deg C, you cant tolerate an ambient of over 55 deg C, which can easily be exceeded in a closed unshaded car's cabin on a calm sunny hot day. Also, how much temperature rise above local ambient does your device's enclosure and mounting location contribute by restricting air circulation? If your bench test was repeated in-situ, you could well see a greater temperature rise. You *DON'T* want to be in a race between how fast the car cabin cools down and the inside of your enclosure heats up from its own dissipation, the 'winner' determining whether or not the max. junction temperature is exceeded.
-
Hi
In Engineering, you need to make sure that thing works in ALL enviromental conditions, particularly the extremes. This is the difference between a hobby project and a commercial product.
Automotive is considered safety critical, so the system have to operate under all conditions to make sure users will be safe. If a system fails, the other system need to compensate for the failure. It is better that no system fail.
You failed to notice what has been said about electronics in enclosures - the inside of the enclosure is hotter than the outside but the electronics in the enclosure has to be operate at that hotter temperature!
-
Do you really need to use a linear regulator? It's a pretty big drop from 13-15v down to 3.3 to do with a linear reg and as you've discovered means you need to deal with a modest amount of heat to get rid of, in what sounds like a confined space with little airflow.
A little switch mode DC-DC buck converter the size of your fingernail could do this.
-
Before you continue the design you definately need to know the requirements for the supply as there are : Load dump-, reverse polarity- and cold crank specs. A discrete design rarely makes sense when it comes to reliability considerations (FIT rates).
You'll find parts designed for that purpose which also comply with AEC-Q100 which is mandatory in most cases. After market systems may not require this though...
-
A little switch mode DC-DC buck converter the size of your fingernail could do this.
EMC could become a nightmare in automotive applications...
-
Hi
You mentioned that this is an automotive application, dashboard in fact.
The test you did shows that the regulator temperature is at 65°C.
Have you considered the environment that automotive has to work in?
Rule of thumb for circuits in enclosures is that ambient in enclosure is 10°C above outside enclosure.
Worste case for automotive is 50°C worstecase, consider car driving through Death valley.
Does a case temperature of 65°C sound reasonable now?
I would allow for a worst case ambient temperature of 70°C under the dash. This figure was what I used when designing mobile radio equipment for installation in cars.
I have measured temperatures in excess of 85°C on top of the dash of a car parked in the sun and that car wasn't even located in Death Valley.
-
Taking all the feedback to heart, I have decided to bite the bullet and replaces the 6.2V zener with a 600mA buck converter. I am using the TPS560430-Q1 (https://www.ti.com/lit/ds/symlink/tps560430-q1.pdf?HQS=TI-null-null-digikeymode-df-pf-null-wwe&ts=1599462934383) which is automotive certified. I plan to set the output voltage to 6.1V which will then go to my main 3.3V linear regulator and a 5V linear regulator I am using to drive the CAN transceiver. The linear regulators will drop and help clean up the power for the final loads. This approach will only costs an additional 6 parts, and will provide for an efficient step down from 14.5V. I ran some initial thermal tests on my dev boards feeding the linear regulators the 6.1V I am planning for and measured only a 8C temp rise on the 3.3V linear reg and 5C temp rise on the 5V linear reg. Muuuuuch cooler!
-
Make sure that you use the FPWM version. If the PFM version sees light load and the switching frequency drops down into the am-band you'll find yourself in hot water...