# EEVblog Electronics Community Forum

## Electronics => Microcontrollers => Topic started by: sci4me on July 28, 2013, 04:42:04 pm

Title: How hot is too hot for regulators?
Post by: sci4me on July 28, 2013, 04:42:04 pm
Hi guys. So, I have a complex design running on my FPGA dev board, and its running at 40 MHz. It gets the fpga to warm up a bit, but its not really hot. However, the voltage regulators? They practically burn your skin if you touch them for more than half a second! I dont have a proper way to measure the temp (unless the board has temp sensors that i dont know about) so, do you guys think it would be safe for me to bump it to 50 MHz? Or should I stay with 40?
Title: Re: How hot is too hot for regulators?
Post by: mariush on July 28, 2013, 04:55:49 pm
The amount of heat a regulator produces is based on how much current go through them and how well you heatsink them.

Going from 40 mhz to 50 mhz may only raise the power consumption by a few milliwatts, but you may already run the regulators much hotter than recommended.

Do the math and estimate how close you are to the maximum allowed.

(Vin - Vout) x current = power dissipation

Datasheet lists how much power the regulator can dissipate, by giving you the thermal resistance (junction) and/or with air or when soldered onto a particular area of copper which is supposed to  act as heatsink. calculate how much area you currently have as heatsink for regulators and estimate if you're stressing the regulator.
Title: Re: How hot is too hot for regulators?
Post by: Berni on July 28, 2013, 04:59:27 pm
Well it's a good idea to use switching regulators for Vcore on FPGAs because they can pull over an amp on it.

But it's very hard to blow a regulator, they usually keep working past 120 deg C and usually have an overtemperature protection in them. So if a drop of water does not boil on the regulator it's not going to throw in the towel just yet. However if you can fix it try to. In general I folow the rule of keeping everything cold enough to hold a finger on it because things getting hot are bad for long term reliability. It also gives lots of headroom in case you run it in a hot environment or put it in a place that does not have enough ventilation.
Title: Re: How hot is too hot for regulators?
Post by: sci4me on July 28, 2013, 05:13:13 pm
Okay, so there are two regulators which can take 125 C and one which can only take 85 according to its datasheet. As far as how much current im drawing, each 10 mhz is ~10mA. As far as how much of a heatsink they have, there are a few standoffs on the board which get pretty warm, and it seems like around half of the pcb is really effected by the heat. This is all based on my fingers so its not very accurate, but I think im just going to stick with 40 MHz for now.
Title: Re: How hot is too hot for regulators?
Post by: mikeselectricstuff on July 28, 2013, 06:13:07 pm
General rule for reliability - it shouldn't get too hot to touch.
If it's using linear regs from 5V to 3.3 and 1.2V then you may be able to reduce the input to maybe 4V to reduce dissipation.

If they aren't well heatsinked (sunk?) then maybe solder some extra copper foil, or replace with chunkier ones on a heatsink. Or as mentioned above, use a switching regulator.

Title: Re: How hot is too hot for regulators?
Post by: SeanB on July 28, 2013, 06:20:07 pm
Stick a fan to blow air past the board so it gets cooling. Do not blow air down on to the board but place the fan at one edge nearest the regulators so air flows across the board. That will cool it a lot better.
Title: Re: How hot is too hot for regulators?
Post by: ElectroIrradiator on July 28, 2013, 06:37:02 pm
It depends on the individual part on how hot is too hot. Many small semiconductor devices have a quite large Rth,jc, so the finger touch test may not be very accurate. Your finger enzymes go 'Ouch!' at about 54 degrees C, yet at that point it is anybody's guess how hot the Silicon actually is.
Title: Re: How hot is too hot for regulators?
Post by: flynnjs on July 29, 2013, 07:23:09 am
Why not put a small heatsink on the regulator to be safe?

Not sure if this is your own design or a dev board but a most
dev boards use switching regulators which don't usually get hot
when run within spec. You might want to investigate.

The other think that can cause regs to get too hot is if the have
gone unstable. Look to see if you have enough capacitance on
both the input and the outout according to the spec sheet.

The data sheet will tell you the thermal properties and you
should be able to work back the the junction temp from your
estimate of case temp.
See: http://en.wikipedia.org/wiki/Thermal_resistance (http://en.wikipedia.org/wiki/Thermal_resistance)

If it's not your own board it might be difficult to tell how good
a sink the PCB is.
Title: Re: How hot is too hot for regulators?
Post by: millerb on July 29, 2013, 08:17:03 am
Okay, so there are two regulators which can take 125 C and one which can only take 85 according to its datasheet. As far as how much current im drawing, each 10 mhz is ~10mA. As far as how much of a heatsink they have, there are a few standoffs on the board which get pretty warm, and it seems like around half of the pcb is really effected by the heat. This is all based on my fingers so its not very accurate, but I think im just going to stick with 40 MHz for now.

An extra 10ma draw probably won't make any bit of a difference. What dev board are you using? Or did you roll your own? Most reasonably designed boards will handle max draw from every device on the board with some extra room for insurance as well. If you have any extra circuits powered off of your board, that may be what's pushing you over the edge.

Understanding the thermal specs on a regulator seems to be one of the most difficult thing to grapple with regarding these devices. Lots of parts out there list nice specs in the datasheet but seem to want to catch on fire with mediocre loads. On a recent project I did a prototype with an LDO rated for 3.3V output @ 1.1A with up to 18V input. With 4V input and 300mA draw it hit over 65C and triggered thermal shutdown. It was a SOP chip with no heat paddle nor a heat pad on the underside. I probably could have seen that coming if I understood the thermal parameters in the datasheet.

I switched that out in my FPGA project for an Analog ADP5040 which has on chip one buck regulator and two LDOs. I used the buck to power the FPGA IO and the first LDO to power the FPGA core. Used the remaining LDO to power a microcontroller, spi flash, i2c flash and clock generator. With that regulator IC it never went 10C over ambient temp with 100mA load on each LDO and 400mA load on the buck despite only having one thin wire connecting the heat pad to ground.
Title: Re: How hot is too hot for regulators?
Post by: alm on July 29, 2013, 04:34:24 pm
It greatly matters which part it is that gets hot. Too hot to touch might be 60°C, a perfectly safe temperature for many semiconductors. So I would not be worried about a MOSFET on a heatsink that is too hot to touch (although I would measure it, 150°C is too high for most), but I would be very worried about an electrolytic cap being too hot to touch (since even the good ones might be only rated for a few thousand hours at 105°C). Also note that max power dissipation usually goes down with temperature (due to the thermal resistance from die to case, so the die is usually even hotter than the case).

I dislike the standard 'cooler is more reliable' rule of thumb. Reducing the temperature of a MOSFET from 200°C to 150°C will probably increase its reliability. From 60°C to 10°C probably not so much. I remember Bob Pease complaining about the standard factor too lower MTBF for every so many degrees increase in temperature rule-of-thumb.
Title: Re: How hot is too hot for regulators?
Post by: Corporate666 on July 29, 2013, 06:03:48 pm
Okay, so there are two regulators which can take 125 C and one which can only take 85 according to its datasheet. As far as how much current im drawing, each 10 mhz is ~10mA. As far as how much of a heatsink they have, there are a few standoffs on the board which get pretty warm, and it seems like around half of the pcb is really effected by the heat. This is all based on my fingers so its not very accurate, but I think im just going to stick with 40 MHz for now.

Thing is, 85C and 125C are junction temps... if the case is getting too hot to touch, it's a cinch that the junction is 20-30 degrees hotter or more.  You can really see how quickly temperature falls off with mass and distance when using a thermal camera.  The die of an LED can be way hotter than the heatsink attached to it.

The datasheet should specify Rth for junction-case, case-ambient and possibly some others.  You can measure current and voltage and figure out how many watts you are dissipating, then do some calcs to guesstimate your junction temps.  But if you don't want to do all that, I would definitely add heat sinking.  Too hot to touch means low reliability, IMO.
Title: Re: How hot is too hot for regulators?
Post by: amyk on July 30, 2013, 07:25:07 am
if the case is getting too hot to touch
"too hot to touch" may mean 60-70, and so the die inside be around 100, which is stil within spec if the max is 125. Power semiconductors in general aren't highly temperature-sensitive and will work just fine even if you can't keep your hand on them.
Title: Re: How hot is too hot for regulators?
Post by: David_AVD on July 30, 2013, 10:52:44 am
A regulator that runs too hot to touch is invariably running too hot for reliable operation.  For a linear regulator reduce the input/output differential.

Even better just change it for a switching type.  There are cheap (< \$5) 3 terminal 1.5A switching regulators these days.
Title: Re: How hot is too hot for regulators?
Post by: ElectroIrradiator on July 30, 2013, 12:46:57 pm
A regulator that runs too hot to touch is invariably running too hot for reliable operation.  For a linear regulator reduce the input/output differential.

Commercially manufactured heatsinks, intended for power applications and fanless/convection cooling, most frequently have their rated thermal resistance, Rth,sa, given for a particular temperature in degrees Celsius. This is because the thermal resistance of a convection cooled heatsink varies considerably with the sink/ambient temperature differential.

Do you know what this temperature is? ;)

(I believe this is what people call a trick question. :D )
Title: Re: How hot is too hot for regulators?
Post by: nukie on July 30, 2013, 02:02:23 pm
Consult datasheet.
Title: Re: How hot is too hot for regulators?
Post by: Corporate666 on July 30, 2013, 06:07:47 pm
if the case is getting too hot to touch
"too hot to touch" may mean 60-70, and so the die inside be around 100, which is stil within spec if the max is 125. Power semiconductors in general aren't highly temperature-sensitive and will work just fine even if you can't keep your hand on them.

What you say is true, but also keep in mind that on pretty much every data sheet I have seen, there is an asterisk under the "maximum ratings" that says that continued operation is not guaranteed at those levels, just that the device is technically capable of operating up to them momentarily.  So if the max junction temp is 125, I would not want to run it at 100 sustained... I think that's much too hot.

It also doesn't leave a lot of overhead for when the product is in the field in an non-air conditioned room in July, or when the fan is clogged up with dust, etc, etc :)
Title: Re: How hot is too hot for regulators?
Post by: nukie on August 01, 2013, 09:14:01 am
Open up a hp34401a, its very warm around the linear regulators, certainly not so good for the electro caps close to that area, so if you have electrolytic caps mounted close you might want to consider a redesign. Heating from components means power waste why not improve the efficiency?
Title: Re: How hot is too hot for regulators?
Post by: Hardcorefs on August 01, 2013, 09:28:52 am
There is a lot of nonsense talked about 'data sheets' and junction temperature....

125 deg. c is the 'absolute maximum',  it is NOT a pointer that you can reliably run the device at this temp or anything near it, generally it is the 'tits up' temperature at which point the devices reliability goes out of the window,

Basically it boils down to this.....
For every 10 deg. c over 25 deg. c, the life expectancy is halved.

This sort of result from an FPGA , can indicate poorly designed logic,  an error in the logic, OVERCLOCKING or an inadequately designed regulator.

One major point often overlooked is that as the DC/DC regulator heats up the voltage regulation drifts.. and 0v1 in the wrong direction can toast the FPGA.

If it is the DC/DC convertor , take a look at the inductors if they overheat things can go bad very quickly...

what sort of 'algorithm' would be designed to allow a bump in the CLK frequency, it's not a 'crypto-currency' is it?

Title: Re: How hot is too hot for regulators?
Post by: alm on August 01, 2013, 09:55:17 am
Basically it boils down to this.....
For every 10 deg. c over 25 deg. c, the life expectancy is halved.
Citation please. This might be true for some sensitive parts, but is absolute nonsense for semiconductors. You can't just expect the Arrhenius equation for a single chemical reaction to accurately model complex systems. And even then, how did you come up with the activation energy of whatever chemical reaction governs degradation of a part, and is it the same for all parts? I think even MIL-HDBK-217F, with its very simplistic model of temperature, is more optimistic than this. Bob Pease even argued that the whole higher temperature (up to 100°C or so) is less reliable is mostly nonsense in his book 'Troubleshooting analog circuits'.
Title: Re: How hot is too hot for regulators?
Post by: brainwash on August 04, 2013, 05:58:33 pm
I believe that half-life rule applies only to capacitors. I believe silicon devices are as happy running at 25C as they are at 80C but imho trying to keep a design running <50C is a good rule to avoid thermal runaways and minimize temperature drifting.
Also, if you design stuff for your home lab everything can be pushed but if the board is meant to sit outside or inside a car (-10 to 80C is normal) that headroom is shrinking away quickly.
Title: Re: How hot is too hot for regulators?
Post by: Hardcorefs on August 05, 2013, 08:45:51 am
Basically it boils down to this.....
For every 10 deg. c over 25 deg. c, the life expectancy is halved.
Citation please. This might be true for some sensitive parts, but is absolute nonsense for semiconductors. You can't just expect the Arrhenius equation for a single chemical reaction to accurately model complex systems. And even then, how did you come up with the activation energy of whatever chemical reaction governs degradation of a part, and is it the same for all parts? I think even MIL-HDBK-217F, with its very simplistic model of temperature, is more optimistic than this. Bob Pease even argued that the whole higher temperature (up to 100°C or so) is less reliable is mostly nonsense in his book 'Troubleshooting analog circuits'.

How about you do the legwork to find papers or research that negates the statement, which if you are right should exist in vast numbers.

Title: Re: How hot is too hot for regulators?
Post by: alm on August 05, 2013, 09:02:52 am
The burden of proof lies with the one making the claim. (https://yourlogicalfallacyis.com/burden-of-proof)

I gave two references already: MIL-HDBK-217F, which gives a much weaker effect than you quote (a factor of 1.2 increase in failure rate from 25°C to 35°C for voltage regulators), and Bob Pease that states that the number should be even lower, and gives another reference supporting his statement.

There's also the basic science of what the Arrhenius equation means and what its parameters represent.
Title: Re: How hot is too hot for regulators?
Post by: brainwash on August 05, 2013, 03:22:32 pm
http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/38512/1/03-2854.pdf (http://trs-new.jpl.nasa.gov/dspace/bitstream/2014/38512/1/03-2854.pdf)

So you get a guaranteed 20 years at 110C continuous, which could mean 80C at the package level.