A little question about an Adjustable Supply

General project help for Adafruit customers

Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.
User avatar
stinkbutt
 
Posts: 593
Joined: Wed Feb 17, 2010 2:40 am

Re: A little question about an Adjustable Supply

Post by stinkbutt »

We're doubling the input voltage. The power dissipation is going up, especially for a linear regulator that at the end of the day, is a glorified voltage divider with some transistors which are acting a lot like resistors anyway. So: If the voltage doubles, either the drop doubles, or the current doubles, or both. There's no way around this thing dissipating more heat, and I can't see any math that results in less then double the power dissipated.

And I was being a dick about the Stockholm thing. I shoulda just posted the above paragraph. I'm sorry, that was wrong. I was wrong.

(I hate non-apology apologies. I think I avoided it there.)

User avatar
ImaginaryAxis
 
Posts: 267
Joined: Sat Nov 13, 2010 3:16 pm

Re: A little question about an Adjustable Supply

Post by ImaginaryAxis »

[quote="stinkbutt"]We're doubling the input voltage. The power dissipation is going up, especially for a linear regulator that at the end of the day, is a glorified voltage divider with some transistors which are acting a lot like resistors anyway. So: If the voltage doubles, either the drop doubles, or the current doubles, or both. There's no way around this thing dissipating more heat, and I can't see any math that results in less then double the power dissipated./[quote]

This is true, if and only if ,the voltage output of the regulator stays the same - say 5V, when the input is doubled. The load current does not have to change either - the load will pull whatever is needed. Specific conditions must be clarified and simply saying "doubling the input voltage" does not necessarily mean the power dissipation doubles - which was my point. However, this is true for quiescent power, but I will put on my Boo Hat™ for that since I often ignore quiescent power (as it is usually very small) when calculating efficiency measurements.

As for the OP, he could also flood the top and bottom of a 2 layer board with 2oz copper. Use vias to connect the tab of the package down to the bottom plane. If he is going to use the PCB as a heat sink - a much more elegant approach IMO.

Seriously, this is a case of you saying "Bellis perennis" and me saying "Daisy". No need to get uppity -- I am sure I will require your assistance when I get my Arduino for Christmas. I vaguely recall what a FOR loop is. 8)

User avatar
stinkbutt
 
Posts: 593
Joined: Wed Feb 17, 2010 2:40 am

Re: A little question about an Adjustable Supply

Post by stinkbutt »

This is true, if and only if ,the voltage output of the regulator stays the same - say 5V, when the input is doubled. The load current does not have to change either - the load will pull whatever is needed. Specific conditions must be clarified and simply saying "doubling the input voltage" does not necessarily mean the power dissipation doubles - which was my point. However, this is true for quiescent power, but I will put on my Boo Hat™ for that since I often ignore quiescent power (as it is usually very small) when calculating efficiency measurements.
You gotta heat sink for worst case. If the regulator's capable of spitting out 1.25V with 18V of input, it's eventually going to be called upon to do that. Besides, the load voltage and current aren't the only source of power dissipation. (In fact, I don't think they're even the most significant one for this thing.)

As you said, there's quiescent current. There's also adjust current. The middle terminal isn't a ground terminal, it's a terminal pinned to 1.25V less than the output terminal. Whatever the adjust current is, it's going to be going across a 1.25V drop.

The internals of this thing are weird as well. There are a gagillion transistors, and apparently most of them are operating in the linear range. And four of them are generating this 1.25V reference voltage using, I dunno Boltzmann's constant or some BANNED. Point is, I don't think this thing can be simply modeled as a garden-variety voltage divider. If it were, it's power dissipation would be massively higher.

And none of this matters. You gotta heat sink for worst case. Frankly, there's a good chance the stock heat sink adafruit includes in the kit would be all this guy needs, because they probably over-engineered the hell out of the breadboard power supply kit and it'd be able to handle twice the input voltage on it's own, but if over-engineering is good enough for them, it's good enough for me.

User avatar
ImaginaryAxis
 
Posts: 267
Joined: Sat Nov 13, 2010 3:16 pm

Re: A little question about an Adjustable Supply

Post by ImaginaryAxis »

stinkbutt wrote:You gotta heat sink for worst case.
Let us walk through the math to see what 'worst case' is - so let's dispel with most of the hand-waving that has been going on.

The regulator specified in the kit is the LM317T, TO-220 package. According to the specification tables the typical output current for a differential output voltage that is <=15V is 2.2A. If the differential voltage is higher than 15V then the current is actually limited. It is ~1.8A for 17V... or to be exact 16.8 (Vin-1.2Vout(min)).

Power dissipation is:
(Vin-Vout)*I_load + (Vin*I_gnd) where I_gnd is not specified. Most likely a sum of Icc and the adjust current which is stated to be 100uA maximum.
[(18V - 1.2)*1.8A] + [(18V *(100uA+10mA)] = ~30W

The maximum temperature rise allowed for 30W is:
Trise = Tj(max) - Ta(max):
125°C - 25°C = 100°C
This is a gross assumption because the ambient temperature will rise when dissipating power, so for fun let's ignore that for now.

The required thermal resistance junction to ambient is:
Trise/P(diss)
100°C/30W = 3.33°C/W

The heatsink requirement is:
T(hs) = Trja - (Trjc + Trch) - let's ignore the thermal contact resistance as I do not see it specified in the heat sink datasheets.
T(hs) = 3.33°C - (4°C/W) = A lousy stinking negative number.

This means a perfect heatsink (does not exist) cannot handle worst case - keep in mind we ignored ambient temperature rise, quiescent and adjustment currents, and contact resistance of the heatsink to the package. Furthermore in a real system there will be other devices dissipating heat and they influence one another! In reality, for this to actually work there would have to be cold air convection along with a heatsink. I suggest moving to the Arctic Circle.

Am I being anal retentive? Absolutely. I understand that for hobbyists it may not be critical to derive equations, but thermal performance is critical to understand. Specific operating boundaries must be stated in order to know what is actually required.

Locked
Please be positive and constructive with your questions and comments.

Return to “General Project help”