Resistor which is thermally coupled to the board?

[Korken]

Hi all!

I am working on a board which has an IMU that is planned to keep at some steady temperature above ambient, and to test the temperature controllers I am going to make a board with a few resistors and and IMU on an thermally isolated island within the PCB.
But I want the power dissipated in the resistors to dissipate into the ground plane of the island, which will be used as the heat spreader, how would one accomplish this?
First thought was to have resistor placed upside down and have exposed copper touching the upside, but for future manufacturing I am unsure how much the assembler will like that, plus the thermal transfer in not optimal.

Are there any better way to do this?

I am going to dissipate (at maximum) 1 W via 4-8 resistors (depending on the heat profile the island in the board gets) that are 0603 size.
This will be kept inside a small plastic housing on the PCB to isolate somewhat to start with, and to protect from airflow.
Are there special resistors/heating elements for this that I have not found?

Thanks for all comments and help!

[daqq]

Are there special resistors/heating elements for this that I have not found?

There are bigger resistors, such as those in DPAK packages that achieve a better thermal coupling. It all depends on your size constraints. Or alternatively if you want maximal contact and high flatness, there are extra wide.

DPAK: http://uk.farnell.com/vishay/dto025c470r0jte3/res-thick-film-470r-5-25w-to-252/dp/2471607
Wide: http://uk.farnell.com/panasonic-electronic-components/erjb3bfr22v/resistor-wide-term-0r22-500-mw/dp/1750972RL

[Korken]

Thank you for the tip!
Sadly I have size constraints which keeps me to 0603 resistors, or around that size.

[Korken]

Perhaps putting dabs of thermal epoxy on the resistors to copper, which is connected to the heat spreader, to act as thermal transfer?

[Tomorokoshi]

How about 2 3/4W 2010's. Depending on the circuit if one side is ground you can then stitch an extended pad to the ground plane using extra vias. Can you post the schematic of the area and some preliminary board dimensions?

What power rating and tolerance of your preferred 0603's?

When enclosed the power rating will need to be derated.

[Korken]

Currently there are no size constraints other than about 10 by 8 mm area (where the IMU needs 5 by 5 mm in the center, including safety distance), but the smaller the island is the easier it is to control its temperature with less power.
The 0603 I looked at are standard 1/4 W resistors on Digikey. Generally it should take small amounts power to keep the temperature, but some more will be needed in the startup phase (1 watt is probably way too much, but I can always reduce it).
The wattage rating of the resistors are for air convection, but with a thermal path to the heat spreader the resistors' power rating should be quite a lot higher.

[daqq]

Well, aside from standard solutions there are exotic solutions - you can use PCB embedded components, you can use a whole layer to act as a "resistor" - just a long, slender squigly line, or you can use print on carbon traces as resistors.

Then there are these fellas: http://www.vishay.com/docs/20043/crcwhpe3.pdf
SMD resistors with both sides of the substrate used as the resistive element. They should be able to dissipate the heat better.

[DaJMasta]

If you're just heating an oven, you don't really need direct thermal transfer, right?  You'd actually want most of the heating to be coming from indirect transfer to minimize the spike in temperature of the IMU when the resistors turn on.  So long as you're sealed in a box, you probably want indirect and even spaced heating elements and a the thermal control feedback sensor to be right next to the device requiring stable heat.  It will take more time to get to temperature that way, but it will probably be easier to control.

[Korken]

ddaq:
Indeed, those are the resistors I was looking at.
Seemed to be quite sturdy, so I think I will make a test board with those first and see how well it works.

DaJMasta:
Indirect heating has good and bad parts.
The good is that the heating is, as you mentioned, slow -- but the phase impact (the time it take for heat in the resistor to go to air and to IMU) can be quite significant, adding instability to the control if no model based approach is applied (LQR / MPC / slow PID).
My idea was to do both, heat up the PCB surrounding the IMU, to get a low heat gradient over the IC, and to heat the air as well to use as a slow buffer. This should allow the use of a simple PID since resistor -> board -> IC should go fast. The good part is that the IMU has build in temperature sensor, measuring the die temperature that I will use.
Not sure if the idea is sound without testing or simulation though

Thanks for all the tips!

[David Hess]

For assembly reasons, I would first try standard resistors but encapsulate them with thermally conductive epoxy after soldering and cleaning.

A higher performance solution would be to use SOT-89 or SOT-223 transistors or even larger ones as heaters since their collector tab is soldered to the board for good heat transfer.

[jbb]

I suggest you have a crack with an array of standard SMTs.  You can carpet bomb the bottom of the plane with them so that the power in each resistor (and there fore temperature difference across the resistor body) is low.  Your solder joints will be quite effective at getting the heat out.

Also (if you have access to a thermal camera), you can experiment with having a lower density (i.e. fewer resistors) in the middle of the plane to avoid a hot spot in the middle.

[wine+dine]

Have you considered using transistors instead of resistors as heating elements?
You can solder the collectors of PNPs to the ground plane for good thermal transfer.

[Korken]

Thank you for the comments!

David Hess:
I see you are on the thermal glue path as well.
For the transistors, I have not thought about that yet.
What directly strikes me is that I need some form of feedback for a constant current driver based on the transistors for regulating the power, or did you have something different in mind?
I like that, as it allows the controllers to have more granularity over the temperature control and the heat transfer properties of transistors are generally quite good.

jbb:
Ahh, indeed! To use a lot on the other side of the board could be a good solution as well!
I was thinking that perhaps waste heat would go to the surroundings or I would have to encapsulate both sides to keep the heat contained.
Though, I will make a prototype of this as well.

And yes, I have a heat camera. 

wine+dine:
Same direction as David I see.
Would you also go with some form of constant current source here?

[babysitter]

I successfully used a moore/hilbert/peano-curve-inspired Cu-trace PCB heater to keep a sensor snuggly warm.
As Cu has a decent tempco of resistance i was using it as the temperature sensor for itself.

babysitter

[Pjotr]

Yes, but for a substantial Cu-resistance in a tight space you need very thin traces which are difficult to manufacture accurately.

M2cents: Use several 0603 arrays on the back of the pcb island. They have sufficient solder area to conduct heat. Then lots of via's and 0402/0603 PT1000 thermistors near IMU. Is this a 2-layer board or 4-layer? 4 layers are of an advantage here and 2 oz copper better than 1 oz.

[Korken]

I successfully used a moore/hilbert/peano-curve-inspired Cu-trace PCB heater to keep a sensor snuggly warm.
As Cu has a decent tempco of resistance i was using it as the temperature sensor for itself.

This would indeed have been the best, but I am limited to 0.15 mm traces sadly.
Using 10 x 10 mm and 2 layers I was able to get a resistance of 0.8 ohms which sadly is too low.
I am aiming of using maximum of 0.2 A @ 5V.

Pjotr:
I have a 4 layer board with copper thickness as 1 oz | 0.5 oz | 0.5 oz | 1 oz
With 0.127 mm minimum trace, but I try to stay on 0.15 mm.
It is the 4 layer specifications of OSHPark which I use for the prototyping.

[David Hess]

David Hess:
I see you are on the thermal glue path as well.

The idea is just to lower the thermal resistance between the resistors and printed circuit board and adding dabs of epoxy after assembly is inexpensive.  Given the low cost of 1 watt and larger surface mount resistors though in reversed case sizes which will have better coupling anyway, maybe it is not worth it.

For the transistors, I have not thought about that yet.
What directly strikes me is that I need some form of feedback for a constant current driver based on the transistors for regulating the power, or did you have something different in mind?
I like that, as it allows the controllers to have more granularity over the temperature control and the heat transfer properties of transistors are generally quite good.

Check out figure 3 of National application note 256 and figure 4 of Linear Technology application note 5 where this is done to regulate the temperature of an IC.  A low value resistor is placed in series with the emitter of the transistor to ground so a low current control voltage at the base can adjust the collector current and thereby the power dissipation.  The advantages of using transistors is that except for the base drive, *all* of the power dissipation occurs at the heater instead of the linear drive circuit and transistors have low thermal resistance compared to most resistors.  I have seen transistors used as heaters this way in temperature controlled ovens for circuits before.

I do not know if it would be worth the extra complexity but the temperature of the transistors themselves can be read back by measuring the forward voltage drop of the base-emitter junction periodically while the collector supply is shut off.

[jt]

I've used these for a similar purpose:
http://www.vishay.com/docs/48114/_pcnm_vmn-pt0492-1606.pdf

Looks to be too larger for your application, but I see that someone already pointed you to a smaller "wrap around" resistor family.

I think my design intent may be similar to yours:

I wanted to heat an area of the PCB top-layer with the resistor populated on the bottom layer.  I made a circular 2oz copper feature on the top side of the board and split the circle into two halves with a small copper keep-out (so that they could be separate electrical nets).  I positioned the resistor on the bottom layer so that it straddled the keep-out and used a liberal amount of thermal vias so heat would be conducted from the resistor pads, through the board, to respective copper half-circle features.  Current was shunted through the resistors through "wagon wheel" traces so the heat would be isolated from the ground and power planes (I only wanted localized heating).

I didn't do a lot of empirical testing to refine the design.  Qualitatively, it lit up quickly on the thermal camera and warmed the surface as needed.

 

[Korken]

Thanks for all the suggestions!
I have ordered an test board on which I put 16 resistors on the backside for testing.

I also added a bunch of NTC thermistors here and there to measure the temperature on the resistor side and the IMU side to compare how well it agrees with the internal measurement of the IMU.
What I hope is to estimate the heat transfer's transfer function to have a model for how control signal will affect IMU temperature.
So for now I will go with a simple second order model, but this will not take into account the heat loss through the heat shield, but hopefully this will be small (and and be improved by adding some ceramic fabric as insulation).

[ESCler]

Hi Korken,

do you have any news on the temperature stabilized IMU? I am also very interested in such a design, to improve the drift.

I remember your post from here: http://forum.taulabs.org/viewtopic.php?f=22&t=200 where you wanted to compare Invensense's MPU-9250 vs ST's LSM9DS0. It was really sad that there was no conclusion at the end. Maybe you can tell us here what IMU you are using right now and if you have finished the comparison. It would be great to here from you.

Thanks.

[Refrigerator]

You might want to use a small MOSFET with a thermal pad instead of a resistor.

[Korken]

ESCler:
It worked very well with just spreading resistors on the bottom under the IMU, and then covering the resistors and IMU is a plastic can to keep wind away.
I put an NTC resistor in the resistor array to measure the temperature on the bottom resistors and then ran a PI controller there, which gets its reference from a simple P controller working on the measured temperature inside the IMU.
The exact performance gain I don't know yet though.

Yeah, that project died from too little time..
But so far I like the LSM6 series sensor, though it is made from multiple dies.
I have built a jig to measure the axis misalignment, which I have characterized the classical MPU-6000 in, and will asap characterize the LSM6DSL.