Soldering Thermal Pads

torsdag den 17. december 2020 kl. 00.05.49 UTC+1 skrev gnuarm.del...@gmail. com:

rge thermal loads. The main problem is a part that has large pads. The guy is laying it out ignoring the suggestions of the data sheet and simply put five 16 mil drill vias to the bottom layer with a lot of copper. That's abo ut like connecting with five thermal breaks, 50 mil wide and 62 mil long.

cking. I've never heard of that. Is that a thing? I've always heard that vi as simply should not be in the pad, rather ring the pad with lots of copper . Then you can get a significant number of them, adequate to connect the bo ttom layer to the top thermally.

while avoiding vias in pads might be a good idea generally, it is done all the time because sometimes that is the only option, like for example a QFN with a thermal pad

and it is commonly shown in datasheets

l.com:

large thermal loads. The main problem is a part that has large pads. The gu y is laying it out ignoring the suggestions of the data sheet and simply pu t five 16 mil drill vias to the bottom layer with a lot of copper. That's a bout like connecting with five thermal breaks, 50 mil wide and 62 mil long.

wicking. I've never heard of that. Is that a thing? I've always heard that vias simply should not be in the pad, rather ring the pad with lots of copp er. Then you can get a significant number of them, adequate to connect the bottom layer to the top thermally.

l the time because sometimes

That doesn't explain how it is done. Also, I think you are referring to v ery small parts with very tiny vias. Even then, the fact that they recomme nd the action in a data sheet doesn't mean it is a good idea. Linear Techn ology has a line of land grid array parts that are a bitch to assemble. I was at the assembly house one day and without prompting I was warned not to use the parts because of the assembly issues.

That's the only way to get the heat out of some parts. A bunch of small vias is the usual approach.

Via-in-pad is fine when either untented, or fully tented.

One side tented is a no-no: the barrel traps flux which bubbles during soldering, guaranteeing a void under the pad.

Full tenting obviously is hard to do. It's a lot of mucking about in most EDA tools, due to how layers are plotted (typically you have to custom draw your own soldermask opening). Even once done, it's not great, there isn't as much soldermask-copper bond area as there is around other footprints so it may flake off during soldering anyway.

And anyway, you haven't really done much, because the patch of soldermask will stick to flux and bubble anyway, and it's poorly conductive so it's basically a void underneath the part regardless of whether it traps gas exactly. (So, better, but not really by much.)

Gold standard is plugged or plated-shut vias. Custom process, plugged requires screening extra epoxy on the board to fill the offending holes, while masking all the others. Plated only works on certain via sizes (and if done poorly, can trap electrolyte, causing failures during soldering). The surface can be smooth, flat and plated -- ideal soldering conditions. Adds lots of labor to the process though (plugging at least).

The usual go-to is untented. Which allows solder to wick down the barrel, so, increase paste a little bit to compensate -- normally blind pads have excess solder anyway (at zero expansion, typical 0.1mm mask thickness, and typical paste solder/rosin ratio) so this isn't a problem. May be annoying for large pads that need a lot of solder anyway (not so much D/2PAKs, as say, the heatsinks often paired with them?).

Obviously if you're doing this by hand, you can inspect the joint in real time, supply enough heat to fully melt everything, and add solder if needed. Doesn't matter. This is only about reflow soldering in production.

Tim

aw

ed.

rge

ive

This is not a small part. This is a 17 x 19 mm MultiPowerSO-30 package mea ning it has three pads on the bottom, two 5x6 and one 5x10 mm. The RMS cur rent will likely be in the range of 4 amps, so a fair amount of heat, poten tially more worst case. This project is not very good at defining actual r equirements.

I've also noticed he hasn't done a great job of laying out the buck switchi ng regulator. The critical EMI generating loops are far from optimal. He routed the flyback diode over to the output connector to find a ground. No t a long way, but that route should be absolute minimum loop area. At leas t he used a $4 part instead of a highly suspect $0.50 TI part. I suggeste d the TI part and he's saying we would need to get an eval board to make su re it works.

The guy never saw an LT part he didn't like. $$$

hing regulator. The critical EMI generating loops are far from optimal. He routed the flyback diode over to the output connector to find a ground. Not a long way, but that route should be absolute minimum loop area. At least he used a $4 part instead of a highly suspect $0.50 TI part. I suggested th e TI part and he's saying we would need to get an eval board to make sure i t works.

Linear Technlogy parts are frequently extremely likeable. They don't tend to be cheap, but if you need the quality, it makes sense to pay the price.

e:

tching regulator. The critical EMI generating loops are far from optimal. H e routed the flyback diode over to the output connector to find a ground. N ot a long way, but that route should be absolute minimum loop area. At leas t he used a $4 part instead of a highly suspect $0.50 TI part. I suggested the TI part and he's saying we would need to get an eval board to make sure it works.

to be cheap, but if you need the quality, it makes sense to pay the price.

Therein lies the problem. "Quality" is not tangible and engineering is abo ut tangible issues. If you can't explain a requirement that only a $4 bout ique chip will meet that a $0.50 part can't then the $4 LT part needs to go . I brought up several of such issues weeks ago when I saw the landscape l ittered with LT including a part that was of no value providing protection against input over voltage and reverse voltage when we have no such require ments. In addition, the other parts were already rated for nearly the same voltage range other than the battery charger buck/boost which could be pro tected by a diode. When we changed the power source to one not using a bar rel jack he relented on that $4 LT part. Then for no clear reason he recen tly changed his mind on the $4 regulator when I showed him pages of similar regulators for under a buck on Digikey.

But I suppose "quality" is worth another $7.50.

Under power-pad parts that have leads all around, I use a topside copper pour inside the pins, and vias down to an inner pour or a thermal antenna on the bottom, or both. As you note, there is usually lots of solder available to slurp down into the un-tented vias, which cuts thermal resistance by a factor of two or so.

A via has less thermal resistance to a nearby power plane, than all the way to the bottom. Layer 2 ground may be just a few mils down.

I theorize that the best pattern is lots of small vias zig-zagged around the edges of the power pad.

But I prefer dpaks or sot-89s whenever possible. Parts with underside power pads, or BGAs, or LGAs, are awful to probe or rework.

ote:

witching regulator. The critical EMI generating loops are far from optimal. He routed the flyback diode over to the output connector to find a ground. Not a long way, but that route should be absolute minimum loop area. At le ast he used a $4 part instead of a highly suspect $0.50 TI part. I suggeste d the TI part and he's saying we would need to get an eval board to make su re it works.

nd to be cheap, but if you need the quality, it makes sense to pay the pric e.

out tangible issues.

With the parts I used, it was smaller offsets, that were right there in the data sheet. Quality - in that sense - is perfectly tangible.

that a $0.50 part can't then the $4 LT part needs to go.

This can depend on who you are explaining it too. Some bosses seem to find it very hard to understand how more expensive chip could possibly be more c ost effective. On one occasion I had to knock up four separate circuit diag rams, using one, two, three and four op amps before my boss could be persua de that the four op amp version was the best choice (and it was the one tha t ended up getting manufactured). The single op amp version needed a very high gain op amp, but Eurotherm had signed up to buy hundreds of thousands of them early on, so the one op amp solution was actually the cheapest - bu t not all that good.

ttered with LT including a part that was of no value providing protection a gainst input over voltage and reverse voltage when we have no such requirem ents. In addition, the other parts were already rated for nearly the same v oltage range other than the battery charger buck/boost which could be prote cted by a diode. When we changed the power source to one not using a barrel jack he relented on that $4 LT part. Then for no clear reason he recently changed his mind on the $4 regulator when I showed him pages of similar reg ulators for under a buck on Digikey.

Can be. Service calls can be very expensive. Not having a requirement to pr otect against over-voltage and reverse voltage can sometimes reflect a lack of relevant experience with ham-fisted installers or users. Some graduate students have a habit of taking everything apart, and putting it back toget her wrong, and then bitching that the device isn't doing what was promised. That was an interesting service call, and the student then bad-mouthed us at couple of conferences that we got to hear about.

switching regulator. The critical EMI generating loops are far from optima l. He routed the flyback diode over to the output connector to find a groun d. Not a long way, but that route should be absolute minimum loop area. At least he used a $4 part instead of a highly suspect $0.50 TI part. I sugges ted the TI part and he's saying we would need to get an eval board to make sure it works.

tend to be cheap, but if you need the quality, it makes sense to pay the pr ice.

about tangible issues.

he data sheet. Quality - in that sense - is perfectly tangible.

That is my point. If you need the "quality" then fine, pay the premium. T his guy reaches for LT parts without even knowing what his requirements are . He just relented on a $4 LT buck switcher when I looked up the cheapest buck switcher in Digikey that vaguely fit the outline. But he doesn't want to use it because he can't simulate it in LTspice! We need something that will deliver about 7 volts from a 10-13 volt supply at 300 mA. That's not rocket science. I responded with a link to a 7808 and a heat sink. It wo uld be better to not use the switcher anyway. There are sensitive circuits on the main board and we will need to take extra care to not get switching noise in them. At least we have some physical separation.

et that a $0.50 part can't then the $4 LT part needs to go.

d it very hard to understand how more expensive chip could possibly be more cost effective. On one occasion I had to knock up four separate circuit di agrams, using one, two, three and four op amps before my boss could be pers uade that the four op amp version was the best choice (and it was the one t hat ended up getting manufactured). The single op amp version needed a very high gain op amp, but Eurotherm had signed up to buy hundreds of thousands of them early on, so the one op amp solution was actually the cheapest - b ut not all that good.

littered with LT including a part that was of no value providing protection against input over voltage and reverse voltage when we have no such requir ements. In addition, the other parts were already rated for nearly the same voltage range other than the battery charger buck/boost which could be pro tected by a diode. When we changed the power source to one not using a barr el jack he relented on that $4 LT part. Then for no clear reason he recentl y changed his mind on the $4 regulator when I showed him pages of similar r egulators for under a buck on Digikey.

protect against over-voltage and reverse voltage can sometimes reflect a la ck of relevant experience with ham-fisted installers or users. Some graduat e students have a habit of taking everything apart, and putting it back tog ether wrong, and then bitching that the device isn't doing what was promise d. That was an interesting service call, and the student then bad-mouthed u s at couple of conferences that we got to hear about.

Did you snip part of my post? I know that somewhere I've explained that al l the circuits exposed to the input voltage are already tolerant of large o ver voltage. We don't want to take the voltage hit on using a diode to pro tect from reverse voltage, but that's also ok since all the devices are als o reverse voltage tolerant other than the buck/boost battery charger which CAN take the diode hit. Hell, he's got a damn diode in the path AFTER the sense resistors that set the charging voltage. I wonder if he factored tha t into the voltage setting. Unfortunately the diode will drop a voltage wi th a temperature factor and the damn battery is already a temperature sensi tive device. He added a temperature sensor to adjust the charging voltage, bit it will be almost two feet from the battery and on a board with virtua lly all the hot power components and next to the very hot motor.

I did manage to get them to use a larger board today. I can't see the deta il of the mechanical design. The drawings aren't very good unless you use a 3D viewer. So the available board size is either 100mm sq right by the m otor or 62x162mm along the side a bit away from the motor. There are two t iny fans blowing on the motor, but I can't say if they will be enough to ac tually cool anything. 2 inch fans don't blow much.

ck switching regulator. The critical EMI generating loops are far from opti mal. He routed the flyback diode over to the output connector to find a gro und. Not a long way, but that route should be absolute minimum loop area. A t least he used a $4 part instead of a highly suspect $0.50 TI part. I sugg ested the TI part and he's saying we would need to get an eval board to mak e sure it works.

t tend to be cheap, but if you need the quality, it makes sense to pay the price.

s about tangible issues.

the data sheet. Quality - in that sense - is perfectly tangible.

his guy reaches for LT parts without even knowing what his requirements are . He just relented on a $4 LT buck switcher when I looked up the cheapest b uck switcher in Digikey that vaguely fit the outline. But he doesn't want t o use it because he can't simulate it in LTspice! We need something that wi ll deliver about 7 volts from a 10-13 volt supply at 300 mA. That's not roc ket science. I responded with a link to a 7808 and a heat sink. It would be better to not use the switcher anyway. There are sensitive circuits on the main board and we will need to take extra care to not get switching noise in them. At least we have some physical separation.

meet that a $0.50 part can't then the $4 LT part needs to go.

ind it very hard to understand how more expensive chip could possibly be mo re cost effective. On one occasion I had to knock up four separate circuit diagrams, using one, two, three and four op amps before my boss could be pe rsuade that the four op amp version was the best choice (and it was the one that ended up getting manufactured). The single op amp version needed a ve ry high gain op amp, but Eurotherm had signed up to buy hundreds of thousan ds of them early on, so the one op amp solution was actually the cheapest - but not all that good.

e littered with LT including a part that was of no value providing protecti on against input over voltage and reverse voltage when we have no such requ irements. In addition, the other parts were already rated for nearly the sa me voltage range other than the battery charger buck/boost which could be p rotected by a diode. When we changed the power source to one not using a ba rrel jack he relented on that $4 LT part. Then for no clear reason he recen tly changed his mind on the $4 regulator when I showed him pages of similar regulators for under a buck on Digikey.

o protect against over-voltage and reverse voltage can sometimes reflect a lack of relevant experience with ham-fisted installers or users. Some gradu ate students have a habit of taking everything apart, and putting it back t ogether wrong, and then bitching that the device isn't doing what was promi sed. That was an interesting service call, and the student then bad-mouthed us at couple of conferences that we got to hear about.

ll the circuits exposed to the input voltage are already tolerant of large over voltage.

"Large over voltage" is a bit unspecific.

verse voltage, but that's also ok since all the devices are also reverse vo ltage tolerant other than the buck/boost battery charger which CAN take the diode hit. Hell, he's got a damn diode in the path AFTER the sense resisto rs that set the charging voltage. I wonder if he factored that into the vol tage setting. Unfortunately the diode will drop a voltage with a temperatur e factor and the damn battery is already a temperature sensitive device. He added a temperature sensor to adjust the charging voltage, bit it will be almost two feet from the battery and on a board with virtually all the hot power components and next to the very hot motor.

It's the temperature of the battery - and in fact the guts of the battery - that influences the voltage you see at it's terminals.

A remote temperature sensor is going to be better than no temperature senso r, but you are talking about a ventilator that is presumably going to be us ed in an air-conditioned hospital, so it does sound a bit odd.

ail of the mechanical design. The drawings aren't very good unless you use a 3D viewer. So the available board size is either 100mm sq right by the mo tor or 62x162mm along the side a bit away from the motor. There are two tin y fans blowing on the motor, but I can't say if they will be enough to actu ally cool anything. 2 inch fans don't blow much.

Presumably there's a volume of air per minute figure for the fans. If you k now the heat coming out the motor, you should be able to work out how much hotter that volume of air will be once it has soaked up that much heat. The universal gas law lets you work it out but there are a lot of conversion f actors to work out along the way.

Going from cubic feet of air at room temperature and pressure to moles of air - which do have a well-defined heat capacity - is tedious.

buck switching regulator. The critical EMI generating loops are far from op timal. He routed the flyback diode over to the output connector to find a g round. Not a long way, but that route should be absolute minimum loop area. At least he used a $4 part instead of a highly suspect $0.50 TI part. I su ggested the TI part and he's saying we would need to get an eval board to m ake sure it works.

n't tend to be cheap, but if you need the quality, it makes sense to pay th e price.

is about tangible issues.

in the data sheet. Quality - in that sense - is perfectly tangible.

This guy reaches for LT parts without even knowing what his requirements a re. He just relented on a $4 LT buck switcher when I looked up the cheapest buck switcher in Digikey that vaguely fit the outline. But he doesn't want to use it because he can't simulate it in LTspice! We need something that will deliver about 7 volts from a 10-13 volt supply at 300 mA. That's not r ocket science. I responded with a link to a 7808 and a heat sink. It would be better to not use the switcher anyway. There are sensitive circuits on t he main board and we will need to take extra care to not get switching nois e in them. At least we have some physical separation.

l meet that a $0.50 part can't then the $4 LT part needs to go.

find it very hard to understand how more expensive chip could possibly be more cost effective. On one occasion I had to knock up four separate circui t diagrams, using one, two, three and four op amps before my boss could be persuade that the four op amp version was the best choice (and it was the o ne that ended up getting manufactured). The single op amp version needed a very high gain op amp, but Eurotherm had signed up to buy hundreds of thous ands of them early on, so the one op amp solution was actually the cheapest - but not all that good.

ape littered with LT including a part that was of no value providing protec tion against input over voltage and reverse voltage when we have no such re quirements. In addition, the other parts were already rated for nearly the same voltage range other than the battery charger buck/boost which could be protected by a diode. When we changed the power source to one not using a barrel jack he relented on that $4 LT part. Then for no clear reason he rec ently changed his mind on the $4 regulator when I showed him pages of simil ar regulators for under a buck on Digikey.

to protect against over-voltage and reverse voltage can sometimes reflect a lack of relevant experience with ham-fisted installers or users. Some gra duate students have a habit of taking everything apart, and putting it back together wrong, and then bitching that the device isn't doing what was pro mised. That was an interesting service call, and the student then bad-mouth ed us at couple of conferences that we got to hear about.

all the circuits exposed to the input voltage are already tolerant of larg e over voltage.

More specific than "protect against over-voltage and reverse voltage".

reverse voltage, but that's also ok since all the devices are also reverse voltage tolerant other than the buck/boost battery charger which CAN take t he diode hit. Hell, he's got a damn diode in the path AFTER the sense resis tors that set the charging voltage. I wonder if he factored that into the v oltage setting. Unfortunately the diode will drop a voltage with a temperat ure factor and the damn battery is already a temperature sensitive device. He added a temperature sensor to adjust the charging voltage, bit it will b e almost two feet from the battery and on a board with virtually all the ho t power components and next to the very hot motor.

- that influences the voltage you see at it's terminals.

Yes, that much is very clear.

sor, but you are talking about a ventilator that is presumably going to be used in an air-conditioned hospital, so it does sound a bit odd.

There is no such presumption of air conditioning. The only assumption is t hat it will be close to the patient so not exposed to extremes. My main co ncern is that the board itself will be rather warmer than the surroundings and so measuring a temperature on the board will create error rather than r educe it.

etail of the mechanical design. The drawings aren't very good unless you us e a 3D viewer. So the available board size is either 100mm sq right by the motor or 62x162mm along the side a bit away from the motor. There are two t iny fans blowing on the motor, but I can't say if they will be enough to ac tually cool anything. 2 inch fans don't blow much.

know the heat coming out the motor, you should be able to work out how muc h hotter that volume of air will be once it has soaked up that much heat. T he universal gas law lets you work it out but there are a lot of conversion factors to work out along the way.

air - which do have a well-defined heat capacity - is tedious.

Don't tell me. I'm the guy responsible for calculating the respiratory vol ume based on the pressure drop across an orifice. Actually, I think that o ne has been put to bed. I had to do a Bill Sloman and go through the work of the guy designing the flow sensor and do the math to show it would not p rovide enough accuracy at low volumes. Mostly it was a matter of the press ure sensor error being a function of full scale pressure rather than the re ading. Now they want to consider a pressure sensor with an error spec on t he measured value, but that's still not realistic at some lower pressure. There's always an offset even with a calibration. I designing the calculat or in the FPGA, so I'm doing the math for a commercial flow sensor that rep orts flow rate directly in SLM. They can make the in-house flow sensor a s eparate project. With the commercial sensor I still need to factor in the absolute pressure, temperature and oxygen content to get actual LPM, so no picnic.